In May I published the case for Xunlei (NASDAQ: XNET) at $6.33: a 78% discount to net asset value, anchored by a 7.8% stake in Insta360, the Shanghai-listed camera maker, worth two and a half times Xunlei’s own market capitalisation, and by a hard date, the 11 June expiry of the lock-up on that stake. The title promised four weeks to close. The date has passed, the shares mark $5.64, the position is down 10.9%, and the discount has not closed [1][6]. The position stays on our list of trades, on humbler terms than the ones I published.

The discount itself behaved: 78% at publication, roughly 75% today. What fell was the net asset value. Insta360 has dropped 36% since the May mark, from 214 yuan to 135.93 [2], and Xunlei’s NAV per share has fallen by about a fifth [7]. The trade lost money the one way the piece spent least time pricing: the catalyst arrived and damaged its own collateral.

What happened

The 11 June expiry freed Xunlei to sell, and it freed everyone else: 56.5% of Insta360’s share register became tradable that day, roughly 37 billion yuan of stock across 28 pre-IPO investors, of which Xunlei’s holding is one line among many [3]. Insta360 was already de-rating from its September peak of 377.77 yuan, a price war with DJI gave the de-rating a fundamental engine through June [4], and the unlock supplied the paper. The stock set a 52-week low of 130.88 yuan on 29 June and closed at 135.93 on 3 July [2]. The May piece named 130 yuan as the level that changes the arithmetic. The mark sits two bad sessions above it.

Management, handed the moment the thesis required, chose patience. On the first-quarter call, two weeks before the unlock, there was no dividend, no timetable for the stake, and one load-bearing adverb: Xunlei would “gradually seek to adjust” its holdings toward a stated ceiling of investment securities at 45% of total assets, excluding cash and government securities [5]. No sale of Insta360 shares has been disclosed since the lock-up lifted. What has been disclosed is a repurchase: on 26 June the board approved a programme to buy back up to $20m of Xunlei’s own shares over the twelve months from 1 July, funded from the cash balance [8].

What I got wrong

The timing failed on its own stated deadline. The body of the piece described an 18-to-24-month monetisation window, but the title said four weeks, and the calendar was wrong: the discount-narrowing flow I expected into the date never appeared, and the shares closed as low as $4.66 before the unlock even arrived [6]. (The title, in hindsight, was the most falsifiable sentence in the piece.)

I priced what the unlock permitted our company to do, and missed what it permitted the other 27 unlocked holders to do. The event that made Xunlei’s exit possible cut the value of the thing being exited by a third, a scenario the published piece did not carry.

The catalyst rested on willingness. The Investment Company Act gives this management a direction and no timetable. The lesson filed in our mistake log: a catalyst date earns the name only when someone is obliged to act on it.

The published return arithmetic inherited the flaw. A base case of 55% to 75% a year was that willingness with a spreadsheet attached; the scenario that assumed the least of management, the bear case, is the one the market delivered.

The position, remarked

On the piece’s own methodology, at the 3 July close [7]: the Insta360 stake, 31.28m shares at 135.93 yuan and 6.77 to the dollar, is worth $628m, against $986m at publication. Net cash is roughly $234m, slightly higher than published. The operating business keeps its trough range of $410m to $745m, midpoint $580m: first-quarter revenue grew 54.1% and subscriptions 26.2%, though management guided to a modest slowdown ahead [5]. Net asset value is therefore about $1.44bn, or $22.65 per ADS, against $5.64 on the screen. The discount is 75%.

Two published facts survive the markdown. Net cash plus the trough value of the operating business comes to about $814m, 2.3 times the $359m market capitalisation, so the camera stake still comes free even marked a third below where I first called it free. And the regulatory pressure has not moved: at current marks investment securities sit near 74% of the relevant asset base against management’s stated 45% ceiling, and the sale required to reach that ceiling is roughly $446m, one and a quarter times the market value of the entire company [5][7].

Where I stand

The trade I published is dead, and pretending the horizon was always longer would be worse than the loss. What remains is a different position: a deep-discount holding with no clock, worth keeping only while that arithmetic holds, and re-examined in public whenever it is tested.

The case for closing is real: book NAV fell 25% in the two quarters to March, Insta360 is fighting a price war at its 52-week low, management’s pace is glacial, and wide discounts persist for years when nothing forces them shut. The case for holding is that at $5.64 the market pays less than the cash and the subscription business alone, the supply event has now happened rather than loomed, Insta360 sits at the May piece’s own bear-case mark, and the Investment Company Act still points in one direction. Selling at the bear-case mark would mean selling the cash and the subscription business below their own worth to be rid of a stake the price already writes off. The honest cost of staying is that the unlock supply has not cleared; it is now permanent.

I am holding, on those re-underwritten terms, with one mechanical exit: a next capital-allocation move of size that is an acquisition rather than a sale or a distribution. A decisive break below 130 yuan on Insta360 is a tripwire of a different kind. It sends the position back to the workbench for a fresh re-underwrite, published here, rather than forcing a sale on its own; a lower mark changes the arithmetic, and it is the arithmetic that decides. The first disclosed sale of Insta360 shares is what upgrades this from a discount we hope closes to one being closed. The watch items posted under the article in June stand, and the first has partially fired: the $20m repurchase is a real, if modest, step toward returning capital, while the pace of the 45% unwind and the next mark on the stake remain open.

References

[1] Stray Narratives Issue 11, An 80% Discount to NAV, Four Weeks to Close, published 14 May 2026 (straynarratives.com/p/an-80-discount-to-nav-four-weeks). Entry $6.33; NAV $28.57 per ADS on 62.86m ADS; net cash $230m; Insta360 stake $986m at the published 13 May mark of 214 yuan; operating business $410m to $745m at trough multiples, midpoint $580m.

[2] Arashi Vision Inc. (Insta360, SHA: 688775) closed at 135.93 yuan on 3 July 2026; 52-week low 130.88 yuan set 29 June 2026; 52-week high 377.77 yuan (September 2025). Stock Analysis and FT market data. Issue 11 published mark: 214 yuan, 13 May 2026, the basis on which the published $986m stake value was computed; the official 13 May close was 220.44 yuan.

[3] Pandaily, Insta360 at a Crossroads: 37 Billion Yuan Lock-Up Expiry Collides with DJI Rivalry, June 2026. 56.5% of shares, worth approximately 37 billion yuan, tradable from 11 June 2026; 28 pre-IPO financial investors; Xunlei at 7.84%.

[4] Caixin Global, DJI, Insta360 Lock Horns in Camera Pricing Standoff, 4 June 2026.

[5] Xunlei Limited, Unaudited Financial Results for the First Quarter Ended March 31, 2026, Form 6-K Exhibit 99.1, filed 28 May 2026, and the accompanying earnings call. Revenue +54.1% year on year to $98.6m; subscriptions +26.2%; cash, equivalents and short-term investments $303.6m; long-term investments $888.6m; total assets $1,415m. Management commentary on holding investment securities below 45% of total assets, excluding cash and government securities, and on adjusting holdings “gradually”; US Investment Company Act of 1940, s3(a)(1)(C), sets the statutory 40% test.

[6] XNET.US daily closes, EODHD: $6.33 on 13 May 2026 (entry); low close $4.66 on 5 June; $5.64 on 2 July 2026, the last session before the US holiday. Move from entry: minus 10.9%.

[7] Refreshed NAV on the Issue 11 methodology: 31.28m Arashi Vision shares × 135.93 yuan ÷ 6.7702 USDCNY = $628m. Net cash $234m ($303.6m cash, equivalents and short-term investments less $69.6m of bank borrowings, Q1 2026 balance sheet). Operating business at the published trough midpoint, $580m. NAV $1,442m ÷ 63.65m ADSs = $22.65 per ADS; market capitalisation $359m at $5.64 on 63.65m ADSs (318,268,921 shares outstanding at 31 March 2026; one ADS is five shares). Investment-securities ratio: $628m ÷ ($1,154m remarked total assets, the Q1 total of $1,415m with the $888.6m long-term investments line remarked to $628m, less $304m cash items) ≈ 74%; sale required to reach the 45% ceiling ≈ $446m.

[8] Xunlei Limited, 2026 Share Repurchase Program, Form 6-K filed 26 June 2026: up to US$20m of ADSs or common shares over the twelve months from 1 July 2026, funded from the cash balance.

If you prefer to listen rather than read, here is a podcast interview version of this same subject:

The American medical-device sector trades at its smallest premium to the wider market since the 2008 financial crisis, and on next year’s earnings it has slipped to an outright discount for the first time in living memory [1]. This is a group of companies that grew revenue from roughly seventy billion dollars to a hundred and seventy-six billion over the past decade, earns operating margins near twenty percent, and sells into the most dependable demand curve in the economy, which is the ageing of the rich world [8]. We are adding a basket of its cheapest quality names, equal-weighted, to the list of trades, because the market has priced a collapse that the businesses themselves refuse to deliver.

Two fears did the damage. The first is that weight-loss drugs will hollow out demand for everything from artificial knees to heart valves to glucose monitors. The second is that the sector’s profitability has quietly broken. Both are checkable against the operating numbers, and the numbers disagree with the price: the drugs are expanding several of these markets rather than shrinking them, and the margin dip everyone extrapolated was a post-pandemic inventory hangover that has already reversed.

How a good business got cheap

The de-rating is real and worth respecting. Large-cap medical devices spent most of the past decade carrying a four-to-six turn premium to the S&P 500 on forward earnings, the going rate for steady growth and recession resistance. That premium has compressed to roughly nothing, and the median name now trades near fifteen times forward earnings against an index above twenty-two [1]. The last time the sector was this cheap relative to the market, the world was busy recapitalising its banks.

Two things drove it. Through the pandemic, hospitals over-ordered devices and consumables, then spent 2022 and 2023 working that inventory back down, so reported revenue growth stalled to under two percent before recovering [2]. And from 2023 onward, every earnings call carried the same question, in one form or another: what does Ozempic do to your volumes. The market has spent two years concluding that weight-loss drugs will empty the operating theatres, and has priced the sector accordingly, which would be the correct response if the operating theatres were emptying.

The engine did not break

Cheap is only interesting if the business behind it is sound, and the ten-year record says it is. Set the pandemic distortions aside and the profitability picture is a cyclical dip that has already recovered: operating and free-cash-flow margins now sit at or above where they were a decade ago, EBITDA margins are at a ten-year high near twenty-six percent, cash conversion has improved, and research spending has held steady at about seven and a half percent of revenue the entire time [2]. The one genuine soft spot is gross margin, which has slipped two to three points over the decade on mix and input costs, an erosion the sector has more than recovered further down the income statement. A business that keeps investing the same share of revenue in its own future while its cash margins climb has not weakened. It has stopped commanding the market’s attention.

What the weight-loss trade got wrong

The GLP-1 fear deserves a real answer rather than a wave of the hand, because the logic is not absurd. Thinner, healthier people should, in time, need fewer of the interventions that obesity drives. The difficulty for the bears is that the operating data has begun to arrive, and for most of the sector it points the other way.

Sleep apnoea is the purest expression of the bear case, and its leading device-maker was punished hardest. The company ran its own analysis across roughly six hundred and sixty thousand patients and found that those prescribed a weight-loss drug were about eleven percentage points more likely to begin therapy, and its device volumes grew double digits straight through the scare [3]. The clinical trial that was supposed to prove the drugs cure apnoea instead found that most patients still had disease serious enough to treat [4]. The drugs are pulling people into the system, getting them diagnosed, and leaving the majority still needing the device.

The pattern repeats across the basket. Continuous glucose monitors were meant to lose their diabetic customers; instead more than two in five users of the leading monitor are now also taking a weight-loss drug and using the sensor to manage the transition [3]. Orthopaedics was meant to lose its knee and hip replacements; in practice surgeons routinely require patients to lose weight before they will operate, so a drug that reliably reduces body mass widens the pool of people eligible for surgery.

There is one place the bears are right. Bariatric surgery, the operations designed specifically to treat obesity, faces a genuine and management-acknowledged decline as the drugs substitute for the scalpel. For the diversified surgical-robotics franchise in our basket that is a small and recoverable share of a far larger procedure base, but it is the one unambiguous loss in the group, and pretending otherwise would be the sort of analysis that gets quoted back to you in eighteen months.

The asset nobody is pricing

There is a longer-dated reason to own these companies that the multiple ignores completely. The device-makers have spent two decades accumulating the one thing the artificial-intelligence boom has made suddenly precious: vast, proprietary, regulated clinical datasets that cannot be scraped from the open web or generated synthetically. The surgical-robotics leader holds the logged record of more than twenty million procedures. The connected-breathing-device maker has billions of nights of physiological data with the contractual right to use it. The diabetes and diagnostics franchises hold continuous, outcome-linked readings at a scale no model-builder can assemble, because a regulator will not accept invented data and a hospital cannot hand its patients’ records to a chatbot. Several of these firms already sell medical AI products the regulator has cleared, built on exactly this data [5]. A market paying extraordinary multiples for the companies that own AI models is paying nothing for the companies that own the one input those models cannot manufacture. We treat this as the free option on top of the value rather than the reason to buy.

The moat has one genuine challenger, and it is not a model-builder. Chinese device-makers are taking share in the high-volume, lower-margin corner of the industry, and they compete on price rather than on data. State procurement has cut the price of routine in-vitro diagnostic tests by as much as three-quarters, and China’s domestic champion in patient monitoring and bench diagnostics is moving upstream into the reagents and instruments Western firms long treated as captive; in gene sequencing a Chinese pair has already overtaken the former Western leader inside China [6]. The pressure falls hardest on the commodity-diagnostics and imaging lines, where the product is a consumable sold to a price-driven state buyer, and lightest on the franchises whose advantage is a regulated, outcome-linked dataset locked inside an installed base: a robotic-surgery record, years of physiological monitoring, a continuous glucose feed. The honest split is that the diagnostics and imaging names in the basket carry a structural margin headwind that quality does not fully offset, while the device-data franchises that anchor the thesis are the ones the substitution reaches last. It argues for leaning on the moat names inside the basket, not against owning it.

Why it re-rates

Cheap and sound can stay cheap for a long time, so the question is what closes the gap. The most concrete answer is that the buyers with the most information and the most capital are already taking these assets off the market outright. 2025 was the busiest year for medical-device dealmaking in more than a decade, including a private-equity take-private of a diagnostics company at roughly eighteen billion dollars and a twenty-one billion dollar acquisition by one of our own basket members [7]. When trade buyers and private equity pay cash for whole companies above where the public market marks them, they set a floor under the equity that sentiment struggles to break through. The slower catalyst is the ordinary one: as the inventory hangover fully laps and earnings revisions turn up, a sector at a discount to the market with a thirty-year demographic tailwind tends to be repriced by the same flows that left it. Those flows are only now beginning to reach devices: the group has turned up sharply off its June low in the past week, led by the diabetes and diagnostics names, though a few days off a six-month bottom is not yet a trend, and that hesitation is the honest timing risk in the trade.

Subscribe now

The position

We are adding an equal-weighted basket of ten names to the list of trades at the open: ResMed, Medtronic, Abbott, GE HealthCare, Boston Scientific, Stryker, Zimmer Biomet, Becton Dickinson, Globus Medical and Dexcom. Six are quality compounders growing earnings and cash flow into the demographic tailwind; three are deep-value or special-situation names where the cheapness and a catalyst do the work; one, Dexcom, is the clearest single embodiment of the thesis, a proprietary-data franchise that the weight-loss drugs turn out to help. Equal-weighted, the basket trades near sixteen times forward earnings, well below both the market and the cap-weighted sector index, which is dragged up by paying thirty to forty times for the two most expensive names in the group. We hold the data moat through its cheaper carriers rather than overpay for the marquee one.

For readers who would rather own the theme in a single line, the iShares U.S. Medical Devices ETF (ticker IHI) delivers the same exposure, with the caveat that its cap weighting makes you pay up for exactly the expensive names the basket is built to avoid. The basket is the better expression of the view, the ETF the more convenient one. A two-line case for each holding sits at the foot of this note.

Stray Narratives is free and carries no advertising, so it travels only as far as its readers send it. If this was worth your time, a like, a restack, or a forward to one colleague who should see it is the whole of our distribution, and what keeps it free.

The position implication

You are buying a sector at a discount to the market for the first time since 2008, with margins at decade highs, a demand curve that compounds for thirty years, acquirers setting a cash floor beneath it, and a proprietary-data moat the AI trade is pricing at zero. What that pays, on a three-year view and entered at the basket’s sixteen-times multiple, is a base case of roughly seventeen percent a year: about half the discount to the market closes, earnings compound near nine percent, and the dividend adds a point. If the discount shuts to parity and earnings run faster, the basket returns about twenty-six percent a year. If it never closes at all and earnings only grind, the basket still returns about six, because the growth and the yield carry it without the re-rating. The re-rating is the option you are handed for free.

That asymmetry is the reason to own it now: the discount exists because the market paid in advance for a catastrophe, and the catastrophe is being refuted in the operating numbers one quarter at a time. The repricing does not require the businesses to surprise anyone. It only requires the absence of the disaster that is already in the price.

References

[1] Forward and historical relative-valuation work on US large-cap medical devices (median next-twelve-month P/E versus the S&P 500 and the healthcare sector), from a sell-side equity research desk, May 2026; cross-checked against a major asset manager’s published note placing healthcare at its lowest relative valuation since 2008. Our own basket multiples are computed from company filings.

[2] Stray Narratives analysis of the ten largest US medical-device companies’ annual filings, 2015 to 2025 (revenue, gross/operating/EBITDA/free-cash-flow margins, research and development, return on equity), aggregated sum-of-parts.

[3] ResMed published analyses of GLP-1 prescription data and CPAP initiation/adherence (IQVIA claims dataset, ~660,000 patients), and company device-volume disclosures, 2025-2026; continuous-glucose-monitor co-prescription figures from the leading manufacturer’s disclosures.

[4] Eli Lilly SURMOUNT-OSA trial results and the FDA approval of tirzepatide for obstructive sleep apnoea, December 2024.

[5] FDA clearances of medical AI/machine-learning device software across the named manufacturers (FDA AI/ML-enabled device list, public).

[6] Chinese medtech competition and pricing: China’s centralised “volume-based procurement” of in-vitro diagnostics (provincial tender data via the National Healthcare Security Administration), the domestic patient-monitoring and IVD leader’s vertical-integration into reagents/raw materials, and the displacement of the incumbent gene-sequencing supplier inside China by a domestic pair (company filings and trade-press deal data; surfaced by a sell-side note, underlying primaries cited).

[7] Medical-device M&A activity, 2025: aggregate deal value and the take-private of a diagnostics company and a large strategic acquisition cited as illustrative (public deal announcements).

[8] US Census Bureau and UN Department of Economic and Social Affairs population projections (US 65+ population toward 71 million by 2030; global 65+ toward ~1.5 billion by 2050).

The basket, name by name

• ResMed (RMD). The cleanest rebuttal to the GLP-1 bear: its own claims data shows patients on weight-loss drugs are more likely to start CPAP, not less, while device volumes still grew double digits. At 16x forward earnings with operating margins above 30 percent, the market is paying late-cycle prices for a structural grower.

• Medtronic (MDT). The lowest-multiple way to own the data moat: several FDA-cleared AI products already run on its proprietary device data, and it converts roughly five billion dollars of free cash flow a year. At 13x forward earnings it is priced as a no-growth utility while revenue has returned to mid-single-digit growth.

• Abbott (ABT). The diversified ballast, spanning glucose monitors, diagnostics and devices, with free cash flow up roughly half over three years to more than seven billion dollars. Its Libre franchise is a GLP-1 beneficiary, as drug users adopt monitoring rather than abandon it.

• GE HealthCare (GEHC). The cheapest large-cap imaging franchise at 13x forward earnings, and the most prolific filer of FDA-cleared AI in the group. Growth is slow, so the case rests on the re-rating of a misjudged spin-off rather than on accelerating fundamentals.

• Boston Scientific (BSX). The Farapulse cardiac-ablation cycle is still driving low-double-digit revenue growth and steady share gains, with free cash flow roughly doubled over three years. At around 13x forward earnings it is the cheapest of the growth names, partly because the stock has sold off sharply in recent weeks on a flagged stagnation in its Watchman left-atrial-appendage line, a setback in a different franchise from the Farapulse ablation business that underpins the growth case.

• Stryker (SYK). Orthopaedic robotics with a genuine data moat: its Mako system has logged more than a million cases a rival cannot replicate, and weight-loss drugs help by clearing the BMI thresholds that gate joint replacement. The 21x multiple is the one premium we pay for quality and moat together.

• Zimmer Biomet (ZBH). The deepest value at under 11x forward earnings, a mean-reversion bet on an orthopaedics recovery rather than a growth story. The same GLP-1 dynamic that worries the bears makes more patients eligible for the hips and knees Zimmer sells.

• Becton Dickinson (BDX). The special situation: 11x forward earnings against a company pursuing a separation that could surface value the conglomerate structure hides. It also carries the most debt in the basket, so this is the position underwritten on the catalyst rather than the margins.

• Globus Medical (GMED). The quality mid-cap: a net-cash balance sheet, a double-digit return on equity, and operating margins climbing back toward the high teens as the NuVasive merger integrates. Small enough to stay a credible acquisition target in a consolidating sector.

• Dexcom (DXCM). The thesis in a single name: a proprietary glucose-data moat, a verified GLP-1 tailwind as drug users adopt monitoring, and free cash flow more than doubled to over a billion dollars. The 27x multiple is the basket’s richest, and the risk to watch is the price war with Abbott, not the drugs.

A crude tanker spends most of its working life earning a return that would embarrass a savings account. Then, for a few months every decade, a chokepoint makes the evening news, crude reroutes the long way round, and the holder books years of profits in a single quarter.

That window is the trade I put on in March. I called it the Wrong Map: a basket of three tanker equities, DHT Holdings, Frontline and International Seaways, equal-weighted, bought as a direct wager on Middle Eastern disruption. I closed it this week with a nice profit:

The reason to own these names was the disruption, and the disruption is resolving. The United States and Iran have signed a memorandum of understanding, the Strait of Hormuz is reopening, and a reopened strait is, with a lag, poor news for ton-miles: crude takes the direct route again, voyages shorten, idle capacity returns, and freight reverts toward breakeven. The war premium I was renting is being handed back.

The arithmetic

When conflict in the Gulf forces crude onto longer routes, ton-miles, the volume of oil multiplied by the distance it travels, rise, and tanker spot rates rise with them. At its early-March peak the Baltic benchmark for the Gulf-to-China route reached roughly $424,000 a day against an operating breakeven near $22,000, the spread that produces the best quarter of a shipping cycle and the record dividends that came with it. It had eased toward $100,000 by mid-June, and the congestion of the reopening keeps it bid for now, but the direction is no longer higher.

Why I am not holding out for the last dollar

Tanker equities have the awkward property of looking cheapest at the exact moment they are about to earn the least. They are priced off the prevailing spot rate, so at the crest of a disruption their trailing earnings look spectacular and the shares screen as the cheapest assets on the market, a multiple usually quoted without the word “trailing” attached. That multiple is a mirage built from a single, unrepeatable quarter, and when freight reverts the earnings behind it leave with the peace. I would rather hand the optically cheap multiple to whoever wants to own the normalisation than sell it to them a month later at a lower price.

There is a genuine floor under freight. Sanctioned Russian barrels, an expanding shadow fleet and structurally longer average hauls should keep ton-miles above the pre-war normal even after the queue clears, which is why the exit may be too early. But a floor a few thousand dollars above breakeven is a rounding error against a rate that spiked to nearly twenty times breakeven. It argues for owning these names at a normalised rate on a quiet day, not for riding them through the risk of a round trip.

Where I stand

I have now closed the Wrong Map tanker basket and all three legs finished in the black, including DHT, which spent most of the trade as the laggard before the reopening congestion firmed rates into June and pulled it level with the others. The other leg of the Wrong Map, the position in rates, runs on a different mechanism and a much longer clock, and it stays on; I will however restructure that trade in the coming weeks to make sure it embeds enough time horizon to materialise.

Another free article, so please like & share if you enjoy it!

In the Central Valley of California, farmers are ripping more than 210,000 acres of almond trees out of the dirt.

They aren’t doing this to rotate crops. They are doing it because the maths of the last decade has completely broken. The University of California, Davis recently ran the numbers on planting a new almond orchard: factoring in land, trees, water, ten years of capex, and debt service, a debt-financed grower now needs roughly twenty years to break even at today’s price deck near US$2.70 a pound, and about twelve years even with no land debt to service [1].

Nobody invests with a twenty-year breakeven. Twenty years is not an investment horizon; it is a sentence. The margin of safety is gone, marginal growers are walking away, and the supply response that historically kills almond bull markets is dead.

An almond tree bears its first crop in year three, reaches full yield around year seven, and produces for roughly twenty-five years before the orchard is replaced. Supply is set the better part of a decade in advance, and it cuts both ways: the planting that should feed 2030 is not going into the ground, and the trees being pulled now will not come back at these economics.

But while California capitulates, the cycle break has opened an asymmetric mispricing 12,000 kilometres away.

I am adding Select Harvests (ASX:SHV) as a trade at the open, at roughly A$3.93. The trade is the global almond cycle break, expressed through the lowest-cost producer outside California, with a revenue tailwind from US-China tariff arbitrage and a share price trading at a steep discount to its hard assets.

Here is why the cycle is broken — and why the downside on this trade is bound by physics rather than by market sentiment.

The Ultimate Capitulation Signal

When an industry quietly stops publishing the data that historically hurt its prices, that is not a methodology change, it is a confession. It is the agricultural equivalent of a struggling tech company quietly retiring quarterly guidance.

In December 2025, the Almond Board of California voted to stop funding the USDA Objective Measurement Report — the rigorous July estimate, built on physical orchard sampling, that historically corrected the looser May grower survey [1]. The timing gives away the motive. The vote came months after the July 2025 estimate landed high and triggered the single largest one-day almond price drop in modern memory [1]. Confronted with the one number that reliably delivered downside surprises, the industry chose to stop publishing it rather than argue with it. Officially, the reason was accuracy. Read it however you like — but an industry quietly euthanising the data that keeps embarrassing its prices is not the behaviour of people who think supply is about to flood.

The data backs the tell. California controls 80% of global almond production, so what happens in the Central Valley dictates the cycle [2]. Three things are happening at once:

• Removals are unprecedented. Land IQ data show more than 210,000 acres — roughly 15% of bearing acreage — pulled since 2022, including about 67,000 in 2024 alone. In 2026, California’s bearing almond acreage fell for the first time since 1995 [2].

• The pipeline has collapsed. Non-bearing acreage, the leading indicator for production three-to-seven years out, fell 26% year-on-year. The supply that should feed the 2028–2032 window simply will not exist [2].

• The marginal acre is uneconomic. Westlands Water District, historically the largest destination for new plantings, planted essentially zero new acres in 2023 and 2024, as water costs sat structurally elevated under California’s Sustainable Groundwater Management Act (SGMA) [2].

Combine that with Australian output running broadly flat toward 2030 as growers replant rather than expand, and structural drought across Spain’s growing regions, and the global supply curve is locked. Meanwhile demand keeps climbing — on tariff-driven Chinese buying, Indian consumption, and the developed world’s determination to put almonds in everything that once contained milk.

Capital Allocation: Utility Over Signalling

The market is pricing SHV on cyclical fear, assuming we are near a top. Management is quietly arbitraging that behavioural error with a textbook capital-return playbook [3].

After a strong 1H FY2026 result, the board announced a 10% on-market buyback — around 14.2 million shares, 10% of the issued capital — and resumed the interim dividend. It also secured a new A$60 million debt facility on favourable terms, not for distressed working capital, but to shrink the float at a depressed valuation [3].

This is the unglamorous version of smart capital allocation: financial utility over market signalling. Buying a dollar of productive almond orchard for eighty cents is not a sophisticated manoeuvre — it is just an unfashionable one. And the mechanics are cycle-proof: a 10% reduction in the share count lifts earnings per share by roughly 11% on a flat earnings base, entirely independent of where almond prices go next.

The Asset-Backing Floor

Replacement-cost analysis is the discipline that stops a cyclical equity from becoming a cyclical trade. On a sum-of-the-parts replacement basis [3][4]:

That is an enterprise value of A$557–803 million. Strip out net debt of roughly A$80 million — down 51% to A$79 million at the last full year — and, across 142 million shares, the implied equity sits at A$3.35–5.10 per share [3][4].

At around A$4.00, the market is paying for the assets at below the midpoint of their own replacement value, with zero operating premium. You get the largest Australian almond producer — A$118–130 million of FY26 base-case EBITDA — thrown in for free on top of the dirt [3].

Unless the orchards literally burn down, the hard assets bind the downside around A$3.30–3.50. Below that, the assets are worth more than the equity, which makes a private buyout increasingly plausible — the kind of valuation that tends to attract people with cheque books and patience. Precedent: Costa Group went private at A$3.20 in 2024 on exactly this sub-replacement logic [4].

Reality Check: The Bear Cases

A thesis is only as strong as its weakest link. Here is what breaks this trade, in order of probability.

1. A Trump-China tariff thaw (~12–15%). Australian almonds have been winning share in China on the US-China tariff differential. A partial agricultural deal in late 2026 or 2027 narrows that gap and reroutes Chinese buyers back toward US origin. Impact: a 15–20% cut to the realised-price assumption.

2. A Murray-Darling water shock (~8–12%). This is the hidden leverage in the model. SHV owns the permanent rights to only about 18% of the water it needs and sources the rest from the allocation market — which is to say, from the spot market and the goodwill of the sky. Under El Niño conditions, Australian allocation prices spike viciously (A$300+/ML) [4]. In a concurrent drought, global almond prices rise but SHV’s unit margins compress as spot water costs bite. Impact:15–25% off EBITDA in the affected year — though the hard-asset floor still protects the equity base.

3. Tail shocks (~3–5%). August-September frost, or a resurgence of the varroa mite that hit Australian hives in 2022. Impact: a 20–30% drawdown.

The unifying point: in every realistic bear case, the asset floor binds. Outside the tail scenario, SHV does not go meaningfully below ~A$3.30 without the assets being worth more than the equity. The downside is capped by physics, not by sentiment.

Position and Expected Return

Adding SHV to the list at the open, ~A$3.93, for a 12-to-18-month window — with optionality to extend into the 2028–2030 production gap if the supply-destruction story fully materialises.

*Fair values recomputed on the corrected 142.1m share count (the 1H result confirmed 142.1m, not the 152m the original model used). Upside rows hold a ~7× EV/EBITDA backed out of the base case; bear and tail rows are anchored to the ~A$3.30–3.50 asset floor. These are total returns over the 12–18 month horizon, not annualised. Probability-weighted, the trade carries roughly +23% to +33%, with the downside floored by the assets — a sharper asymmetry than the original model showed, because fewer shares lift every per-share figure and the buyback is retiring stock below replacement value.*

What would make me add: California’s 2026 crop confirmed at ≤2.65 billion pounds; Murray-Darling allocation easing below A$120/ML; FY26 EBITDA tracking above A$130 million. What would make me trim: a US-China deal that narrows the tariff differential; spot water sustained above A$250/ML for two quarters; FY26 EBITDA tracking below A$95 million.

Bottom line. The cycle is not at its peak. The cycle has broken. New orchards do not get planted at twenty-year breakevens, the pipeline has collapsed, and the industry has stopped publishing the number that used to keep it honest. Select Harvests is the cleanest way to own that break: capped downside bound by physical assets, asymmetric upside, and a management team compounding capital while everyone else is reading the cycle backwards. There is, occasionally, a free lunch in markets. It is usually buried under something nobody wants to look at — in this case, more than 210,000 acres of dead trees.

Subscribe now

References

[1] University of California, Davis — 2024 Almond Cost Study (Sacramento Valley & Southern San Joaquin editions), for the new-orchard breakeven horizon; Almond Board of California position reports, including the December 2025 vote to discontinue funding the USDA Objective Measurement Report.

[2] Land IQ Standing Almond Acreage estimates (commissioned by the Almond Board of California; the USDA NASS standalone California almond acreage survey was discontinued), 2022–2026 — bearing removals (more than 210,000 acres / ~15% since 2022; first bearing-acreage decline since 1995, 1,401,097 → 1,385,870), the ~26% drop in non-bearing acreage to the 2025 final (142,306 → 104,900), and Westlands Water District / Land IQ planting data; California ≈ 80% of global production (International Nut & Dried Fruit Council). (SGMA — California’s Sustainable Groundwater Management Act — is the regime behind the elevated marginal-acre water costs.)

[3] Select Harvests 1H FY2026 results announcement and presentation, 28 May 2026 — FY26 EBITDA guidance, the 10% on-market buyback, A$60m debt facility, resumed interim dividend, and orchard / processing / water-entitlement asset disclosures.

[4] Asset-floor anchors — Australian agricultural land and water transaction benchmarks (incl. Murray-Darling Basin entitlement and allocation pricing) and the Costa Group / Paine Schwartz Partners take-private at A$3.20 (completed early 2024; asset-backed precedent).

When a technology concentrates wealth this completely in the hands of capital owners, the political system reacts, as it has at every comparable point in modern history. AI does not ask for stock options, does not unionise, and does not complain about the return-to-office mandate, which is precisely why the reaction will not come from inside the companies. The previous issues in this series traced where economic value migrates as the cost of cognition collapses: toward what AI cannot make — the human element, physical limits, and verification capacity.

The central investable question of the next decade is what the political system will do to the winners.

Morrow’s Diagnosis

In the years around the Great Financial Crisis, Robertson Morrow, a macro thinker and former CIO at Peter Thiel’s Clarium Capital, wrote a paper titled “The Bull Market in Politics” — the paper this issue takes its name from [2]. His argument was simple: going forward, government influence over trade, immigration, social policy, and the decisions of war and peace would matter more to investors than business fundamentals ever would.

He was exactly right, and his thesis perfectly describes the post-AI economy. When the economic engine stops working for the median voter, policy swings take over the wheel.

What is currently breaking that engine? We are watching the bimodal compression of cognitive labour begin [3]. It is driven by two opposing forces acting on the market simultaneously:

• The Pull of Commodification: Capital pushes anything codifiable toward zero cost. If a task can be modelled, AI does it for fractions of a cent. The cost curve for that work collapses.

• The Pull of Irreducibility: What AI cannot do — cumulative judgment, signature liability, primary-source relationships, taste, and verification authority — becomes aggressively more scarce and exponentially more expensive.

The middle will get eaten from both sides. The routinely competent professional — the junior associate, the mid-tier analyst, the junior radiologist — is structurally exposed: their commodifiable output is absorbable by the model, while they lack the accumulated reputation and liability-bearing capacity that defines the top tier. As AI capability matures, the two ends of the economic spectrum will pull violently apart, and the middle will vanish.

Roughly 13% of US cognitive jobs, over 21 million workers, sit directly in the path of that compression [4]. Wage data reveals polarisation of this kind only in retrospect, on a lag of years, and the deployment wave is barely three years old. What shows up so far is concentrated in the occupations where adoption ran first. But the real political danger isn’t happening at the bottom of the income ladder. It is happening in the upper-middle-class waiting room that this compression is emptying.

Elite Overproduction: The Holding Pen is on Fire

This brings us to Elite Overproduction, a concept pioneered by historical macro-sociologist Peter Turchin [5].

Turchin studies history through the lens of population dynamics, and his thesis is dangerously relevant today: late-cycle societies produce far more credentialed “elite aspirants” than there are elite slots to absorb them. The American numbers tell the story plainly. The number of US lawyers tripled from roughly 400,000 in the mid-1970s to 1.2 million by 2011, while the population grew only 45 percent. MBAs over the same window grew six-fold [6]. The supply of credentialed aspirants for elite positions has run miles ahead of the fixed supply of positions themselves.

For decades, the economy kept these surplus elites pacified with high-paying, middle-management cognitive jobs. It was a massive holding pen for ambition. AI is the match held to it.

Nothing radicalises a population faster than a newly minted professional who realises their expensive master’s degree merely qualifies them to supervise a chatbot. When you have too many ambitious, highly educated people fighting over a rapidly shrinking pool of status-bearing jobs, they don’t form a support group. They fracture the political system.

Historically — whether you are looking at the late Roman Republic, pre-revolution France, or the late Russian Empire — intra-elite competition destroys institutional trust. The frustrated elite aspirants turn populist, harnessing the anger of the broader working class to tear down the very structures that locked them out.

The Fujiwara Effect: When Labour Throws a Brick

If Elite Overproduction explains the anger of the professional class, the Fujiwara Effect explains the mechanics of how that anger rewrites the rules of the market.

Borrowed from meteorology, where it describes two or more cyclones merging into a single larger system, the Fujiwara Effect has been made the organising metaphor of Viktor Shvets’s recent work as Global Strategist at Macquarie Capital, most fully developed in The Twilight Before the Storm [7]. In Shvets’s reading, three cyclones are converging: three decades of neoliberal policy, deep financialisation, and the disruptive Information Age now turbo-charged by AI. The central observation is grim: as these forces compound, the “winner-takes-all” dynamic becomes absolute, and the marginal economic utility of human labour collapses. First they came for the blue-collar assembly lines; now they have come for the white-collar spreadsheets.

But Shvets makes a lethal political observation: the changing value of votes.

When a median worker realises their labour no longer gives them any leverage over capital, they remember they still possess one asset the market hasn’t fully depreciated: their vote. And they tend to throw it like a brick.

The democratic system’s traditional mechanism for absorbing inequality is to translate economic discontent into gradual policy shifts. But when labour’s leverage falls this fast, gradualism dies. Silicon Valley billionaires suddenly discovering a deep, philosophical interest in Universal Basic Income isn’t altruism [8]. It is fire insurance. The serious versions are already being drafted, from corporate “displacement taxes” to an AI dividend fund modelled on Alaska’s [9]. They can read the room. They understand that a society where capital owns everything AI cannot make, and labour fights over the shrinking pie of what it can, is a society begging for a wealth tax.

The Catalyst: The Sovereign Accounting Trap

The sociological anger is the fuel, but the spark that ignites the political snapback is pure accounting. Modern Western governments fund themselves primarily through taxing labour (income and payroll taxes). Capital is highly mobile and taxed at much lower effective rates.

Every point of GDP that migrates from salaries to capital income leaves the payroll-tax net with it. If the bimodal compression runs its course, the sovereign tax base hollows out at precisely the moment the political system requires trillions of dollars to pacify a fractured, jobless middle class (via UBI, bailouts, or aggressive stimulus).

How does a sovereign fund a massive redistribution with a shrinking tax base? It cannot do it through gradual policy. It is forced into draconian asset taxes, financial repression, and aggressive currency debasement. This is the mechanical proof for why the coming crisis resolves via inflation.

The Inevitable Snapback: History’s Rhythm

No serious analytical framework predicts that accelerating disparity at record levels resolves itself gracefully. The tension always breaks, and it breaks via state intervention.

Look at the resolution of the three prior US secular bear markets:

• The 1930s (Deflationary): Resolved through the New Deal. A massive, state-executed restructuring of financial regulation, forced by the undeniable political pressure of mass unemployment.

• The 1970s (Inflationary): Resolved through Volcker’s extreme disinflation, after the failure of Nixon’s wage-price controls [10].

• The 2000s (Deflationary): Resolved through Quantitative Easing and financial repression [11]. This was a massive wealth transfer to capital owners to solve a banking crisis. It saved the system, but the bill for that disparity is now coming due.

A forced redistribution is coming. The only question for an investor is how to position capital before the politicians finish drafting the legislation.

The Investable Playbook: What This Means for Capital

The political question is not if the state will intervene to redistribute wealth, but where it will aim when it does.

• Overweight the Diffuse Scarcities (Real Assets & The Human Element): Capital owners of what AI cannot make — relational-sector luxury brands and physical-input chokepoints — hold pricing power that is durable to inflation. More importantly, they possess political camouflage. A politician is highly unlikely to nationalise an ultra-luxury handbag boutique, and they cannot easily expropriate a fragmented network of prime agricultural land. The populist case against them is weak.

  *The Macro Caveat:* Do not mistake political camouflage for invincibility. During an inflationary bust, the state will not outright steal the copper mine, but it will absolutely levy windfall taxes on the excess margin or impose price caps. You own these assets to survive the currency debasement, but you must actively manage the margin risk.

• Overweight Verification Infrastructure: If AI drops the cost of cognitive output to near-zero, the economic value migrates entirely to the entity that stamps it as true. Think of the global testing, inspection, and certification giants — Swiss-based SGS, French-based Bureau Veritas, or British-based Intertek. An AI can optimise a global supply chain flawlessly, but it cannot legally certify that a shipment of medical supplies meets EU safety standards. The liability, and therefore the economic moat, remains human.

• Underweight the Coordination Layer (The Expropriation Target): This refers to the digital middlemen. These are the platforms that don’t actually produce the goods or provide the service, but extract rent by controlling the matchmaking — think Amazon’s marketplace, Uber, App Stores, and the digital advertising duopolies. They are massive, highly visible, centralised, and deeply unpopular. When an angry political system needs a scapegoat to appease a fracturing society, Big Tech has a giant bullseye painted on its back. Margin-crushing regulation or antitrust dismantling of these platforms is a populist dream.

• Hedge with Store-of-Value (With Cyclical Caution): When the political resolution inevitably demands wealth redistribution, the fiat currency bears the cost. Gold and structural currency diversification compound significantly when governments inflate their way out of a social crisis. However, this is not a one-way street; precious metals typically suffer brutal liquidity drawdowns during initial market shocks before ripping higher in the policy response.

  ---

  What This Is Not

  To ensure we are trading on market realities and not political emotion, we must define what this framework does not claim:

  • It is not a prediction of exact timing. These are secular trends. In terms of sequencing, there are actually growing arguments — which I will cover in a coming issue — that we will likely see a cyclical, deflationary recession first. That recession will act as the catalyst, pushing the political system to panic and finally triggering the inflationary policy bust.

  • It is not a partisan political warning. State intervention historically comes from both sides of the aisle. Nixon (a Republican) imposed wage-price controls; FDR (a Democrat) built the New Deal. The math forces the state’s hand regardless of who occupies the office.

  • It is not a doomsday argument. The historical pattern is resolution, not total collapse. If you are familiar with Neil Howe and William Strauss’s The Fourth Turning [10], you recognise this rhythm. Late-cycle secular bear markets force an institutional reset, producing new social contracts and redirecting capital flows. Properly positioned capital does not just survive these transitions — historically, it multiplies.

    Subscribe now

    ---

    The Bottom Line

    We are living in a late-cycle society where the marginal worker’s economic leverage is falling vastly faster than the political system can comfortably absorb. Bimodal compression and Elite Overproduction are the disease; populist redistribution is the historical cure.

    The most dangerous assumption an investor can make today is that the benign regulatory environment of the 2010s will survive the brutal economics of the late 2020s. Asset prices reprice the exact moment a political mechanism becomes visible — not when the bill is signed into law.

    The people who get this right will hold the assets the resolution rewards: the diffuse, inflation-resistant owners of the human element, physical limits, and verification moats. The people who get it wrong will hold the massive, highly visible coordination-layer platforms, watching their margins compress against regulatory caps while the wealth they thought they had concentrated quietly migrates away.

    Thanks for reading! Subscribe for free to receive new posts and support my work.

    ---

    References

    [1] Chief Justice of India Surya Kant’s remarks in open court, 15 May 2026, and the rise of the satirical Cockroach Janta Party (founded 16 May 2026 by Abhijeet Dipke; roughly 19–22 million Instagram followers within a week, surpassing the ruling BJP’s official account): BBC News, CNN, The Guardian and Al Jazeera, 19 May – 8 June 2026. Graduate unemployment: nearly 40% of Indian graduates under 25 are jobless, per Azim Premji University research cited by CNN, 22 May 2026.

    [2] Robertson Morrow, “The Bull Market in Politics” (Clarium Capital, circulated mid-to-late 2000s). Morrow was CIO of Clarium Capital. Primary attestation in George Packer, “No Death, No Taxes: The Libertarian Futurism of a Silicon Valley Billionaire,” The New Yorker, 28 November 2011, which discusses the paper and the thesis within a profile of Peter Thiel. Morrow’s later by-line at The American Conservative (2002–2005) provides corroborating biographical detail.

    [3] Stray Narratives, Issue 12: “Who Gets Paid After AI,” May 2026. The four-mechanism framework and the bimodal compression that runs through each.

    [4] Stray Narratives, Issue 06: “The Noise Economy,” April 2026. The 13% / 77% workforce split.

    [5] Peter Turchin, End Times: Elites, Counter-Elites, and the Path of Political Disintegration, Penguin Press, 2023. The elite-overproduction framework is developed across the book; earlier statements appear in Ages of Discord (2016).

    [6] American lawyer counts ~400,000 (mid-1970s) to 1.2 million (2011), per the American Bar Association, cited in Peter Turchin, Ages of Discord: A Structural-Demographic Analysis of American History (Beresta Books, 2016) and End Times (Penguin Press, 2023). MBA six-fold expansion over the same window per the same Turchin sources. Secondary summary in Noah Smith, “Blame Rich, Overeducated Elites as Our Society Frays,” Bloomberg Opinion, 20 November 2013.

    [7] Viktor Shvets, The Twilight Before the Storm: From the Fractured 1930s to Today’s Crisis Culture (Boyle & Dalton, August 2024). Shvets is a global strategist at Macquarie Capital; the book is cited here as a public work. The Fujiwhara Effect itself is a 1921 meteorological term (named after Sakuhei Fujiwhara) describing the interaction of two or more cyclones; Shvets adopted it as the organising metaphor of his macro thesis, deployed across both The Twilight Before the Storm and his earlier The Great Rupture (2020). His three-cyclone reading (neoliberalism + financialisation + Information Age) is laid out at viktorshvets.com/the-fujiwara-effect and discussed with Alan Kohler in The New Daily, “The Fujiwara Effect — three cyclones of modern world,” 26 August 2024.

    [8] Silicon Valley figures advocating Universal Basic Income, including Sam Altman’s Y Combinator Research UBI experiment and public commentary, 2016 onward.

    [9] Alap Shah, “The Global Intelligence Crisis, Part Three: The Path Forward” (alapshah1.substack.com, March 2026) — the “American Prosperity Compact”: a corporate displacement tax funding income security and an American AI Dividend Fund modelled on the Alaska Permanent Fund.

    [10] Paul Volcker’s Federal Reserve tightening cycle, 1979–1982, taming inflation following the failure of Nixon’s 1971 wage-price controls. Federal Reserve archive and historical record.

    [11] Federal Reserve Quantitative Easing programmes commencing November 2008; the ensuing decade of financial repression as the macro-policy response to the Global Financial Crisis. Federal Reserve communications and successor literature.

    [12] William Strauss and Neil Howe, The Fourth Turning: An American Prophecy, Broadway Books, 1997. The generational-cycle framework.

• The Law of the Post-AI Economy: What an algorithm can codify drops to near-zero cost; what it cannot produce becomes exponentially more valuable. The wealth freed up by collapsing commodity costs doesn’t vanish—it migrates directly toward scarcity. The engine is psychological, not technological. Comin and Lashkari find that roughly three-quarters of the century-long shift in consumption is driven by income effects rather than Baumol’s cost disease: as people get richer, they stop paying for utility and start paying for human provenance and social signal.

• Three scarcities sit on the far side of the cost collapse, each scarce for a different reason and each compounding as AI improves: the Human Element (provenance, exclusivity, curation), the Physical Limit (copper, power, and land that cannot be synthesised), and Verification (the judgment and liability that underwrite cognitive output).

• The labour market splits in two: the junior doing commodity work loses leverage to the model, the senior holding the licence and the liability keeps the margin, and the middle is squeezed. One honest correction to our earlier work — the early hollowing-out of entry-level hiring tracked remote work more closely than AI. The AI displacement is additive, and still ahead of us.

The conclusion is to own the chokepoints, not the AI software winners: human-element brands with waitlists, physically constrained real-asset producers, and the verification moats of audit, certification, and named-principal liability. Late in the Kondratiev cycle, with the next secular bear likely inflationary, these hold pricing power while commoditized producers compress to zero real margin.

Here below is the AI Podcast version of this same article:

Starbucks tried to automate. A $112 billion company selling one of the most standardised products in the modern economy spent years systematically removing labour from its stores.

Then, this spring, CEO Brian Niccol reversed course. He brought back handwritten notes on cups. Ceramic mugs. The return of great seats. More baristas, not fewer. He reinvested in the hospitality, not the coffee. [1]

Read carefully, that decision points at the entire post-AI economy. Starbucks reached for what AI could not replace, because what AI could replace was no longer the thing customers were actually paying for.

The consensus story about artificial intelligence assumes that when machines can produce anything humans can produce, the jobs disappear and human economic value goes to zero. That view is fundamentally backward.

The principle of the new economy is entirely mechanical, and it fits into one sentence: What AI can produce gets cheap; what AI cannot produce gets exponentially more valuable.

AI is a production technology. It compresses the cost of producing the things it can codify. Anything fungible, anything modular—the cost curve collapses, and the price follows. But as the cost of these commodities drops to near-zero, the freed-up capital and consumer spending do not disappear. They migrate.

The most important investable question of the next decade is where that money flows. The data shows it flowing directly toward three distinct scarcities that possess moats an algorithm fundamentally cannot cross.

The Engine: The Psychology of Wealth

We are consistently told that services dominate the modern economy because they resist automation. That is mostly false. In 2021, economists Diego Comin and Danial Lashkari showed that this price effect—Baumol’s cost disease—only accounts for about a quarter of the macroeconomic shift. The other 75% is driven by income effects. [2]

When income rises, people stop paying for pure utility and start paying for social signals. You can see this clearly in US consumer data: the wealthiest 20% of households don’t just consume a larger volume of goods than the bottom 20%; they disproportionately shift their wealth toward categories with a heavy relational component—in-person dining, entertainment, and exclusive services. [3]

This is the macro engine of the post-AI economy. The structural reallocation of wealth runs on human psychology, not just technology. Even if AI can automate a luxury service for pennies, wealthy consumers will still shift their spending toward the human alternative—because paying for the human alternative is the point.

Which brings us to the first scarcity.

1. The Human Element (The Psychology of Scarcity)

Nobody who buys a $3,000 Armani suit is buying a more efficient way to stay warm. They are buying the brand, the social meaning, and the fact that other people know what it is but cannot have it.

As utility gets cheap, the premium on social signalling skyrockets. Demand shifts toward goods whose entire value is derived from human provenance, exclusivity, and taste.

Economists Alex Imas and Graelin Mandel recently ran an experiment using physical art prints that serves as the single most important data point for understanding consumer behaviour in the AI era. When human-made artwork was made exclusive (one physical copy versus many), its value jumped by 44%. But when AI-generated artwork was made exclusive, that premium collapsed to just 21%. [4]

Sit with that finding. The mere fact that an algorithm was involved in producing the object destroyed the exclusivity premium. AI involvement is inherently anti-exclusive. It cannot capture the premium that comes from human provenance because that premium is defined entirely against AI involvement.

Furthermore, as AI drops the cost of generating content and products to zero, supply becomes infinite. When supply is infinite, the ultimate scarcity is the Filter. The market will pay massive premiums for human curators, tastemakers, and editors who filter the deluge of AI-generated noise and dictate what is actually worth our attention.

2. The Physical Limit (Geological Scarcity)

AI moves at the speed of code. The physical world moves at the speed of geology.

You can prompt-engineer a million lines of code in seconds, but you cannot prompt-engineer a new copper mine, a gigawatt of baseload power, or a new water basin. We are currently watching the algorithmic, exponential growth of AI compute collide directly with a physical supply curve that refuses to budge.

Look at the agricultural equivalent: productive land and water rights. In California’s Central Valley, the breakeven on planting a new almond orchard has stretched to twenty years because water is so physically and regulatorily constrained. [5] The marginal acre is uneconomic. The supply curve is effectively dead. A coming issue will lay out the investment case on this specifically.

Just as AI destroys the exclusivity premium of luxury goods, it is completely powerless to synthesise base reality. Whether you are looking at copper miners with depleting reserves, LNG export terminals with physical chokepoints, or grid hardware manufacturers with multi-year backlogs, the thesis holds.

3. Verification (Reputational Scarcity & The Balance Sheet)

If the demand side of the new economy is driven by the human element, the supply side is driven by verification and risk.

AI is aggressively commoditising the raw materials of cognitive work: drafting, data analysis, calculation, and pattern recognition. When the cost of producing complex cognitive output drops to near-zero, the output itself is no longer the product. What stays valuable is the signature and the balance sheet.

An AI can draft a flawless 50-story architectural blueprint for a fraction of a cent. But an AI cannot assume liability if the building collapses. It cannot go to jail. It cannot hold a licence. Similarly, an AI can mathematically underwrite a massive commercial real estate loan, but the AI cannot take the financial loss if the borrower defaults. Risk capacity is strictly constrained by human or institutional capital.

The medical opinion, the audit letter, the legal interpretation, the structural stamp—the underlying work product gets cheap, but the cumulative human judgment and financial liability that underwrites it stays scarce.

This dynamic will produce a bimodal compression in the labour market as AI capability matures. The junior associate doing commodity drafting loses leverage to the model; the senior partner holding the liability licence captures the margin. The middle gets squeezed, the top compounds—and the speed of that compression will be the central question of the next five years.

Where the Capital Should Sit

The investment framework over time will not be about picking the AI software winners. It will be about owning the chokepoints of the post-commodity economy.

• The Human Element (Product as Signal & Filter): Businesses where the human element, curation, or exclusivity is the product itself. Handbags with multi-year waitlists. Concert venues with finite capacity. Family-controlled luxury houses with a century of brand equity.

• The Physical Limit (Dead Supply Curves): Businesses where the input is physically constrained. Copper miners with depleting reserves. Energy infrastructure and grid hardware manufacturers with multi-year backlogs.

• The Verification Moat (Irreducible Signature): Businesses on the supply side of cognitive production where the human signature and balance sheet are the units of value. Global testing, inspection, and certification (TIC) giants. Audit firms with senior-partner sign-off authority. Architecture firms with named-principal liability.

The Bottom Line

The consensus assumes that because the commodity form is completing itself—delivering products with zero human involvement at near-zero marginal cost—the value of human economic output goes to zero.

In reality, cheap commodities free up real income. That income flows directly toward comparative preferences, physical constraints, human curation, and liability-bearing signatures.

The macro environment makes positioning for this reallocation urgent. We are sitting on the late-stage side of the Kondratiev long cycle. [9] Historical patterns—moving through 1930s deflation, 1970s inflation, and 2000s deflation—suggest the next secular bear market will likely be inflationary. Real-asset businesses and luxury brands with pricing power survive inflationary regimes by definition. Commoditised producers compress against zero real margin.

But there is a looming, darker question here. If AI concentrates wealth into the hands of capital owners, and systematically compresses the wages and leverage of the middle class, the tax base of the sovereign state collapses exactly when society demands a bailout.

What happens politically when a highly educated, newly displaced class gets angry? That is the subject of a coming issue.

Subscribe now

References

[1] Starbucks “Back to Starbucks” turnaround under CEO Brian Niccol: the reversal of in-store automation in favour of hospitality — return of ceramic mugs, handwritten notes on cups, and condiment bars, alongside increased barista staffing (staffing raised at ~3,000 stores; a 700-store coverage pilot). Niccol’s remark that the chain had “ran like a manufacturing facility” per Fortune, 23 March 2026; programme detail per Starbucks corporate, “Back to Starbucks,” 2025, and Reuters reporting, May 2025.

[2] Diego Comin, Danièle Lashkari, and Martí Mestieri, “Structural Change with Long-Run Income and Price Effects,” Econometrica, Vol. 89, No. 1 (2021). Income effects account for roughly three-quarters of the long-run shift in consumption structure; the Baumol price / cost-disease channel for about a quarter.

[3] US Bureau of Labor Statistics, Consumer Expenditure Survey, 2022 — top-quintile versus bottom-quintile expenditure on relational categories (in-person dining, entertainment, personal services). Corroborated across the universe of US consumer spending by Joachim Hubmer, “The Race Between Preferences and Technology,” Econometrica, 2023.

[4] Alex Imas and Graelin Mandel, experimental findings on AI involvement and the exclusivity premium for human-made versus AI-generated artwork — a +44% exclusivity premium on human-made prints versus +21% on AI-generated prints. Summarised in Alex Imas, “What Will Be Scarce?”, Ghosts of Electricity (Substack), 2026, with companion technical note at aleximas.com.

[5] California’s Sustainable Groundwater Management Act (SGMA), now in enforcement across the San Joaquin Valley, with irrigation districts shutting deep-well agricultural pumps from 2026; the Public Policy Institute of California projects that pumping reductions could take up to ~20% of San Joaquin Valley farmland out of production by 2040. Almond orchards (3–4 acre-feet of water per acre, and non-fallowable once planted) are the marginal casualty.

[6] Peter John Lambert and Yannick Schindler, "The Broken Ladder: AI, Remote Work, and Early-Career Hiring," SSRN working paper, 2026 (analysis of ~243 million new hires and ~407 million job postings across the US, UK, Canada and Australia, 2017–2025); when work-from-home and generative-AI exposure are estimated jointly, the work-from-home effect on the junior-hire share holds while the AI coefficient attenuates toward zero. Analysis sharpened by Paul Kedrosky.

[7] A major global investment bank’s proprietary adoption survey (April 2026; ~800 companies across banking, software, tech hardware, semiconductors and professional services in the US, UK, Germany, Japan and Australia, all 12-month-plus AI adopters): a net ~5% reduction in roles over the prior year, with eliminations and non-backfilled positions concentrated in early-career staff — roughly six times the rate of the most experienced. Anonymised per house sell-side policy.

[8] US Census Bureau, Business Trends and Outlook Survey (BTOS), 2026: fewer than 20% of US firms report any AI use in a given two-week window; over 90% report no employment effect from AI over the prior six months; among firms reporting an effect, more cite increased than decreased employment. The most AI-exposed sectors account for roughly 15% of total US employment.

[9] Kondratiev long-cycle framework as documented across roughly a century of bond-yield and secular equity-cycle data; current late-cycle position per a long-cycle macro strategist’s May 2026 work (anonymised per house attribution policy). The three prior US secular bear markets — 1930s deflationary, 1970s inflationary, 2000s deflationary — bracket the pattern.

Markets anchor to trauma. When an asset class burns investors badly enough, as Chinese equities did through the tech crackdowns, zero-COVID lockdowns, and the property collapse, the consensus simply stops looking at the data. The narrative hardens into a single word: uninvestable.

But the market’s refusal to look is exactly where the alpha lives.

Right now, China is quietly handing us three massive, structural tailwinds firing simultaneously:

• Cloud and AI pricing power is expanding.

• The consumer is shifting from defensive hoarding to equity participation.

• The real estate drag is finally finding a floor.

The country has not seen these three pillars align in over a decade.

Yet, because the crowd is still trading the trauma of 2022, the biggest beneficiary of the global AI boom is being priced as if the AI revolution is passing it by entirely.

At the open, I am adding KWEB (China internet) and MCHI (broader China) to the model portfolio, sized for a 12-to-18-month window. MSCI China is currently trading at roughly 12.6x consensus 2026 earnings. Compare that to the S&P 500 at 21.4x. They are posting broadly comparable EPS growth, yet you are getting the Chinese tech-internet segment, which is actually growing faster than the US aggregate, at a 41% discount [1].

I am not playing for a dead-cat bounce but a potential structural rerating. Here is the case for why I am making this move.

Pillar 1: The new era of tech pricing power

To understand the Chinese tech opportunity, you have to look at the global compute market. Compute is structurally scarce. Through 2024 and 2025, the AI story was about land grabs and volume as companies subsidized access to build their user bases.

Today, that phase is over. Pricing power has taken the wheel. Anthropic stopped subsidizing enterprise plans this spring, and per-token prices have jumped [2]. Customer demand has vastly exceeded installed capacity globally, meaning the providers now hold the pen on price. The race-to-the-bottom on per-token cost has officially reversed.

The West operates under the assumption that China is locked out of this dynamic due to semiconductor sanctions, but the data says otherwise.

According to Stanford’s 2026 AI Index, the US-China frontier-model performance gap has collapsed from 17.5% in 2023 to just 2.7% today [3]. Chinese models are running roughly 12.6 trillion tokens per week (quadruple the US volume in recent tracking) [4], and they are doing it at 8% to 14% of the unit cost of leading US closed-frontier models [5]. China is running the AI economy on a vastly steeper, cheaper domestic demand curve.

When a scarce input meets immense demand, you get multiple expansion and profitability acceleration at the same time. Alibaba is the cleanest proof of concept. In their FY26 Q4, cloud and AI revenue grew 38% year-over-year. But more importantly, cloud and AI adjusted EBITDA grew 57% [6].

That 19-point spread between revenue growth and profit growth is the sound of operating leverage firing on all cylinders. The capex is already yielding cash. As Alibaba CEO Eddie Wu put it on the 13 May earnings call:

“There isn’t a single card on our service that is idle... we have a lot of customers still waiting to access the service.” [6]

Pillar 2: The saver-to-investor rotation

The persistent bear argument against China is the lack of a “bazooka” fiscal stimulus to revive the consumer. What this argument misses is that a second stimulus event isn’t necessary. The capital is already sitting on the household balance sheet; it just needs a reason to move.

Over the last few years, the Chinese consumer played aggressive defense. But the premium tier is turning. Alibaba’s 88VIP membership (their top-decile capture) hit 62 million in Q4, with growth accelerating to 24% [6]. Merchant ad spend on the platform is accelerating [6]. This is wallet-share capture among the knowledge-economy professionals in Hangzhou, Shenzhen, and Shanghai who are already reaping the AI windfall.

The mechanism that turns this premium-tier spending into broad market participation is financialization. There is a towering 50 trillion yuan wall of one-year-plus deposits maturing in 2026. In Q1 alone, 7.6 trillion yuan landed in personal deposits [7].

But savers are waking up to the reality that parking cash in defensive assets is a dead end. In January 2026, the China Securities Depository and Clearing Corporation reported 4.9 million new individual investor accounts, the highest monthly print since the stimulus peak in late 2024 [8]. The retail investor is beginning to speculate in equities again on a scale we haven’t seen since the 2014-2015 cycle. The cash is migrating to the stock market and the impact compounds.

Pillar 3: Property is clearing

I am not making a sweeping, optimistic call on Chinese real estate. The aggregate numbers across 70 cities are still in deflation [9]. But to use the aggregate as a reason to avoid China is lazy analysis.

What matters to the broader equity indices is that the drag is effectively capped, and in the areas that matter, the market is clearing.

First, Tier-1 resale has officially turned positive. In April, Shanghai, Beijing, Shenzhen, and Guangzhou all printed positive month-over-month price action [9]. Shanghai just recorded its highest secondary-home transaction volume for the month of April in a decade, surpassing the 2019 pre-crisis peak [10]. The recovery is highly specific to the AI-windfall cohort cities, and it is widening.

Second, and more importantly, the supply side has been obliterated. New construction starts are down more than 70% from their peak [9]. Developers are solely focused on completing existing projects.

This mirrors the exact mechanics of the Spanish and Irish real estate markets in 2014 and 2015. When you completely halt new supply, the residual inventory gets absorbed, prices flatten, and the sector stops being a black hole for GDP. The math points to a transition in 2027 where property actually shifts from a drag to a contributor. Until then, it is simply no longer the weight that sinks the ship.

The reality check: what could break this trade

I take the failure points seriously. Here is what can go wrong:

The AI capex air pocket. This is the most legitimate threat. The defense against a tech bust is the $1.9 trillion in aggregate Remaining Performance Obligations (RPO) logged by the Big Four hyperscalers. But that backlog is dangerously concentrated (Oracle’s $455B total RPO is roughly $300B in OpenAI commitments) [11]. Historically, suppliers re-rate the moment a major customer guides capex down, not when a bust is confirmed. If a US hyperscaler trims its 2027 capex guide by 10-15%, the Asian supplier base will compress 20-30% within a week. The Chinese internet leg is partially insulated by domestic demand, but it would not escape the gravity.

Geopolitics. A 41% multiple discount already prices in a heavy geopolitical premium. However, a Trump escalation (a blanket Section 301 tariff above 60% and an expanded Entity List) is the modal bear path. It would be worth a 10% to 15% net drawdown.

The policy regime. Central Huijin (the state fund) operates as a PBoC-backstopped floor, but it doesn’t buy for upside [12]. The implicit bet of this trade is that Beijing maintains its current pro-AI posture. If they revert to 2021-style platform crackdowns, the trade requires immediate liquidation.

The 90-day watchlist. I am actively monitoring the mid-June NBS release for any Tier-1 city resale flipping back to negative, alongside May PPI data and Alibaba cloud growth (a dip below 30% breaks the momentum thesis). I am also watching for any formal Beijing action constraining AI commercialization.

The playbook and expected returns

This is a timing-arbitrage trade, sized to capture the closure of the multiple gap rather than relying on heroic EPS revisions.

The Base case assumes a partial gap closure to roughly 15x forward earnings on MSCI China against a stable US multiple. The Bull case adds a re-rating to 17-18x plus upward EPS revisions on AI-cloud acceleration. The Bear case accounts for the AI capex air pocket intersecting with a regulatory shock.

JD.com is the multiple-recovery candidate at the defensive end of the KWEB basket. It trades at roughly 9 to 10 times consensus 2026 earnings against a balance sheet closer to a fixed-income instrument than to a Chinese e-commerce platform. The market is pricing it as if the franchise has permanently lost share to PDD and Meituan, which is not what the operating numbers say. JD bounds the downside of the basket if the broader AI rerating thesis takes longer than 12 months to play out.

Alibaba is the multiple-recovery candidate at the offensive end of the KWEB basket. The 57% cloud and AI EBITDA growth that anchored Pillar 1 is being priced at roughly 12 times forward earnings, against US peers trading at 25 to 30 times for materially less acceleration. The market still has Alibaba in the 2022-platform-crackdown frame even as the cloud segment compounds. If pricing power on compute holds, this is the name in the basket where EBITDA growth flows into the multiple first.

Bottom line

The financial media is treating Chinese equities as uninvestable value traps. But true value traps are defined by deteriorating fundamentals, shrinking margins, and capital flight. They do not exhibit 57% EBITDA growth in frontier technologies, a rapidly steepening domestic demand curve, and millions of retail investors rotating out of cash and into equities.

The dislocation between the dominant market narrative and the actual cash flows being generated on the ground is the widest it has been in ten years. You do not need a geopolitical miracle or a massive government bailout for this trade to work. You simply need the market to realize it is aggressively pricing in an apocalypse that is no longer happening.

References

[1] MSCI China and S&P 500 consensus forward P/E and 2026 EPS growth: MSCI factsheets (30 April 2026); S&P 500 consensus via FactSet aggregate.

[2] Anthropic enterprise pricing communications, spring 2026; OpenAI, Google Cloud, and AWS public API price lists through Q1-Q2 2026; NVIDIA earnings commentary on GPU pricing and inference wait-lists.

[3] Stanford Institute for Human-Centered AI (HAI), 2026 AI Index Report — frontier-model benchmark performance gap.

[4] OpenRouter weekly token-consumption analytics, late March / early April 2026; reporting via Bloomberg secondary coverage.

[5] DeepSeek V3.2 and GPT-5.2 published inference pricing per million tokens; benchmark comparisons from Inference.net independent compilation; Anthropic Claude Sonnet 4.5 published rate card.

[6] Alibaba Group Holding Limited, FY26 Q4 earnings release, supplemental financial information, and earnings call transcript (CEO Eddie Wu management commentary), 13 May 2026.

[7] People’s Bank of China, Quarterly Financial Statistics Q1 2026, released 13 April 2026.

[8] China Securities Depository and Clearing Corporation (CSDC), Monthly Investor Account Data, January 2026 release.

[9] National Bureau of Statistics of China releases, April 2026 cycle: 70-City Sales Prices of Residential Buildings (18 May), Producer Price Index (10 May), and new-construction-starts series.

[10] China Real Estate Information Corporation (CRIC), April 2026 monthly secondary-home transaction data.

[11] Hyperscaler RPO disclosures, Q1 2026: Microsoft, Alphabet (Q1 2026 Form 10-Q cloud RPO), Amazon, Oracle (FY26 Q1, $455bn total RPO). Aggregate roughly $1.9 trillion across the Big Four.

[12] Bloomberg reporting on Central Huijin Q1 2026 position adjustment and PBoC backstop commitment, April 2026; PBoC press communications on the Central Huijin lending facility.

Key Takeaways

• AI generates four observable mechanisms across the cognitive economy: recursive output proliferation, constant experimentation as time-sink, verification burden, and attention fragmentation. Each carries its own bimodal logic and its own compressed middle.

• The audit profession ran one of these mechanisms first. Between 2002 and 2026, audit work commodified entirely; the signature on Page 1 kept 100% of the pricing power. The cognitive economy is now running all four at once.

• The slice of the workforce this hits is roughly 13% of the US labour force, the proceduralised cognitive roles Issue 06 named [11]. The other 77% (food service, retail, construction, healthcare support, transport) is not running this trade.

• The unemployment question is a sequencing question, not a directional one. AI eliminates roles at machine speed; new economic categories crystallise at institutional speed. The gap is where the shock lives.

• The equilibrium across each mechanism is bimodal: a small top tier of cumulative-judgment-holders retains pricing power, a large bottom tier becomes interchangeable, the middle goes away.

Here below is the podcast version of this article produced with NotebookLM:

What an auditor sells, in 2002 and in 2026

In 2001, an auditor sold thoroughness. The audit itself was the product, the partner’s signature at the end was a procedural formality, and selling thoroughness is what you do when you have nothing else to sell.

Enron collapsed in late 2001. Arthur Andersen, the audit firm that had blessed its accounts, collapsed with it. A market eighty-nine years in the building evaporated in nine months because the signature had not done what the signature was supposed to do.

Sarbanes-Oxley followed in 2002. Buried inside the thousand pages of other reforms was the change that mattered: the audit partner whose name appeared on the opinion was now personally responsible for it. The signature became the unit of liability; the workpapers, which for ninety years had been the product the audit firm sold, stopped being the product. Signed accountability did.

Twenty-four years later, the audit field work is largely automated. AI handles anomaly detection, transaction sampling, reconciliation, control testing, journal-entry analysis. The audit teams of 2026 are smaller, more senior, supervisory. What they sell is the same thing they sold in 2003: the partner’s name on Page 1. The room got smaller. The signatures got more expensive.

If you have ever signed an audit letter, or sat across a table from someone who has, this is the transition you have already lived through. The audit profession ran the cleanest of four mechanisms first, because verification was the dominant binding constraint and Sarbanes-Oxley named it in legislation rather than letting the market discover it slowly. The cognitive economy is not so lucky. It is running all four at once, with no Sarbanes-Oxley to clarify any of them.

The slice of the cognitive economy this transition will run through is roughly 13% of the US workforce but that 13% is going to create havoc and the market is still in denial.

As Ken Griffin of Citadel appears to have just noticed, finance will be hit hard. Financial and macro research, probably does not even need the audited signature approval on the content created. Its forecasting ability has always been poor not from lack of work or effort, but simply because forecasting complex adaptive systems is more art than science. However plentiful the hallucinations of the AI model, they will likely be no less error prone than the approximations and assumptions of the analysts. As he said, “there is no theoretical position to retreat to”, especially in finance.

The data is starting to show up where the framework predicts it should. The unemployment rate for workers under thirty-five holding a Master's degree has reached its highest level in nearly twenty years. Business-school applications are dropping fast enough that specialised degrees are now openly discounted; the higher-education credential factory boom that defined the 1990-to-2020 period is ending in real time.

The four mechanisms

A previous piece [10] named the four mechanisms AI produces in cognitive labour markets. Each runs in parallel; each creates its own bimodal compression; each carries its own scarce role; each is at a different stage of legibility.

1. Recursive output proliferation. AI outputs become inputs become outputs. The half-life of any specific knowledge advantage collapses, because what one analyst learns this quarter is in everybody’s training set the next.

2. Constant experimentation as time-sink. Every team running parallel trials on tools, workflows, agents, and models. Time spent evaluating what to use displaces time spent on the work itself.

3. Verification burden. When an AI can produce anything, verifying becomes the bottleneck. Citations may be hallucinated, code may compile and still be wrong, drafts may pass spell-check and still be incoherent on a re-read.

4. Attention fragmentation. The bottleneck is no longer capability but choice: what to attend to, in a stream that has lost most of the curatorial gates that used to do that work for us.

Subscribe now

Recursive output proliferation

The premium on having access to a piece of analysis falls because the cost of producing equivalent analysis falls. Each output becomes training data for the next model release, and each release lowers the floor under what a generic AI summary can produce on the same topic. Volume rises; average per-unit value falls, perhaps a long way.

Sell-side equity research is the cleanest live case. The morning notes that used to anchor a buy-side institution’s intraday positioning are now, within a quarter, ingested, paraphrased, and surfaced as AI-summarised feeds to subscribers who never paid the issuing firm [15]. The note still exists; the informational asymmetry it used to carry does not. The half-life of an analyst insight has compressed from quarters to days.

Academic working papers run the same loop. A preprint posted on arXiv on Tuesday is summarised, cross-referenced, and embedded in literature-review queries by Thursday. Authorship is preserved; what has changed is the duration over which the paper functions as the reader’s first encounter with its argument.

Inside the cognitive workplace the loop runs faster. An internal research memo, once written, is fed into the firm’s AI tooling within a week and surfaced as context to every junior analyst’s queries thereafter. The analyst who wrote it has handed her firm a permanent productivity asset and lost the ability to be paid for being the only person who could have written it. Two quarters later, she is competing with a tool trained on her own outputs.

What remains scarce is what the training set cannot reproduce. A radiologist with a million scans has calibration the model cannot have until it is trained on a comparable distribution, and the distribution itself is the constraint. A senior credit analyst with thirty years of cycle memory holds judgement compressed from outcomes the model has no access to. A primary-source researcher with cultivated relationships at central banks, regulators, or operating-company management produces information the model has never seen because it was never written down [1][3]. The conference-room conversation that never makes it to email, the off-record deal-room exchange, the boardroom briefing that exists only in the partner’s memory, the dinner with a CFO whose remarks are not minuted: these are the channels that stay scarce by structurally avoiding the recursive loop. The professional whose value sits in those channels remains paid for being trusted; the professional whose value is fully digitised has handed it to the next model release.

The bimodal: signal-extractors and primary-source-holders keep their pricing power; output-producers compress; the middle, the analyst whose work the next model release will absorb at no cost to the customer, goes away.

Constant experimentation as time-sink

There is a tax on attention, the AI experimentation tax, and it takes the form of keeping up with platform churn. The tax compounds quietly because the comparison to the prior tooling regime fades from memory. People remember being busy; they do not always remember whether the busyness was productive.

An engineering organisation ran on a stable IDE in 2022. In 2023, GitHub Copilot was the new shared assumption. In 2024, Cursor displaced Copilot for many teams. In 2025, Claude Code displaced Cursor for several of those same teams. Each marginal model release shifts the relative ranking enough to justify a re-evaluation. Each evaluation costs a team a week of attention; each team runs roughly four per year. The aggregate looks like a tax of one month per engineer per year, paid in deferred work, before any productivity benefit from the chosen tool is counted.

The same dynamic runs one layer up. Agentic frameworks, MCP servers, orchestration layers, retrieval-augmented stacks: each is now under live evaluation at most cognitively intensive firms, and the cost of the evaluation is the senior engineer-week that is not spent on the work itself. The output of these meetings is rarely a new tool; it is usually the firm’s collective decision to keep using roughly the same tool, slightly upgraded. The tax was paid for an answer the firm already had.

What remains scarce is the meta-decision-maker. The senior practitioner who can credibly say “we are not re-running this comparison; the gain is below the cost of evaluating it” removes the tax from the teams underneath. Standard-setters who pick a stack and force the firm to live with it for two years rather than two months recover the attention the experimentation tax was draining. These are not popular roles. They are paid because the alternative is the entire firm in a permanent state of evaluation.

The bimodal: meta-decision-makers and standard-setters win. The busy-but-not-deciders, running the comparison for the fifth time, lose the attention that used to do the work itself. The middle, the team-lead who keeps comparing because no one more senior will stop the comparison, is the one whose week vanishes into the tax.

Verification burden

When an AI can produce anything, verifying becomes the bottleneck. Citations may be hallucinated, code may compile and still be wrong, drafts may pass spell-check and still be incoherent on a re-read. This is the mechanism the audit profession lived through first, and the one Issue 01 named as the Trust Economy quadrant of the Verification-Substitution Matrix [7]. It is now running across the rest of the proceduralised cognitive economy with no Sarbanes-Oxley to organise it.

If you have ever spent a week waiting on a single senior signature while the underlying work was done in an hour, you have already met this bottleneck.

• Legal. Bottleneck: a senior partner has to verify the AI-drafted brief before it can be filed. Lawyers in Texas, New York, and California have been sanctioned for filing briefs with AI-fabricated citations [4]. A senior partner described the new workflow as “a whole second associate-level workforce we did not have to hire two years ago, doing nothing but verifying what the first one produced.” A second associate-level workforce, except non-billable, and somehow this is filed under productivity gains.

• Medical. Bottleneck: the AI cannot be the named clinician of record on the diagnosis. Diagnostic AI has reached or passed expert-level accuracy on benchmark datasets in radiology, dermatology, and pathology. The radiologist who signs off is still the named accountable clinician; the signature is the unit the legal and insurance systems can act on. AI produces more candidate findings per patient that must be ruled in or out, and the rate at which a radiologist can rule findings in or out has not changed. The AI is faster, the radiologist is the same speed, and the patient queue gets longer.

• Software. Bottleneck: a senior engineer has to credibly approve the pull request. Code volume is up; the engineer count is down; senior time in code review is up sharply. Engineering leaders describe the new bottleneck as “the rate at which a senior engineer can credibly approve a pull request,” and that rate is the same as it was in 2022. The career path that produced senior engineers now runs through the tier the AI is eating, which is the sort of thing the strategy memo never quite gets to.

The 2025 macro numbers fit. US GDP grew 2.7% with 181,000 net jobs added, the worst non-recession year for net employment since 2003 [5].

The bimodal: senior verifiers retain pricing power; junior producers compress, because the bottom rung of the cumulative-judgment ladder is now performed by the model.

Attention fragmentation

Email becomes agent-to-agent. Search becomes AI-mediated. Content becomes infinite. The bottleneck is no longer capability but choice: what to attend to, in a stream that has lost most of the curatorial gates that used to do that work for us.

Google’s AI Overviews now displace a meaningful fraction of the organic-search referral that used to send users to underlying publishers [14]. The user gets an answer; the publisher gets nothing. The previous bargain (the publisher writes, search indexes, the user clicks through, the publisher monetises the click) has been quietly rewritten so that the click is no longer needed. The gate (the search-results page with its ten organic links) has been replaced by a single curated answer in which no publisher’s URL needs to appear. This is the channel-controller dynamic Issue 04 named [9], one layer further down the stack.

Inbox email is running the same compression. Agent-to-agent message flow (the AI assistant that triages, summarises, and replies on behalf of its user) reduces the inbox from a stream of human signals to a stream of pre-processed summaries. The pre-processing removes senders the user no longer needs to hear from, which sounds like an upgrade; it also removes the curation-by-attention that used to make the inbox a signal of who was paying attention to whom. The signal degrades; the time saved feels real; the relationships do not survive.

Long-form content is the most obviously infinite. Every newsletter, podcast, and video channel the user might subscribe to now produces at the rate the user used to consume across all of them combined. The cost-of-production constraint has fallen; consumption has not expanded to meet the new floor. What used to be a curation problem is now a triage problem.

The scarce role is the tasteful curator operating at scale. The Substack writer whose subscribers trust their selection enough to read what is recommended and skip what is not, the podcast host whose guest selection is itself the product, the editor whose name on a piece of content is the reason it was opened: these roles function as the new choice-architecture. The user delegates attention to them because the alternative is to drown in the unfiltered stream [2].

The dynamic mirrors music after the phonograph: a small star tier whose names were the asset, a great middle buried. Network effects in content distribution concentrate attention on a handful of writers, podcasters, and curators whose distribution is near-universal; the majority is statistically invisible not because the work is worse but because the volume of competing output makes any individual title disappear against the average.

The bimodal: tasteful curators with cumulative reader trust retain pricing power; generic distributors lose. The middle, the title whose curation is now indistinguishable from any other generic gate, goes away. The Google homepage of 2012 is the cognitive economy’s church-choir-and-dance-hall of 2026: the middle distribution layer, eaten by the platforms above and the algorithmic curation below.

Pieces like this one are fighting the same fight, against the same deluge. The fact that the reader has made it this far means the curatorial gate worked at least once today, which writer and reader (?) can perhaps both notice with some relief.

The gap between elimination and emergence

The unemployment question is a sequencing question, not a directional one. AI eliminates roles at machine speed across all four mechanisms. New economic categories crystallise at institutional speed, which is slow, and in forms nobody currently knows how to describe.

The audit profession is the reminder that the gap can be long. The transition from Sarbanes-Oxley in 2002 to the senior-and-supervisory 2026 audit firm took twenty-four years, with workforce contraction front-loaded and the new equilibrium back-loaded. Junior auditors absorbed the loss; senior partners absorbed the rebalanced rent; the audit-manager tier in between, whose function had been to verify juniors who were no longer being hired, was the slowest-resolving casualty.

The four mechanisms now run in parallel without the legislative cue Sarbanes-Oxley supplied. There is no act of Congress arriving to clarify that the AI summary is not the named author of record; no statute defining the verification ratio at which an AI-generated brief becomes a sanctionable filing; no licensing body declaring the meta-decision-maker the new partner-track role. Institutions form around the gap slowly, and the gap is the shock.

I do not try to predict the length of the gap. I am naming the mechanism by which the unemployment outcome is determined: elimination runs faster than emergence; the unemployed are the working-age population in the gap; the gap is set by the slower of the four institutional clocks. Whichever mechanism resolves slowest sets the duration of the dislocation.

Bottom Line

Two forces, opposite directions. Capital pushes work toward commodification, the way the phonograph pushed singing: the singing stayed, the leverage dissolved, the great middle of working singers in church choirs and dance halls mostly went away [8]. Accountability pushes the surviving slice toward irreducibility, the way Sarbanes-Oxley pushed the audit partner’s signature toward irreducibility. AI accelerates the first and sharpens the asymmetry; the mechanism is the same in each of the four domains, the costume different.

The equilibrium is bimodal. A small top tier (cumulative-judgment-holders, primary-source-holders, meta-decision-makers, verifiers, tasteful curators) keeps its pricing power. A large bottom tier (the analyst the next model release absorbs, the team-lead running the fifth comparison, the junior producer whose verification ladder has been kicked away, the generic distributor) becomes interchangeable. The middle hollows out across all four. What would falsify this is a stable verification ratio across firms over time, or a regulatory shift commodifying accountability the way limited liability commodified ownership in the nineteenth century. Neither has happened yet.

The luxury sector of cognitive labour compresses into a single tier of Hermès at the top and a long tail of interchangeable at the bottom [12]. The middle goes away.

If you are reading this from inside the affected 13%, the question I keep returning to is which fields defend themselves and which do not. No checklist gets this fully right (too many variables, too long a horizon), but three tests do most of the work for me, and I think they are the right starting point:

1. Does the field reward judgment that takes decades and cannot be compressed?

2. Is the junior-to-senior ladder still being funded? No juniors means no future seniors.

3. Does credentialing intensity or regulatory accountability prevent the AI shortcut at the entry tier?

The defensible fields by all three: medicine and surgery, senior engineering disciplines where firms still invest in junior training, specialty research, track-record-based investment management, skilled trades outside the AI-cognitive frontier. The cautious fields are the ones the shock is hitting hardest at the junior end: financial compliance, legal junior-associate work, junior software engineering, content production, white-collar BPO. The ladder breaks at the bottom rung; the field shrinks even where the senior tier remains scarce.

Paul Kedrosky is right that “unprecedented in human history” is a small-sample argument; the compression dynamic does not need precedent to operate. This is the elite-overproduction half of Peter Turchin’s framework [13]: a credentialed professional class larger than the elite positions available to absorb it, alongside popular immiseration. A cognitive class compressed into a star tier and an interchangeable tier across four mechanisms at once is what elite overproduction looks like in real time. Turchin’s evidence base does not predict revolution; it predicts volatility, delegitimisation, and a search for scapegoats. A generation of credentialed professionals is walking into that configuration without the vocabulary to recognise it. Some of them are reading this. I am one of them, having run the three filters above on my own field and chosen not to elaborate on which is missing.

The signature is what is being bought. The draft is now cheap. The audit profession ran one of the four mechanisms first, between 2002 and 2026; three more are running through the cognitive economy in parallel, in different costume, without the legislative cue that organised the first. When verification and cumulative judgement are the scarce inputs across all four, a new economic class forms around their certification, and a market emerges priced around it. Each of these four mechanisms warrants its own deep treatment, and future notes will take them on individually, starting with recursive output proliferation. I named the four mechanisms I observe today; further mechanisms may surface as the AI impact on the economy continues to take shape, and my framework will be extended as they do.

References

[1] Sangeet Paul Choudary, Reshuffle, platformthinkinglabs.com, 2024-2026. The coordination-layer thesis.

[2] Alex Imas, “What Will Be Scarce?”, Ghosts of Electricity, 2026. The relational-sector reframe of post-commodity scarcity.

[3] Arvind Narayanan and Sayash Kapoor, AI as Normal Technology, Knight First Amendment Institute at Columbia, 15 April 2025. The three speed limits of invention, application, and diffusion.

[4] Mata v. Avianca, Inc., 22-cv-1461 (S.D.N.Y. 2023). Canonical first sanctions case; subsequent sanctions through 2024-2026 in jurisdictions including Texas, New York, and California are documented in legal trade press and federal court orders.

[5] “Unprecedented ‘Jobless Boom’ Tests Limits of US GDP Expansion”, Bloomberg, 18 February 2026. 2025 GDP growth and the 181,000 jobs added per BLS Employment Situation and BEA NIPA tables.

[6] Anthropic ARR of ~$30bn and headcount ~3,500 per Reuters / Bloomberg, April 2026. Salesforce fiscal-2024 revenue $34.86bn / 72,682 employees per Salesforce 10-K, February 2024.

[7] Stray Narratives, Issue 01: “The Verification-Substitution Matrix”, February 2026. The Trust Economy quadrant.

[8] Stray Narratives, Issue 02: “Adoption Asymmetry”, March 2026. The bargaining-power-shift mechanism: wage suppression arrives first through eroded leverage, not direct displacement. The phonograph case study is set out in the body.

[9] Stray Narratives, Issue 04: “The Channel Controllers”, March 2026. AI capability absorbed by channel controllers as toll collectors and platform operators.

[10] Stray Narratives, Issue 10: “What We Got Wrong About the Internet (and May Get Wrong Again)”, May 2026. The four-mechanism passage on which this piece’s framework is anchored.

[11] Stray Narratives, Issue 06: “The Noise Economy”, April 2026. The 13% / 77% workforce split.

[12] Alex Imas and Kristóf Madarász, Mimetic Dominance and the Economics of Exclusion, Quarterly Journal of Economics, 2024. Willingness-to-pay doubles with exclusion.

[13] Peter Turchin, End Times: Elites, Counter-Elites, and the Path of Political Disintegration, Penguin Press, 2023. The two-precondition framework (popular immiseration plus elite overproduction) is developed across the book; the empirical base across 20+ societies and 5 centuries is set out in the introductory chapters.

[14] Search referrer trend data for AI Overviews displacement of organic traffic to news and reference publishers, 2024-2026, as reported by Similarweb, Press Gazette, and publisher-side disclosures (BBC, Mail Online, News Corp earnings calls).

[15] Salesforce / Bloomberg / consumer AI-search ingestion of third-party sell-side research: industry trade reporting in Institutional Investor and Risk.net, 2025-2026, documenting analyst-research summary surfacing in chatbot products and the resulting compression of research half-life.

If you have no time to read and prefer to sit back, relax, and listen to this article in a podcast format by Notebook LLM (only 5 minutes long!):

The trade

Buy Xunlei (NASDAQ: XNET) for 6 US dollars before the 11 June 2026 lock-up unlocks on its Insta360 stake [1]. You are paying $398m for a Chinese internet operator whose Shanghai-listed equity stake is worth $986m at the 13 May close [2], whose net cash is $230m after netting bank debt [1], and whose operating business is worth another $410m–$745m on current Chinese-internet trough multiples.

That is a 78% discount to net asset value with a hard catalyst date and a published lock-up schedule.

Twenty-two years in the attic

Xunlei was a download manager in 2003, the first to make peer-to-peer file transfer fast enough to matter in China. It IPO’d on Nasdaq in 2014, lost the founder in 2017, pivoted into cloud, then live streaming, then short-form video, then accelerator subscriptions. Each pivot left a residual asset on the balance sheet.

One residual was a strategic equity position in Insta360, taken in the early 2010s when Xunlei was still a venture-portfolio business. It compounded silently while management chased four other things. Insta360 IPO’d on Shanghai’s STAR Market on 11 June 2025. Xunlei held 8.73% pre-offering, 7.84% after dilution, and approximately 7.8% at year-end 2025 per the audited filing [1].

This is what fifteen years of Chinese-internet pivoting looks like. It looks like an attic full of strategic equity stakes nobody is paying attention to. Most of them are worth nothing. One, by accident, is worth more than two and a half times the entire company holding it.

What you are buying, before the camera stake

Skip the camera stake for a moment. Pretend it does not exist.

For $398m you get a Chinese internet operator with $462m of revenue growing 42.5% in 2025 [1], $230m of net cash on the balance sheet, and a real operating business in three pieces.

Subscriptions. Xunlei Accelerator. $155m revenue, +16% growth, mid-twenties operating margins. Closest comps in the current trough are Tencent Music at 3.2× P/S, Bilibili at 2.2× P/S, iQIYI at 0.3× P/S [3]. At 1.5–2.5× revenue today, the subscription cash cow is worth $230m–$390m. Through-cycle (when the comps re-rate to 2020–2022 levels) the same business is worth $385m–$545m. Either way, it alone clears more than half the market cap.

Live streaming and IVAS. $170m revenue, +97.5% growth, including the inorganic Hupu contribution from Q2 2025. Hupu is the leading Chinese sports media platform, acquired May 2025 for RMB 500m, which sets a $69m floor under that piece. Closest comp: JOYY at 1.4× P/S [3]. At 1.0–2.0× revenue today this segment is worth $170m–$340m. Through-cycle, $425m–$680m.

Cloud. Twenty per cent residual in Onething, after Xunlei sold the controlling 50% to Kingsoft Cloud for RMB 125m in March 2026 [1]. Pro-rata value: $10m–$15m. The cleanest piece of capital allocation management has done in years was deciding they were not the right operator. They walked.

Operating business at current trough multiples: $410m–$745m, midpoint $580m. Through-cycle: $820m–$1,240m. We anchor the headline visual to the trough-midpoint because we want a number a sceptical reader cannot dismantle on a Bloomberg comp screen. The honest disclosure is that the trade improves substantially if Chinese-internet multiples re-rate.

The operating business and net cash exceed the market cap before Insta360 contributes a dollar. Even at the trough-low end ($410m of ops biz), net cash plus operating business clears $640m, against a $398m market cap. The camera company is technically free. It is also actually worth most of the value.

The Insta360 stake, the kicker

Insta360 holds 67% of consumer 360-degree cameras and is now joint leader in action cameras alongside DJI [4]. GoPro, the once-dominant incumbent, is the third name and falling. Q2 2025 revenue grew 41% to $190m. Trailing-twelve-month growth: 77%. The stock closed at 214 yuan on 13 May 2026, down 43% from its September peak of 378 yuan, well above the IPO offer price of 47 [5]. The chart is messy but the business is fine. Check out this video if, like me, you are too old to know what Insta360 is.

Xunlei’s stake: approximately 7.8% per the FY2025 20-F, equivalent to 31.28m Arashi Vision shares [1][2]. Audited carrying value at 31 December 2025: $1,051m. Mark at 13 May 2026: $986m.

The lock-up bucket is the standard 12 months under STAR Market rules. The 36-month rule applies only to controlling shareholders (Shenzhen Yiwei at 29.94%); the 24-month rule applies only to the underwriter’s investment arm. Xunlei sits in the 12-month bucket per the prospectus shareholder-commitment table [6]. The lock-up unlocks on 11 June 2026, less than a month away.

After the lock-up, Xunlei can sell. PRC corporate income tax (25%) plus Cayman repatriation withholding (10%) plus capital-control timing means roughly fifty cents of every gross dollar reaches ADR holders in the central case [7]. Insta360 is the kicker on a position the operating business already underwrites. You get the cake and the icing.

The risks honestly

Five concerns. They are also, mostly, the discount decomposition.

The lock-up could be voluntarily extended. It is not, per primary text. Xunlei is in the standard 12-month bucket. Management’s Q3 2025 commentary, “no plans to sell Arashi Vision shares due to regulatory requirements,” is consistent with the active 12-month restriction. Q4 commentary shifted to “determine the pace based on capital market conditions,”which is the right vocabulary for a board preparing to sell.

Insta360’s share price could correct further. It could. The doomsday case stress-tests against an IPO-band retracement to 70 yuan. There the trade loses money. I am not pretending otherwise.

Management never returns the cash. This is the structural risk. Twenty-two years without a dividend. Cumulative buyback through end-2025: $6.5m on a $20m authorisation [8]. At that pace, the authorisation will be exhausted some time in 2031. The market is pricing this management like the founder is still in the chair. He has not been since 2017. Current CEO Jinbo Li, in role since April 2020, launched the first two buyback programmes in company history (2023 and 2024). The pace is glacial. The direction is right. The Sohu/Changyou going-private at an 82% premium in April 2020 [9] is the comparable success precedent for a Chinese-ADR-holdco-to-portfolio monetisation. It is one data point, not a frequency.

Capital controls bite. They do. I model them in each scenario. The bear case (20% reach-through) is already heavy.

HFCAA delisting. Currently dormant. The PCAOB removed mainland China and Hong Kong from its inspection blacklist in December 2022 [10]; Xunlei does not expect to be identified as a Commission-Identified Issuer. The base case does not require this to remain dormant. Reactivation is the high-asymmetry tail.

The discount sits at 78% of net asset value because three reasons unwind mechanically (lock-up inertia, dormant HFCAA cohort drag, small-cap liquidity friction) and two are structurally priced (capital controls, management-distribution discipline). The market may be right that current management will inherit founder-era inertia. It may also be looking at the wrong twenty-two years.

How to express the trade

Stock: Linear, lower-leverage, suitable for a multi-month hold through the monetization window.

Call options: Higher leverage, time-bounded, binary downside. June, July, October, and January 2027 expiries each capture a different stretch of the catalyst window. Out-of-the-money calls magnify the stock returns 3–5×, with proportional risk of going to zero. Calls magnify everything, including being wrong. Reader picks the variance.

Expected return

Today’s $6.33 close is the entry. Hold through the 18–24 month monetisation window after 11 June 2026.

The visual headline of 4.5× is the gap between current price and net asset value. The realised return is the gap times the friction stack: tax leakage (25% PRC CIT + 10% Cayman withholding ≈ 50% reach-through), partial monetisation (30–40% of the stake sold over 18 months in our central case), and the residual unmonetised stake remaining exposed to whatever Insta360 trades at after the window closes. The visual depicts NAV. The IRR depicts what survives the friction stack.

Bull (Insta360 to ~300 yuan, 60% of stake monetised in 18 months, 50% reach-through, ops biz re-rates toward through-cycle): 125–160% per annum. Calls magnify 3–5×.

Base (Insta360 holds ~214 yuan, 35% monetised over 18 months, 50% reach-through, ops biz partially re-rates): 55–75% per annum.

Bear (Insta360 to 130 yuan, 20% monetised, 30% reach-through, no ops re-rating): even the bear case has a positive 15–25% per annum.

Doomsday (Insta360 to 70 yuan plus HFCAA escalates plus management distribution stalls plus operating margin compression): −15% to −30% per annum. This trade can lose money.

The path is staged: the first four weeks carry a discount-narrowing flow as the market reads the catalyst date; the next twelve carry the actual monetization; the eighteen to twenty-four months carry the operating-business compounding alongside.

Re-read points: Arashi Vision below 130 yuan or above 250 yuan changes the math; the first disclosed Insta360 sale by Xunlei confirms management willingness; Q1 and Q2 2026 earnings.

Bottom Line

A Chinese-domiciled internet operator trading at less than a third of audited net asset value. A Shanghai-listed equity asset on the balance sheet worth almost two and a half times the entire market cap. A subscription cash cow alone worth more than half the company is selling for. A four-week catalyst published in a prospectus.

Not a free lunch, there are none, but worth a closer look don’t you think?

Position disclosure

At time of publication, I hold a long position in XNET expressed via stock and call options.

References

[1] Xunlei Limited, Form 20-F for FY ending 31 December 2025, filed with the SEC on 28 April 2026. Consolidated Balance Sheets (cash $157.0m + short-term investments $148.2m + restricted cash $0.8m = $306.0m liquid; bank borrowings current $37.2m + non-current $38.4m = $75.6m). Note 6 (long-term investment in Arashi Vision: $1,050.6m fair value at 31 December 2025; approximately 7.8% equity interest). Note 12 and segment-revenue disclosure for Subscriptions, Cloud Computing, Live Streaming.

[2] Mark-to-market 13 May 2026: 31.28m Arashi Vision shares (= 401m total shares × 7.8% per Stock Analysis / Yahoo Finance public records) × CNY 214 closing price ÷ USD/CNY 6.791 (EODHD) = US$985.5m.

[3] EODHD fundamentals pull, 2026-05-13 close. Tencent Music P/S 3.24×, Bilibili P/S 2.21×, iQIYI P/S 0.30×, JOYY P/S 1.38×, NetEase P/S 4.83×.

[4] Mordor Intelligence, 360-Degree Camera Market; Grand View Research, Action Camera Market. Market-share commentary on Insta360 / DJI / GoPro.

[5] Stock Analysis (stockanalysis.com), Arashi Vision Inc. (SHA: 688775); Xunlei Limited (NASDAQ: XNET). Share price history and trading metrics.

[6] Lexology, In review: governing rules for IPOs in China. Hankun Law and KPMG STAR Market lock-up framework summaries; Arashi Vision STAR Market listing announcement (上市公告书), 11 June 2025; 36氪, “影石创新科创板上市”, 11 June 2025.

[7] PwC, China Tax Summaries: Corporate Income Tax, capital-gains treatment of listed-equity sales by PRC corporate sellers; Cayman / Hong Kong dividend-withholding treatment.

[8] Investing.com, Xunlei sets new $20 million share repurchase program, 4 June 2024; subsequent quarterly buyback disclosures in 20-F.

[9] Sohu.com Limited, Sohu.com Announces Completion of Changyou Going-Private Transaction, PR Newswire, 17 April 2020. Comparable Chinese-ADR-holdco-to-portfolio monetisation precedent.

[10] U.S. Public Company Accounting Oversight Board, Statement on PCAOB inspection access in mainland China and Hong Kong, 15 December 2022.

Research, not investment advice. Stray Narratives is the publishing arm of an investment-research operation; positions are disclosed alongside the analysis where they exist. Past performance does not guarantee future returns. Full disclaimer at straynarratives.com/p/disclaimer.

• The historical record on technology shocks shows a recurring pattern: the first-order impact people argue about turns out to be smaller than the second-order impact almost nobody saw coming.

• Three precedents (the talkies of 1930, the IT consensus of 1973, and the internet of 1995) all show the pattern, and the AI debate of 2026 already has the look of a fourth iteration.

• Issue 06 The Noise Economy introduced friction elimination as one form of unforeseen second-order effect; this issue argues that noise itself, the recursive proliferation of AI-generated outputs, is plausibly another, and there will be more.

• The contemporary anchor is the 2025 divergence: US GDP grew 2.2% with only 181,000 jobs added all year, the worst figure for any non-recession year of expansion since 2003, and Anthropic in 2026 is producing nearly the same revenue as Salesforce did in 2024 with around five per cent of the headcount.

• The political response is already visible in New York’s vote for Mamdani, where voters experienced the divergence directly: GDP rising, markets soaring, material lives stuck.

Here below is a podcast version of this article produced with Notebook LLM, hope you enjoy it!

Twenty-four thousand musicians

In 1927, The Jazz Singer arrived in cinemas and the live-music industry began to evaporate. Roughly twenty-four thousand musicians worked in theatres across the United States and Canada at the time, accompanying silent films night after night with whatever combination of piano, violin, drums, and brass the local theatre could afford. By 1930, the American Federation of Musicians estimated that around twenty-two thousand of those jobs had vanished [1]. Almost the entire trade, in three years. In some markets the destruction was even faster: New York and Cincinnati both lost between half and three-quarters of their working musicians inside the first two years of sound.

They did not go quietly. In 1930 the American Federation of Musicians spent the equivalent of ten million dollars (in today’s money) on a national campaign begging audiences to refuse “canned music” and to keep going to theatres that still employed live performers. Joseph N. Weber, the union’s president, predicted that the public would never accept “like-less, soulless, synthetic music.” Edward More, music critic of the Chicago Herald Tribune, agreed: the films, he said, “have a long way to go before they can duplicate living musicians.” The whole defence rested on a single load-bearing assumption, which is that the new technology would have to match the old one to displace it. The films had to play violin.

They never did. The films did not have to. As Dror Poleg put it in his essay on this episode, disruptive technology rarely seeks to replicate. It sidesteps and makes the old standards redundant [2]. Recorded sound made music cheap and accessible to a vastly larger audience, and most of that larger audience turned out to be perfectly content with what Weber had dismissed as soulless. Weber was, in a narrow sense, proved right: the public never did warm to soulless synthetic music. They simply turned out to be indifferent to the difference, which proved entirely sufficient. The threat to the trade was never a robot violinist. It was a different technology entirely, one that bypassed the contest the musicians thought they were in.

The same script ran a generation later in print encyclopedias. Britannica had been the authoritative reference work for two hundred and forty-four years when, in 1993, Microsoft launched Encarta, a digital encyclopedia replicating Britannica with multimedia and a CD-ROM price point. The threat looked clear, and the threat was the wrong one. Eight years on, when the actual challenger arrived, it was Wikipedia: free, crowd-sourced, infinitely updatable, and not trying to be Britannica at all. Encarta died in 2009. Britannica stopped its print edition in 2012. The threat to two and a half centuries of authoritative reference publishing came from a different category of object entirely, one that had no interest in being authoritative in the old sense.

This is the pattern of unforeseen second-order effects, and the reason the talkie episode and the encyclopedia episode keep repeating themselves in the historical record is that they are a feature of how technological transitions actually work, not a one-off. A useful shorthand for what’s happening: call X the predicted impact of a new technology, the thing the consensus debate of the time fixates on; call Y the actual dominant impact, the thing that turns out to matter most in retrospect. For the talkies of 1930, X was “machines that play music well enough to replace the live ones” (which never had to come), and Y was “recorded sound, sidestepping live performance entirely.” X and Y rarely coincide. Y is usually bigger. And Y is almost always something nobody at the time of X was writing about. The argument that follows is that we are now in the next iteration of the same pattern, with AI in 2026 standing in for the talkies in 1927, and the internet in 1995 standing in as the most recent precedent of how badly the X-versus-Y misalignment can go.

The 1973 mirror

Forty-three years later, the consensus was the symmetric opposite, and it was wrong in the same way.

Hobijn and Jovanovic, in their 2000 paper on the IT revolution and the stock market, record the 1973 expectation as a stylised assumption in their model: information technology, the consensus held, would substitute for white-collar cognitive labour the way mechanisation had substituted for blue-collar physical labour [3]. The subsequent record inverted it. IT complemented cognitive workers rather than displacing them; the cognitive-skill premium widened through the 1980s and 1990s; the wage gini diverged. The bigger story, the one nobody was writing about in 1973, turned out to be the personal computer, the web, social media, the attention economy, the wholesale reorganisation of how information moves through a society. The first-order story (white-collar substitution) was the consensus debate. The second-order story (the reshaping of attention itself) was the one that mattered, and it was almost entirely unforeseen.

This is the pattern of unforeseen second-order effects, and the reason the talkie episode and the encyclopedia episode keep repeating themselves in the historical record is that they are a feature of how technological transitions actually work, not a one-off. A useful shorthand for what’s happening: call X the predicted impact of a new technology, the thing the consensus debate of the time fixates on; call Y the actual dominant impact, the thing that turns out to matter most in retrospect. For the talkies of 1930, X was “machines that play music well enough to replace the live ones” (which never had to come), and Y was “recorded sound, sidestepping live performance entirely.” X and Y rarely coincide. Y is usually bigger. And Y is almost always something nobody at the time of X was writing about. The argument that follows is that we are now in the next iteration of the same pattern, with AI in 2026 standing in for the talkies in 1927, and the internet in 1995 standing in as the most recent precedent of how badly the X-versus-Y misalignment can go.

A note on what Hobijn and Jovanovic actually do here. Their paper records the 1973 expectation as an empirical fact about what the consensus of the time held. The point of citing them is the observation that the consensus of the time was directionally wrong, and that being directionally wrong about a technology shock has been the historical baseline.

The scope, and what Issue 06 already mapped

A scope limit before going further, because the temptation to over-extrapolate from one’s own corner of the economy is almost the defining vice of writing about AI from inside the knowledge professions.

Issue 06 The Noise Economy established the relevant numbers [9]. Around thirteen per cent of the US workforce sits in roles with high friction-elimination exposure, the proceduralised cognitive layer that the AI debate fixates on. Seventy-seven per cent sits outside that frame entirely: food service, retail, manufacturing, transport, healthcare support, government, education. The mechanisms I’ll develop below are about the knowledge economy. Many sectors will be little affected by what feels, from inside the affected slice, like a massive threatening revolution. I’m inside that slice. You probably are too. That fact, on its own, ought to make both of us a degree more cautious about the breadth of any claim that follows.

Issue 06 also did half the analytical work this issue extends. Friction elimination was the argument that a whole layer of professional activity exists not because it creates value at the point of transaction but because it navigates an opacity AI is now liquidating. Take the opacity away and the activity disappears, with no displaced worker because there was nothing left to displace. Noise, the candidate this issue offers, is a different mechanism in the same family: friction elimination removes activity; noise multiplies it. Both are shapes of unforeseen second-order effect, both bypass the displacement debate, and both belong to the same affected slice. The working assumption is that there are more shapes still ahead.

The unforeseen Y, applied to AI

The first-order story for AI is the displacement-at-speed-X debate that fills the column inches. Will agents replace lawyers in three years or fifteen? How many call-centre seats will be gone by 2027? The debate is real, the numbers matter, and Issues 01 through 06 of this series have been engaged in it. The question this piece sets down is whether, as in 1930 and 1973, the dominant effect turns out to be something quite different and quite a lot bigger.

The candidate this issue offers is noise, and the closest precedent for what noise actually does to a working life is something every reader has lived with for thirty years.

Email is the precedent. Email was sold as a productivity tool, and it was a productivity tool, for a few months. Then the volume became unconstrained. Anyone could send anyone a message at zero marginal cost, and the cost of that asymmetry has accumulated into a working day in which knowledge workers spend a third or more of their hours triaging messages they did not ask for. Email did not replicate paper memos; it made the entire structure of paper-memo norms obsolete. The unforeseen Y of email, in 1995, was not the productivity gain we expected. It was the productivity tax we now live with. AI noise is the next iteration of the same dynamic, run at a generation’s worth of speed.

A few mechanisms underneath that label, observable now:

• Recursive output proliferation. AI outputs become inputs become outputs. The half-life of any specific knowledge advantage collapses, because what one analyst learns this quarter is in everybody’s training set the next. The premium on having access to a piece of analysis falls because the cost of producing equivalent analysis falls. There is a great deal more of everything, and a great deal less to distinguish any one piece from any other. The shape of the resulting market is one in which volume rises and average per-unit value falls, perhaps a long way.

• Constant experimentation as time-sink. Every team running parallel trials on tools, workflows, agents, and models. Time spent evaluating what to use displaces time spent on the work itself. There is a tax on attention that takes the form of having to keep up with the platform churn, and the tax compounds quietly because the comparison to the prior tooling regime fades from memory. People remember being busy; they do not always remember whether the busyness was productive.

• Verification burden. When an AI can produce anything, verifying becomes the bottleneck. Citations may be hallucinated, code may compile and still be wrong, drafts may pass spell-check and still be incoherent on a re-read. Concrete cases are now a regular feature of professional practice. Lawyers in Texas, New York, and California have been sanctioned in the past eighteen months for filing briefs containing AI-fabricated citations, including, in one case, a complete fictitious appellate decision the firm had not noticed before submission. Federal judges now routinely spend several paragraphs of their sanction orders explaining, for the benefit of the bar, that the cited authorities do not exist, a sentence that until 2023 would have been a parody of itself. Software teams report a steady stream of AI-generated code that compiles and ships and is later discovered to have invented an API call that does not exist or a function whose stated behaviour does not match its actual logic. Hiring managers describe an arms race of AI-generated CVs and cover letters tuned to defeat applicant-tracking systems, where the marginal cost of producing an apparently-qualified candidate has fallen close to zero. In each case, signal-discrimination becomes the scarce capability, and it is exactly the capability that was previously externalised onto the production process: the editor, the legal reviewer, the senior analyst whose job was to know which numbers to trust. Strip out the production friction and the verification cost shifts up the chain, where it is harder and more expensive to apply at scale.

• Attention fragmentation. Email becomes agent-to-agent. Search becomes AI-mediated. Content becomes infinite. The bottleneck is no longer capability but choice: what to attend to, in a stream that has lost most of the curatorial gates that used to do that work for us.

The recent precedent for what this dynamic does at cultural scale is social media, and the version of that precedent worth getting right is one the consensus often blurs. Social media as a mass cultural phenomenon was not, on its own, an internet story. The internet was the substrate. The carrier turned out to be the iPhone, twelve years later: a device nobody in 1995 had thought about, which made always-on access universal and turned social media from a niche curiosity into the dominant capture mechanism for human attention. The unforeseen Y, in retrospect, was the smartphone, not the internet. AI’s equivalent carrier is not yet named.

What the four mechanisms above describe is what current AI noise looks like in working life. They are diagnostic, not prophetic. The harder claim, and the one this piece commits to, is that the cultural and economic centre of gravity of the AI era will form somewhere none of the four mechanisms describes.

The pattern of which technologies actually scale into mass cultural phenomena is consistent enough to be falsifiable. A technology becomes culturally dominant when it satisfies three conditions simultaneously: an engagement loop that captures attention, network effects that compound value as adoption rises, and a clear path to monetisation. The internet on its own satisfied two of the three. The internet on the iPhone satisfied all three, and the cultural revolution that followed was the result.

AI in its current form, foundation models accessed through chat interfaces, satisfies one or two of the three. Foundation-model companies have (still to be proven at scale) monetization , weak network effects, and largely utilitarian rather than addictive use. Anthropic and OpenAI are fundamentally building infrastructure businesses, with the engagement loops as a secondary consideration. That may become a profitable category but not the category that becomes the dominant cultural force. Where AI satisfies all three is the application layer that Tristan Harris has begun calling the attachment economy: AI companions, AI relationships, embedded AI tutors, AI confidants [8]. Engagement loops, in their most direct form, the relationship-attachment kind, run a subscription clock that does not stop. Network effects emerge as the AI accumulates context about its user and the user invites the relationship into shared spaces. Monetisation is the subscription itself, plus the data, plus the ecosystem lock-in. All three conditions, in the same product category. Replika is the early version. Character.ai is another. Most of the specific companies in this space will not survive in their current form. The category they sit inside almost certainly will.

The practical test follows from the framework. When you read about a new AI product in 2026, ask three questions: does it create an engagement loop, does it benefit from network effects, does it have a clear monetisation path. If the answer is yes to all three, you are looking at a candidate Y. If the answer is yes to one or two, you are looking at infrastructure. The current investment debate is almost entirely focused on infrastructure. The category that almost nobody is arguing about, because it still sounds small from inside the 2026 vocabulary, is where the cultural and economic centre of the AI era almost certainly forms. The closest a careful reader can get to the actual unforeseen Y is to look at which technology is starting to satisfy all three conditions, and to take that category more seriously than the infrastructure debate currently does.

The 2025 divergence

The unforeseen Y is already partly observable, and the data are starting to fill in.

US GDP grew 2.7% in 2025. Net employment over the same year was 181,000 jobs added, the worst figure for any non-recession year of expansion since 2003 [4]. Bloomberg published a piece in February 2026 under the headline Unprecedented “Jobless Boom” Tests Limits of US GDP Expansion, and the accompanying chart showed real GDP and nonfarm payrolls (two lines that moved in near-lockstep for over three decades) diverging in a way that, in Bloomberg’s words, “has never happened this far into expansion” [4]. What makes the configuration anomalous is that this is not a recession picture. There is no economic contraction holding payrolls down. The expansion is real, and it is happening without the headcount that used to come with it.

The killer specific is the per-employee revenue stat. In fiscal year 2024, Salesforce, by some distance the most successful pure-software company of the last decade, generated nearly thirty-five billion dollars of revenue with around seventy-three thousand employees, which is roughly half a million dollars of revenue per worker, strong by any pre-AI software standard. Two years later, Anthropic is producing roughly thirty billion dollars of revenue at an annualised run rate, with around three and a half thousand employees, which is closer to eight and a half million dollars of revenue per worker. Same software industry, same broad addressable market, two years apart. Salesforce has roughly twenty times the headcount producing nearly the same revenue, and the per-worker output at the AI-native firm is around eighteen times what Salesforce achieved at peak [5]. Put another way: if Salesforce in 2024 had operated at Anthropic’s 2026 per-worker productivity, it would have employed about four thousand people. The remaining sixty-nine thousand jobs were not, by any 2024 measure, redundant. They were the standard cost of running a company. The economy underneath that comparison is one in which output and employment are decoupling at the company level for any firm willing to build around the new tooling.

Beyond the macro charts, the pattern is showing up at named-CEO level. Klarna disclosed in early 2024 that an AI assistant was doing the work of around seven hundred customer-service agents, and the company’s headcount fell from roughly five thousand to three and a half thousand over the following year while revenue rose. Duolingo’s chief executive said in January 2024 that around ten per cent of the company’s contractors had been let go because AI was now doing their work. Goldman Sachs has visibly compressed junior banking-analyst hiring through 2025, with multiple reports of cohort sizes shrinking and some entry-level functions being routed through AI tooling instead [6]. None of these is a layoff event in the conventional sense. They are quiet adjustments to the production function, and they accumulate.

What macroeconomists have been arguing should not happen this far into an expansion is happening. The production function underneath the headline numbers has changed, and the change is showing up in the GDP-versus-payrolls chart, in the per-employee revenue gap, and in a steady drumbeat of named-CEO disclosures that AI is now doing work the company used to staff for. The first-order story (AI substitutes for cognitive labour at speed X) is one frame. The second-order story (the relationship between economic output and human employment is being rewritten in real time) is plausibly bigger, and is happening regardless of whether you believe the first-order story or not.

The political response to the divergence is already visible. New York elected Zohran Mamdani as mayor in November 2025 in part because, as Dror Poleg observed, voters experienced the divergence firsthand: the economy was growing, the stock market was soaring, and yet their material lives were not getting better [7]. San Francisco, in the same season, elected a different sort of mayor and started liberalising zoning. Two cities, two different responses, both confirming that the structural break is real and that voters can feel it before economists are willing to call it.

A short personal note

I’ve spent thirty years in finance, which is one of the sectors most exposed to the noise mechanism I’ve just described, and I read for a living the kind of analysis that AI now produces in volume. The best calibration of any thesis I write on AI’s impact is to remember that I’m squarely inside the affected slice and probably overweighting my own experience by a margin I cannot quite measure.

I have tried to write this article under a new format where it should be easier to capture the subject of the post from the very beginning and also provide a summary at the end. If you like this new format please “like” the post, it means a lot to me.

Bottom Line

In 1995, the case for the internet was that it would let us work and consume more efficiently. It did, eventually, in that narrow sense. The bigger story took twelve more years to arrive, and arrived through a device nobody in 1995 had thought about: the iPhone, which made the always-on internet into a pocket-sized reality and turned social media from a niche curiosity into the dominant capture mechanism for human attention. The unforeseen Y, in retrospect, was not the internet at all. It was the smartphone.

The case for AI today reads almost identically to the 1995 case for the internet. The first-order story is well-understood and being argued about in every research note and every Substack. It is also, probably, not the story that will matter most in retrospect. Whatever the dominant impact of AI turns out to be, it is almost certainly not on anyone’s spreadsheet right now. Watch for what is being built with all this capability that nobody designed for. The next piece in this AI series asks the question that follows: when noise dominates, where does signal go, and who pays for it?

References

[1] American Federation of Musicians, Music Defense League records, 1930. The 22,000 jobs lost figure is the AFM’s own contemporaneous estimate of theatre-musician unemployment by 1930, and is the canonical historical record cited across the labour-history literature on the talkies transition. The campaign and its underlying figures were widely covered at the time, including in Smithsonian Magazine‘s retrospective on the Music Defense League.

[2] Dror Poleg, Who Will Get Canned in 2026?, drorpoleg.com, 30 December 2025. The “disruptive technology rarely seeks to replicate; it sidesteps and makes the old standards redundant” framing in Section 2 is Poleg’s, and the article’s reading of the talkies episode owes its shape to his piece.

[3] Bart Hobijn and Boyan Jovanovic, The Information Technology Revolution and the Stock Market: Evidence, NBER Working Paper No. 7684, May 2000. Assumption 2 of the model records the 1973 consensus that IT would substitute for white-collar cognitive labour the way mechanisation had substituted for blue-collar.

[4] Unprecedented “Jobless Boom” Tests Limits of US GDP Expansion, Bloomberg, 18 February 2026. The chart of real GDP versus nonfarm payrolls and the “never happened this far into expansion” framing are drawn from this piece. The 2025 GDP growth of 2.7% and the 181,000 jobs added (worst non-recession year since 2003) are from US Bureau of Labor Statistics Employment Situation reports and Bureau of Economic Analysis NIPA tables, both publicly available.

[5] Anthropic annualised revenue run-rate of approximately thirty billion dollars as of April 2026, and headcount of approximately three and a half thousand, reported by Reuters and Bloomberg, April 2026. Salesforce fiscal-year 2024 revenue of $34.86 billion and year-end employee count of 72,682, per Salesforce’s audited 10-K filing, February 2024.

[6] Klarna headcount and AI-customer-service disclosure: public statements by CEO Sebastian Siemiatkowski via X and Bloomberg interviews, 2024-2025. Duolingo contractor reduction: public statement by CEO Luis von Ahn, January 2024 earnings communications. Goldman Sachs junior banking-analyst hiring compression: multiple Bloomberg and Financial Times reports through 2025.

[7] Dror Poleg, Welcome to the Jobless Boom, drorpoleg.com, 18 February 2026. The Mamdani-divergence interpretation cited in the 2025 divergence section is from this essay.

[8] Tristan Harris, Center for Humane Technology. The “attachment economy” framing extends the engagement-economy critique of social media to AI specifically, with AI companions and AI-mediated relationships as the central category. From Harris’s recent public appearances and writings.

[9] Stray Narratives, Issue 06: The Noise Economy, 12 April 2026. The friction-elimination argument with the 13% / 77% workforce split is developed in this issue.

The mandate event, already priced

The market voted before it counted the votes.

Magnum was distributed to Unilever’s shareholders the way unwanted assets always are: handed across and let go of, fast, by funds whose mandates did not allow an Amsterdam-listed ice cream stock to stay on the book. By 22 April, the share price had drawn 34% off its February high. The Amsterdam line sat as the eighteenth-most-shorted name in the STOXX 600 at 19% of free float (the share count actually available for trading), and the listing-day valuation already trailed a European broker’s pre-listing fair-value range by €2-3bn.[3][3a][3b][4] The supply was being priced before any operating evidence had a chance to land.

Then the operating evidence landed. The Q1 2026 trading update on 30 April printed organic growth of +4.5%, volume +2.9%, with Europe & ANZ at +4.3% volume on -0.3% price. That is unit pull, not promotional defence. FY2026 guidance was reaffirmed.[24] The price moved on the print, the cleanest leg of the near-term trade taken in a single session, much of it absorbed by short-cover.

The forced-seller overhang is no longer hypothetical. It has landed, it has been priced, and it has begun to unwind. That changes the question. The bull case is no longer a claim that the forced-seller compression will happen. It is a claim about what comes after the compression that has already been paid for.

Worth stating out loud: a Q1 trading update discloses revenue. It does not disclose margin. The bull case turns on margin. So the print confirms the volume thesis the bears were betting against, narrows the structural claim by exactly that much, and leaves the load-bearing question unresolved. The 1H 2026 release in late July is the first window in which the load-bearing question gets answered.

The category sits in pressure

Magnum is not a standalone case. The whole branded CPG aisle is under pressure. The US and UK middle-class consumer has traded down to discount and private-label brands at a pace the incumbent operators have finally been forced to admit on calls. The Wall Street Journal in November 2024 led on the fact: households earning over $100,000 are now the fastest-growing cohort at Aldi, Lidl, and the US dollar chains.[5] Six-figure households shopping at the discounter is not a recession story. It is a structural one. Private-label has moved from recessionary backstop to default preference, and has survived wage growth and improving sentiment without giving any of the share back.

The issuer side runs parallel. Kroger’s 10-K carries Our Brands sales above $32bn at 90%+ household penetration with 900 new items in the year, which is the language of a strategic asset, not a defensive line.[6] Kraft Heinz, P&G, Unilever, and Colgate have each said versions of the same thing on calls for the past two years: the price gap to private label has widened past the sustainable range.[7] Bronnenberg, Dubé and Sanders’ 2018 NBER work on scanner data, including ice cream as a tested category, found that consumer blind-test preference does not distinguish private-label from national brand.[8] That finding is a decade old. What has changed since is the arrival of retailer cohorts that have professionalised private-label packaging to a point where the informational gap has narrowed.

The ice-cream evidence is issuer-recorded, in the parent’s own voice on the parent’s own call. Hein Schumacher, then-CEO of Unilever, on the Q4 2023 results call, verbatim: “very disappointing year for Ice Cream with price elasticity in the in-home channel much more negative than that seen in other categories and with a strong consumer down-trading to private label.”[9] That is Magnum’s inherited book.

The investment case turns on a single question. Are Magnum’s specific moats (brand occasion, cabinet distribution, category archetype) the exception inside the pressure, or the next item on the list?

The near-term trade

Issue 01 mapped what we called the Kill Zone: when a product is easy to compare on price and quality, and cheap to switch between brands, one or two players end up taking nearly all the market.[23] An AI-powered shopping assistant that reads ingredient lists and price gaps at the shelf makes products more comparable by definition. Functional FMCG sits inside the zone. The middle-class downtrading documented above is not only cyclical; it is verifiability shifting under the brand premium.

Ice cream is not a uniformly functional category, which is the Magnum-specific hinge. The franchise outside the in-home tub (the supermarket multipack eaten at home) is emotional, seasonal, experiential, impulse-weighted. The purchase is not an optimisation problem the assistant resolves, because the consumer is not at the shelf with time to read. That is the structural defence the bull leans on. The bear reads the same fact pattern as a partial defence at best: the in-home tub channel is c. 60% of revenue, and that is where the 2023 trade-down (the moment branded ice cream lost shoppers to private label on price) landed.

The forced-seller arithmetic is where bull and bear briefly agree. About 60% of the float had to be sold inside six to twelve months. Natural buying capacity was around half of that.[4] Between 8 December and 22 April, the stock did what supply-and-demand demanded. The 34% drawdown, the 19%-of-float short register, and the multi-billion-euro discount to broker fair value are the same fact in three notations.

A 19% short position at the eighteenth-most-shorted seat in the STOXX 600 is a crowded trade. Crowded bear trades work until they do not, which is to say they work until you find yourself wedged in next to seventeen other people on the same side of the boat. That is what 30 April looked like in the post-print session.

The next forced flow is a buy. Once Unilever clears the first batch from its retained 19.85%, the share count freely available for trading rises enough to lift Magnum into full benchmark weighting on STOXX 600, MSCI Europe, and FTSE 350. The index funds tracking those benchmarks then have to buy. That mechanical buying is typically 1-3% of market cap inside a 1-3 day window, which on today’s level implies $80-240m of committed flow into a register that the short position has already squeezed.[4]

The base case from here is +5-15% over six to twelve months on residual overhang absorption, first-tranche disposal, index recalibration, and short-cover on any further upside catalyst. The range is -10% on disposal slippage plus a Q2 miss to +25% on a first-tranche announcement landing inside the inclusion window. Both bull and bear concede this leg.

The 12-month inflection

Past index inclusion, the consensus dissolves.

Two questions, defensible from the same fact base, define the bull and bear past 12 months.

The bull’s question. Do Fernando Fernandez at Unilever plc and Peter ter Kulve at Magnum execute the disclosed €500m productivity programme on the disclosed calendar, and does the resulting margin structure compound toward a 19-22% adjusted EBITDA band by FY2028-FY2030?

The bear’s question. Does category drift (cabinet atrophy, GLP-1 frequency compression, private-label compounding under verifiability pressure) reassert after the near-term mechanics unwind, and does the plc-form ceiling structurally cap the margin arithmetic below Froneri’s accessible range?

The remainder presents each at its strongest.

The longer-term bull

For ten years, the ice cream division ran an operating margin of 9.6-12.1%, capped by a matrix organisation that allocated marketing budget across detergents, deodorants, tea, and ice cream from the same group pot. Cabinet investment in Manila lost to deodorant media in Detroit, year after year, in board meetings the ice cream people did not always attend. That is not a management failure. It is what happens to ice cream inside a diversified FMCG conglomerate, and it is why ice cream never earned its category-native margin structure inside Unilever.

Magnum’s standalone machinery is now in place. Single-category board chaired by ter Kulve. Single-category P&L. The €500m productivity programme is disclosed in three tranches (supply-chain, overhead, technology), with €250m already delivered by end-2025, and the Q1 update reporting “good progress... on track for the full year”.[12][24] Half is in the bag before management has had a single contested earnings print to defend.

The reference point is Froneri, a privately-owned ice-cream maker (joint venture between Nestlé and the private-equity firm PAI Partners) operating in the same category with the same dairy, cocoa, and packaging cost base. Froneri runs at roughly 26% EBITDA margin on FY2024 revenue of €5.53bn.[13] On 2 October 2025, PAI completed a €3.6bn continuation vehicle at an enterprise value of approximately €15bn, 10.5x trailing EBITDA, with the operator-aligned capital electing to extend rather than exit nine years into the hold.[10][14] That is the market-clearing price for a dedicated ice-cream operator, paid by investors who have read the internal books.

The Froneri perimeter is now expanding. On 19 February 2026, Nestlé confirmed the phased sale of its remaining ice-cream operations (Canada, Chile, Peru, China, Malaysia, Thailand; ~CHF 1bn turnover; brands including KitKat ice cream) to Froneri.[25] The reference benchmark is consolidating in real time, on a calendar that overlaps Magnum’s first two earnings prints. That fact strengthens both sides of the argument: it confirms the destination the bull names andwidens the moat the bear says Magnum cannot replicate.

The margin bridge from Magnum’s FY2025 15.9% to Froneri’s 26% breaks into addressable layers and a structural residual.

Chart A — The margin bridge: Magnum FY2025 15.9% to a reachable 19-22% by FY2028, with Froneri's 26% as the ceiling.

The bands caveat the magnitude, not the direction. Roughly 400-700bp of the gap is addressable, 100-200bp is not, and Magnum has a three-to-five-year programme running at the addressable portion. At a 60% retention rate (high end of productivity-programme precedent, although ter Kulve already has €250m of the €500m in the bag), the reachable margin lands in the 19-22% band by FY2028-FY2030. Above company guidance, below Froneri by the plc-inaccessible residual, and above today’s entry multiple by enough to compound into terminal valuation. Half the bull case is paid for already. The other half is what the next two earnings prints test.

The cabinet network is the execution lever a dedicated operator with single-category decision rights actually has.[15]

Chart B — Cabinet economics: Magnum at €2,640 per cabinet vs Froneri at €3,950 per cabinet.

The bull case rests on two named executives. Fernandez sits at the plc parent, weighted toward the disposal calendar. Ter Kulve sits as the dedicated operator with a single-category P&L and a productivity programme on his desk of which half is already in the bag. The reader is asked to make a judgement on these two people, on the €500m disclosure, and on what the delivered €250m implies about the remainder.

The longer-term bear

On 27 August 2025, in a fireside chat, ter Kulve said in his own voice as the incoming CEO that the global cabinet fleet had been declining “approximately 2% a year” before the current reinvestment programme.[16] One sentence; the whole separation narrative, said unguarded. Three million cabinets contracting at sixty thousand a year for the Unilever decade is not an under-investment problem a productivity programme fixes. It is the physical signature of a channel quietly withdrawing floor while nobody at group level was looking.

Three mechanics compound the contraction. Consolidation in convenience retail (corner shops and petrol stations merging or closing) shrinks the number of locations willing to host a branded freezer. Migration of ice-cream sales into fast-food chains drops Magnum’s sticks into a case shared with Froneri and Häagen-Dazs. Grocery-aisle substitution of branded impulse sticks by own-label take-home tubs pulls the household occasion off the branded-cabinet page entirely. The conclusion is not a slower-growing fleet but a shrinking one in a channel whose three largest structural hosts are each quietly withdrawing floor.

The cabinet-specific capex disclosure, of which there was none across the Unilever decade, is the second thing worth saying. A parent that did not treat cabinet capex as a material disclosure theme for ten years is a parent that did not treat the cabinet as a strategic asset. Which is why the rebuild is not the release of an under-invested asset; it is a reconstruction project on something whose prior owner did not think was worth talking about.

The plc-form ceiling is the structural reason the drift reasserts. Three features make it a form-factor problem, not an operational one. Froneri’s pro forma leverage of 5.7x sits on the back of a June 2025 €3.9bn debt recap that funded a €4.4bn shareholder dividend.[18] Froneri’s incentive architecture is a single-purpose Luxembourg holdco with equity-linked management participation at operating-company level. Froneri’s governance is a closely held JV with two strategic owners and no quarterly disclosure to a free-float register. Each one of those is a structural advantage a triple-listed plc cannot replicate by design. The ceiling is the visible surface of an operating model a plc cannot copy.

The royalty-free Häagen-Dazs asymmetry is the second structural piece. Froneri operates Häagen-Dazs in the United States and Canada under a 99-year paid-in-full royalty-free licence running to 2110.[1][19] Häagen-Dazs holds over 38% of the US premium snacking segment at brand sales of roughly $1.4bn.[20] Froneri carries zero ongoing royalty cost on the single largest premium brand in the single largest premium ice-cream market. Every Magnum stick cabinet in North America competes at direct cabinet-level against a premium-brand operator whose brand cost on its largest US asset is effectively zero. The $641m Nestlé paid in December 2001 was the market-clearing price for perpetual access to the US premium-tier leader. Magnum does not have an equivalent asset, and no productivity programme can create one. The Froneri perimeter expansion announced 19 February 2026 widens the asymmetry, not narrowing it.[25]

GLP-1 (the Ozempic class of weight-loss drugs that suppress appetite) is the demand-side compounding layer. Adoption is tracking 2-3% of developed-market adult population by 2030 in the central scenario. Calorie reduction reported at 15-30% across all food categories. The bull frame bounds the category impact at 1-5% by FY2030 on calorific compression alone. The bear reading on the frequency axis is sharper. GLP-1 does not substitute the Magnum for something else. It eliminates the second and third Magnum from the weekly basket. One quarter of post-elasticity volume recovery does not settle that question; it narrows the mechanical claim and leaves the structural claim intact.

The disposal overhang is the five-year structural drag, not a twelve-month window.

Chart C — Disposal precedent timeline: Haleon, Kenvue, Solventum, Magnum.

The pattern is consistent. The first disposal is the start of the calendar, not the approach to recovery, which is why one operational beat is informative about the operating asset and silent on the disposal mechanics.[21][22]

The judgement

The market overpaid for the supply. The earnings will eventually price the asset. We now have one print’s worth of evidence that the volume thesis the bears were short on has narrowed.

The bear case kept the rest: cabinet contraction, plc-form ceiling, royalty-free Häagen-Dazs asymmetry, GLP-1 frequency compression past 2027. None of those is refuted by Q1. They are still on the page, and they remain the strongest possible test of the position.

I think the asset is attractive at this level. The forced-seller drag is mostly past. The dedicated operator has half the productivity programme banked. The reference comparable, Froneri, is widening its perimeter in real time. None of that requires the bear case to be wrong. It requires only that the operating evidence between now and FY2028 keep doing what the first print started to do.

I should have started buying earlier. Looking at the chart now, the eighteenth-most-shorted seat in the STOXX 600 was the trade, the 19%-of-float short register was the trade, and the broker-fair-value gap was the trade. The chance to take it without paying for the rerating-already-in-progress closed in a single April session, while I was still asking whether the operating evidence was reliable enough to lean on. The cleaner risk-reward window was the day before the print, not the day after.

I’m taking a position now anyway, on the bull case rather than the rebound. That ter Kulve delivers the €500m. That the cabinet network responds to single-category decision rights. That the index-inclusion mechanic provides the next forced flow upward. And that 18-24 months from here, the multiple gap to Froneri compresses by enough to compound past the entry-point premium I am paying for having waited.

None of that is settled by one quarter. None of that is what the position is taken on. The position is taken on the €250m delivered against the €500m promised, on the cabinet rebuild plus single-category decision rights, on the index-mechanic flow ahead, and on the 18-24 months in which the operating thesis either lands or it does not. The first 1H print, in late July, is the first window in which the position has to defend itself.

The expected return

Taking today’s level as the entry, and assuming margin lands in the lower half of the 19-22% band by FY2028 with Magnum trading at roughly 9x adjusted EBITDA (a measured discount to Froneri’s 10.5x continuation-vehicle clearing price), the base case implies an IRR of roughly 20-25% per annum over a 24-30 month horizon.

The bull case (margin reaching 22%, multiple recovering to 10x as Froneri’s perimeter expansion makes the comparable harder to discount) lifts that IRR to roughly 35-40%. The bear case (margin stalling at 17% and the multiple compressing to 8x on disposal slippage and a Q2 miss) compounds to about 8-12%. Positive, but no longer differentiated from a benchmark return.

The path is staged. The first six to twelve months carry +5-15% from forced-seller absorption plus the index-inclusion mechanic. The remainder is in the margin delivery between FY2026 and FY2028. Re-read points: the 1H 2026 print in late July, the FY2026 full-year results in February 2027, and the first index-inclusion announcement window.

I should add: I like ice cream. Summer is coming. If there was ever a time to own an ice cream stock, this is it.

Appendix: Häagen-Dazs reconciliation

The Häagen-Dazs trademark is owned globally by General Mills, Inc. (acquired via its 2001 purchase of The Pillsbury Company, which had owned the brand since 1983). General Mills licences the trademark royalty-free and exclusively to Nestlé and its authorised sublicensees for ice cream and other frozen-dessert products in the United States and Canada, under a 99-year paid-in-full licence established when Nestlé bought out General Mills’ 50% interest in Ice Cream Partners USA in December 2001 for $641m; the licence runs to 2110. In December 2019 (closed Q1 2020), Nestlé sold its US ice cream business, comprising Dreyer’s, Edy’s, Drumstick, Outshine, Skinny Cow and, importantly, its contractual Häagen-Dazs US and Canadian licence rights, to Froneri (the Nestlé / PAI Partners JV) for $4.0bn.[1][2] General Mills owns and operates Häagen-Dazs directly outside North America and Japan, manufacturing most of the rest-of-world volume at its Tilloy-lès-Mofflaines (Arras) plant in France. In Japan, Häagen-Dazs is produced and sold by Häagen-Dazs Japan, Inc., a joint venture majority-owned by Suntory (40%) and Takanashi Milk Products (10%) alongside General Mills (50%, inherited from Pillsbury, 1984). The Magnum Ice Cream Company holds no Häagen-Dazs rights in any territory; Häagen-Dazs is a direct premium-tier competitor.

References

[1] Nestlé USA, “Nestlé USA Acquires Fifty Percent Ownership Stake in Ice Cream Partners USA from General Mills,” press release, 26 December 2001 (99-year paid-in-full royalty-free Häagen-Dazs US and Canada licence; $641m consideration). Retrieved via nestle.com press-release archive.

[2] PAI Partners, “PAI Partners-backed Froneri announces acquisition of Nestlé’s US ice cream business for $4.0 billion,” press release, 11-12 December 2019. Retrieved via paipartners.com.

[3] The Magnum Ice Cream Company N.V., listing materials, 8 December 2025; €12.80 reference price; €7.8bn initial market capitalisation. Retrieved via Food Business News and Food Navigator listing coverage.

[3a] MICC NYSE and Euronext Amsterdam market data, close 22 April 2026, $13.07; 52-week range $13.06-$19.93; all-time high $19.87 recorded 11 February 2026. Retrieved via the Magnum Ice Cream Company N.V. investor relations page (themagnumicecreamcompany.com/investors) and the NYSE primary listing market-data page.

[3b] AFM Register of Net Short Positions (afm.nl), filings as of 17 April 2026, aggregate disclosed net short position equal to 19% of free float; STOXX 600 ranking via the primary European stock-loan data provider’s public weekly summary page.

[4] Stage A v2 and Stage B Demerger Precedent memo, BA Imogen, April-May 2026, internal. European broker pre-listing equity range €10.1-10.8bn (Irish Times, BNN Bloomberg, MarketScreener, ESM Magazine, 8 December 2025).

[5] The Wall Street Journal, “America’s Middle Class Embraces Discount Brands,” November 2024.

[6] The Kroger Co., Form 10-K, fiscal year ended 1 February 2025. Retrieved via SEC EDGAR.

[7] Kraft Heinz, Procter & Gamble, Unilever, Colgate-Palmolive FY2024 and Q1-Q3 FY2025 earnings transcripts.

[8] Bronnenberg, Dubé and Sanders, “Consumer Misinformation and the Brand Premium: A Private Label Blind Taste Test,” NBER Working Paper 25214, 2018.

[9] Unilever plc, Q4 2023 results transcript, 8 February 2024. Retrieved via unilever.com/files/transcript-unilever-q4-2023-results-8-february-2024.pdf.

[10] PAI Partners, continuation vehicle completion press release, 2 October 2025; €3.6bn equity transaction; ~€15bn EV; 10.5x FY2024 EBITDA.

[11] Unilever plc, FY2023 and FY2024 full-year statements; Unilever demerger circular, October 2025.

[12] The Magnum Ice Cream Company N.V., FY2025 full-year results, 12 February 2026; €7.91bn revenue; 15.9% adjusted EBITDA margin; €500m productivity programme disclosure; €250m delivered end-2025. Capital Markets Day 2025, 9 September 2025.

[13] Stage B Froneri Reconstruction, BA Imogen, 29 April 2026, internal. Underlying primary sources: Froneri Lux Topco S.à r.l. FY2023 consolidated accounts (Luxembourg RCS B241537); Nestlé FY2024 JV disclosure.

[14] Private Equity Wire, “PAI Partners secures €3.6bn for Froneri,” 3 October 2025.

[15] The Magnum Ice Cream Company N.V., prospectus, December 2025.

[16] Peter ter Kulve, CEO-designate, TMICC, Fireside Chat with Warren Ackerman (Barclays), 27 August 2025. Retrieved via Unilever IR posted transcript (unilever.com/files/tmicc-fireside-chat-video-transcript.pdf).

[18] Private Equity Insights, June-July 2025 coverage of Froneri €4.4bn shareholder payout and €3.9bn debt financing; Bloomberg, “Haagen-Dazs Maker Froneri to Raise €4 Billion Debt as PAI Seeks to Keep Stake,” 3 June 2025.

[19] General Mills, Inc., Form 10-K fiscal year ended 25 May 2025, Häagen-Dazs trademark disclosure. Retrieved via SEC EDGAR.

[20] Dairy Reporter, “Froneri’s success shows why Unilever is right to spin-off ice cream,” 27 October 2025.

[21] Haleon plc, Admission Prospectus, 18 July 2022; GSK and Pfizer disposal schedules.

[22] Kenvue Inc. and Johnson & Johnson disposal chain, IPO 4 May 2023 at $22.00; Kimberly-Clark acquisition 3 November 2025 at $21.01. 3M / Solventum spin-off 1 April 2024.

[23] Stray Narratives, “The Sorting Machine,” Issue 01, 4 March 2026.

[24] The Magnum Ice Cream Company N.V., “Q1 2026 Trading Update,” 30 April 2026. Retrieved via assets.unileversolutions.com/v1/147150923.pdf.

[25] Nestlé S.A., phased sale of remaining ice-cream operations to Froneri, 19 February 2026.

Research, not investment advice. Stray Narratives is the publishing arm of an investment-research operation; positions are disclosed alongside the analysis where they exist. Past performance does not guarantee future returns. Full disclaimer at straynarratives.substack.com.

There is a particular kind of investing frustration that is worse than being wrong. It is being right, structurally, fundamentally, analytically right, and watching the market ignore you with complete serenity while you pay the opportunity cost in silence. Uranium investors have been living in this frustration for three years.

The structural case for uranium has been broadly correct since at least 2022. Supply runs structurally short of demand. The deficit is widening. The world’s largest producer keeps cutting guidance. Reactor construction is accelerating. Utilities must contract. There is no substitution, no switching, no clever workaround. And yet: the spot price peaked at $107 per pound in February 2024 [1] and has since drifted. Equities have underperformed indices. Meanwhile, for a period in late 2024, Dogecoin’s market capitalisation exceeded the entire tradable uranium equity universe by roughly two billion dollars [2], which tells you everything you need to know about what the market was thinking about and what it was not.

The question I want to address is not whether the thesis is right. It is whether the clock has finally aligned with it, and, more usefully, where in the nuclear investment universe the thesis is actually still available to buy.

The Accounting Argument

The structural case can be stated briefly because it does not require much elaboration. It is closer to accounting than forecasting.

Global uranium demand runs at approximately 180 million pounds per year, consumed by 440 operating reactors and 65 more under construction [3]. Primary mine supply produces roughly 140 to 150 million pounds per year [3]. That gap does not close easily: new uranium mines take ten to fifteen years from discovery to production, and a decade-long bear market between 2011 and 2021 starved exploration. The pipeline of new supply is thin relative to what the demand trajectory requires.

Kazakhstan, through its state producer Kazatomprom and its joint-venture partners, is responsible for approximately 45 percent of global primary supply. Kazatomprom confirmed its 2026 production guidance cut across its Q4 2025 operations update in January and its 2025 full-year results in March [4]. This follows a multi-year pattern of underperformance driven by sulphuric acid supply chain constraints that show no signs of resolving. Their inventories sit at a ten-year low: four months of production [4]. These are not the numbers of a producer about to rescue the market.

The one element that makes uranium demand unlike almost any other commodity is that it is contractually predetermined. Utilities sign long-term supply agreements years before they need fuel. Once a reactor is built, it runs for sixty years regardless of the uranium price. Demand is not a guess. It is a schedule. The question has never been whether the market would eventually tighten. It has always been when utilities would be forced to recognise what the schedule requires.

The Market Has Already Voted, Just in the Wrong Place

Here is the observation that changes the shape of the question.

This is the kind of error the publication has described before as a wrong-map problem: the market is reading the map confidently, just not the map that matters.

The market has not, in fact, ignored the uranium thesis. It has embraced it enthusiastically. It has just placed its bets in the wrong instrument.

Nuclear equities and developers have been re-rated substantially over the past three years. NexGen Energy, the company sitting on what may be the largest and highest-grade undeveloped uranium deposit in the world, trades meaningfully above its base-case net asset value at current uranium prices. The stock requires the full bull scenario, uranium at $150 per pound or more, simply to avoid losing money at today’s price. Cameco, the senior miner and the closest thing the sector has to a blue chip, is priced at base-case fair value, which sounds reasonable until you realise that base case offers limited upside from current levels and the bear case still takes forty-seven to sixty-six percent off.

SMR stocks, the small modular reactor developers, are in a category of their own. These are companies whose products will not reach commercial scale until the 2030s at the earliest. The stocks trade on narrative, and the narrative is at least a decade ahead of the physical reality.

The result is a strange inversion. Across the nuclear investment universe, the market has already answered the question “will the thesis play out?” with a fairly confident yes. It just expressed that answer through the equities rather than through the commodity.

The uranium price itself has not moved to reflect this. Physical uranium, held through vehicles like the Sprott Physical Uranium Trust, trades near net asset value [8]. The bull case upside, if the contracting cycle fires and prices move toward $120 to $150 per pound, is roughly seventy-seven to one hundred and twenty-four percent from current levels. The bear case, if the thesis is wrong or delayed, is a loss of twenty-two to twenty-five percent.

That is a meaningfully different risk profile than a developer priced for resolution. The equity market has been enthusiastically pricing in a thesis that the commodity market has not yet confirmed. When those two things diverge, one of them is wrong. They are rarely both right.

What Would Actually Signal the Clock Has Aligned

The structural case being correct and the clock being correct are separable skills. Uranium has a long track record of making fools of people who confused the two. So what would actually signal the thesis is firing, rather than merely strengthening?

Three conditions matter:

1. Utility contracting at or above replacement rate. The replacement rate is approximately 180 million pounds per year. In 2024, global contracting ran at roughly 90 million pounds: half the replacement rate [5]. The majority of that was Chinese and Ukrainian utilities, not the Western utilities whose behaviour drives the price. Q4 2025 showed the first real contracting inflection. That is a necessary beginning, not a sufficient signal.

2. Kazatomprom’s inventory buffer exhausted. Guidance cuts are necessary but not sufficient. They tell you supply is running slow. What matters is when the buffer is gone. At four months of production and declining, that point is getting close. When it arrives, the physical tightness becomes unavoidable rather than theoretical.

3. Long-term contract price moving decisively above current levels. The long-term price sits at approximately $90 per pound [7]: already the highest level in history at the beginning of a contracting cycle, which is itself an extraordinary fact. But utilities signing contracts in volume at $90 and above is qualitatively different from the price sitting at $90 because nobody is signing anything.

Where Each Signal Stands

The honest assessment against that framework:

• Kazatomprom’s guidance cut is confirmed as structural, not temporary. The first condition is advancing materially. NexGen received its construction licence from the Canadian Nuclear Safety Commission on 5 March 2026 [6]. The largest undeveloped uranium deposit in the world is now a project under construction, removing what had been the thesis’s most significant binary risk. These are real developments.

• And yet the contracting inflection of Q4 2025 remains, by definition, an inflection from a very low base. Utilities remain substantially under-contracted relative to their forward needs. The long-term price has not moved decisively. The Sprott trust is roughly at net asset value, not at a sustained premium, which would signal Sprott’s buying mechanism re-engaging in earnest.

The trigger has not fired but the conditions for it are the most advanced they have been.

The Genuine Uncertainty

The uranium market has a specific and well-documented failure mode: the “imminent inflection” call. The number of analysts who correctly identified the structural deficit and incorrectly identified the timing is large. Utilities have demonstrated a remarkable capacity to remain under-contracted for longer than any reasonable person expected. They have kept finding levers (secondary market purchases, existing inventory, creative procurement) that have deferred the moment of reckoning further than the supply-side arithmetic suggested was possible.

There is also a consensus problem. Every major research house is bullish uranium. Two significant catalysts, the NexGen licence and the Kazatomprom confirmation, are now public information, available to every market participant. When a trade becomes consensus, the variant perception that generates returns has largely been captured. The setup is not the same as it was in 2020 when the thesis was lonely.

This is not an argument that the thesis is wrong. It is an argument that “the thesis is right” and “now is the optimal entry point” are different claims, and the second is harder to make than the first.

Being right and being paid are separable. When a structural case is lonely, the early mover earns both the fundamental return and the re-rating as consensus arrives. When the case is consensus, the re-rating has already happened, or is happening, and only the fundamental return remains. The structural reality stays intact. The edge compresses anyway. The sharpest version of this for uranium is right thesis with wrong timing: you hold the position for eighteen months of cost of carry and foregone returns elsewhere, then exit flat when the thesis finally fires for the next cohort.

What the Disciplined Investor Does

The consequence is that instrument selection matters more than the timing call. Sprott at NAV preserves the asymmetry at the commodity level even if the equities have already de-rated the edge.

The answer to “right thesis, uncertain clock” is not to wait for certainty. Certainty arrives after the move. By the time the contracting cycle is visibly firing, with utilities scrambling, spot price spiking, and headlines appearing, the easy money is already made.

But neither is the answer to bet heavily on precise timing. The track record of “imminent inflection” calls in uranium specifically argues against that.

The distinction that matters is between conviction sizing and timing sizing. If the structural case commands genuine conviction, and the accounting above suggests it should, the position should reflect that conviction. But it should be sized for the possibility that the clock is still twelve to twenty-four months from aligning, not sized as though the move is imminent. There is a cost to being early, and that cost is real and should be acknowledged honestly rather than wished away.

The epistemically cleanest vehicle remains physical uranium via Sprott. It provides direct exposure to the commodity without operational risk, without relying on any individual miner to execute a multi-year construction project in a difficult jurisdiction on a schedule and within a budget. Miners offer leverage to the upside. The leverage cuts both ways, and most of the accessible miners have already priced in a version of the bull case that the commodity itself has not yet confirmed.

The reason Sprott at NAV is the right vehicle, and not merely an available one, is mechanical. When the units trade at a premium to net asset value, the trust issues new units at the market, uses the proceeds to buy physical U3O8, and stores it in licensed facilities. Spot demand gets created mechanically, and the trust’s accumulation itself tightens the market. The NAV premium is therefore the leading indicator: when it widens, Sprott is buying, and the feedback loop is firing.

When the units trade at or near NAV, the at-the-market issuance stops. No new units, no new uranium purchases, no mechanical spot bid. The mechanism has not broken. It is dormant, waiting for the premium to return. Buying near NAV is buying ahead of the bid, not alongside it.

Sometimes the most sophisticated thing you can do in a crowded trade is to go back to the underlying. The equity market has been buying the thesis. The commodity has not moved yet. One of those is closer to the price that will matter.

References

[1] UxC Uranium Market Outlook, historical spot price series, February 2024. [Spot peak of $107/lb.]

[2] CoinGecko and Sprott Uranium Miners UCITS ETF constituent market-cap aggregation, Q4 2024 snapshot. [Dogecoin market cap vs tradable uranium equity universe.]

[3] World Nuclear Association, “World Nuclear Power Reactors & Uranium Requirements,” updated 2026. [Operating reactor count, reactors under construction, annual uranium demand, and primary mine supply figures.]

[4] Kazatomprom Investor Relations, “4Q 2025 Operations and Trading Update” (January 2026) and “2025 Full-Year Results” (March 2026), kazatomprom.kz. [2026 production guidance cut; inventory at four months.]

[5] UxC Contracting Volume, calendar year 2024. [Global utility contracting at approximately 90 million pounds.]

[6] Canadian Nuclear Safety Commission, Rook I construction licence decision, 5 March 2026, nuclearsafety.gc.ca. [NexGen construction licence issuance.]

[7] UxC and TradeTech long-term price indicators, April 2026. [Long-term price at approximately $90/lb.]

[8] Sprott Asset Management, Physical Uranium Trust (U.UN/U-UN.TO), NAV pulled 2026-04-20, sprott.com. [Premium to NAV of +0.56%, i.e. trust trading at net asset value.]

There is a question worth asking before the quarterly results, before the capex disclosures, before the infrastructure pipeline numbers. It is not whether AI works. It does. It is not whether AI will make businesses more efficient. It will. The question is whether making everyone more efficient at the same time, using the same tools at roughly the same cost, makes anyone more profitable.

Electricity is the answer. When rural electrification arrived in the early decades of the last century, it transformed productivity across every sector it touched. Farms, factories, shops, offices, all became measurably more efficient. The companies that owned the wires and the generators did not become the dominant businesses of the century. The value flowed to the users of electricity, not its suppliers, because access became universal and the efficiency gains competed away through lower prices and higher output expectations. The wire was not the business. What the wire enabled was not the business either. The business was what you built once reliable power was a given and your competitors had it too.

AI is the new wire. Access to it is becoming a baseline operating condition. The productivity gains it delivers will be real and they will be available, at broadly comparable cost, to every firm in every sector simultaneously. That is not a reason to be pessimistic about AI. It is a reason to be precise about where the value actually goes.

It does not, in the main, go to the companies building the infrastructure.

A brief detour through history

The historian Carlota Perez spent considerable time studying what happens when genuinely transformative technologies arrive. Canals, railways, rural electrification, fibre optics. Her finding, somewhat inconvenient for anyone hoping for a clean story, is that bubbles and golden ages tend to arrive in sequence rather than as alternatives. The bubble is not the opposite of the revolution. It is, rather annoyingly, the precondition for it.

The railroad bubble of the 1870s wiped out a remarkable number of companies and banks[1]. It also settled the American West and accounted for sixty-two percent of total US market capitalisation by 1900. Both things were true simultaneously. Arguing that AI is a bubble is not the same as arguing AI does not matter. It is simply arguing that the path from here to the golden age runs through some territory the current valuation of the sector has not fully priced.

What the market has not fully priced is not the technology. It is the transformation of the companies building it.

What the hyperscalers are becoming

Imagine a toll road. The owner built it, owns it, and collects tolls from everyone who uses it. The economics are straightforward and rather pleasant. The marginal cost of one more car is close to zero. As traffic grows, margins expand. This is what the large technology platforms have been for the past twenty years. Asset-light, high-margin, and protected by the simple fact that everyone needed to use the road.

Now imagine the government tells the toll road owner that it must also build the next hundred miles of highway, fund the electricity grid that powers the tunnel lighting, sign fifteen-year fixed-price contracts for construction materials, and accept a regulated return on the new sections. The original toll revenue is still coming in. But the company is now simultaneously a construction firm, a utility, and a regulated infrastructure operator. Its balance sheet looks different. Its risk profile looks different. And a sensible investor, asked to value it, would reach for a rather different multiple.

This is what is happening to the hyperscalers. The five largest technology companies by capital expenditure are forecast to spend the equivalent of more than two percent of US GDP this year. To put that in context: it exceeds the peak railroad buildout of the nineteenth century, the shale energy boom, and the fibre optic spending of the dotcom era. Historic is now the correct word.

The reason only the largest incumbents can build this infrastructure is not mysterious. The bottleneck has shifted from acquiring chips to securing land, substations, transformer capacity, and the utility relationships that allow a data centre to move from contracted power, which is merely a promise, to connected power, which means actual electrons flowing into actual buildings. When challenger capacity is cancelled, as nearly half of planned US data centre builds this year are expected to be delayed or cancelled, it is the smaller operators who fall away first. The incumbents retain pricing power precisely because physical execution is so hard to replicate.

So only they can build it. And building it is changing what they are.

The technology platforms have historically derived their monopoly power from three sources: economies of scale, network effects, and proprietary technology. AI is eroding all three simultaneously. Economies of scale are undermined because AI lowers the fixed cost of software development for everyone, including the hyperscalers’ own customers. The Catholic Church initially welcomed the printing press as a useful tool for spreading its message. The printing press had other ideas. Network effects are threatened by AI agents that sit between the user and the platform, potentially reducing what is now a coveted destination to a repository of content. Proprietary technology is undermined by the degree to which the underlying AI research has remained open source.

The result is a double movement. Software margins compressing from below, as the intelligence commodity deflates and barriers to entry fall. Infrastructure costs accumulating from above, as the physical build-out obligation grows. The balance sheet in the middle absorbing pressure from both directions at once.

The market has begun to notice. Two years ago, every dollar of AI capital expenditure announced was rewarded with two dollars of additional market capitalisation. A year ago, the ratio had fallen to one for one. It has since reversed. Forward multiples across the hyperscaler class have been compressing, the market beginning to price these businesses not as asset-light software platforms but as capital-intensive infrastructure operators. It is not saying they will fail. It is adjusting the multiple to reflect what they are becoming.

Not all hyperscalers sit in the same position. The class looks uniform from a distance but three structural variables separate the well-positioned from the vulnerable: custom silicon, cloud revenue, and edge inference capability. The companies that combine all three occupy a fundamentally different economic position from those that have one or none.

Custom silicon is the most widely discussed. Google has developed custom tensor processing units at a scale that makes it substantially independent of external chip supply. Amazon’s custom chip programme, covering its Graviton processors, Trainium AI chips, and Nitro networking infrastructure, has reached a comparable position. Both companies are avoiding what might be called the Nvidia margin tax: the substantial gross margins that a company dependent on external silicon pays to its chip supplier before it earns anything from its own customers. The custom silicon route is not without risk, but it restructures the cost base in a way that external dependence cannot.

Cloud revenue is the variable the market has not fully separated from the capex story. Amazon, Microsoft, and Google operate cloud platforms that generate external revenue from the same infrastructure they are building. Every dollar of data centre capex serves both their internal AI workloads and their paying cloud customers. The infrastructure is a cost centre and a revenue centre simultaneously. Meta and Apple do not have cloud businesses. Their infrastructure spending is pure cost against advertising revenue or hardware margins, with no external customer base to amortise the investment. The distinction matters because it determines whether the capex obligation is self-funding or parasitic on existing margins. A company building infrastructure that other companies pay to use is in a structurally different position from one building infrastructure that only serves its own products.

Edge inference is the variable that almost no one is pricing correctly, and it cuts both ways. If AI processing migrates from centralised data centres to devices, running on the neural processing units now embedded in every flagship smartphone and laptop, it reduces long-term demand for centralised compute. Apple is better positioned for this shift than any other company in the class. Its on-device inference capability, built on years of custom silicon for iPhones and Macs, means it can deliver AI features without routing every query through a data centre. Google and Qualcomm are investing heavily in the same direction. The implication for the infrastructure thesis is not that edge AI kills it, but that it shortens the constraint window. Training and heavy agentic workloads remain centralised for the foreseeable future, but the assumption that inference demand scales linearly into data centres deserves more scrutiny than it is receiving. Edge erosion is a risk to the duration of the physical constraint, not to its existence.

The combination is what matters. Google and Amazon have custom silicon, cloud revenue, and growing edge capability. They sit in the strongest structural position. Microsoft has cloud revenue and corporate integration but depends on external silicon, a vulnerability it is addressing but has not yet resolved. Meta has custom silicon ambitions but no cloud business and no meaningful edge presence, making its infrastructure spend the most concentrated bet in the class. Apple has the strongest edge AI position and the most advanced device silicon but no cloud business, which means it captures the inference shift without participating in the infrastructure buildout at all.

The position of Microsoft deserves specific attention because it is the most complicated case. It is the most deeply embedded of any technology company in corporate operational infrastructure. Office 365, Teams, and Azure are not easily substituted, and that integration gives it a genuine platform durability that the infrastructure cost argument does not simply override. Some of the multiple compression across the sector reflects the broader re-rating of software businesses rather than a specific infrastructure judgement. A company with Microsoft’s level of corporate penetration may prove more resilient in the AI transition than a clean reading of the electricity parallel would suggest.

Four points, stated plainly. First, the hyperscalers are being repriced from software to infrastructure economics. This is a direction, not a destination, and the transition is incomplete. Second, within the class, the companies best positioned to absorb the double squeeze are those that combine custom silicon, cloud revenue, and edge inference capability. Third, cloud revenue is the underappreciated buffer: it turns infrastructure capex from a pure cost into a revenue-generating asset, and the market has not yet fully differentiated between hyperscalers that have this buffer and those that do not. Fourth, edge AI compresses the duration of the centralised infrastructure constraint, which matters for how long the physical bottleneck thesis remains operative, and for which companies emerge strongest on the other side of the buildout.

The honest counterargument

There is a scenario in which compute scarcity itself becomes a source of pricing power. If the physical bottleneck proves as durable as the evidence suggests, and if only the incumbents have the execution depth, then the survivors may find their earnings revised upward even as their multiples compress toward utility economics. Inference prices across the major platforms have been edging upward rather than downward in recent months. Token credits are being rationed. API pricing is being revised. The consensus has not yet noticed.

The mechanism is traceable. Providers absorbed cost gaps through 2025 to win market share. High-end compute costs then surged. Agentic workflows arrived, consuming tokens at multiples of traditional chat interfaces, and the cost buffers that cloud providers had been willing to carry ran out. The compute repricing is not a projection. It is already visible in how the major platforms are pricing their inference products.

The difficulty with dismissing this counterargument is that it does not operate on the same time horizon as the electricity parallel, and conflating the two produces a false resolution. The infrastructure repricing thesis, compute becomes structurally more expensive, the physical asset layer captures a larger share of AI profits, operates on a three-to-five year horizon. The electricity parallel, intelligence eventually commoditises, value flows to users, operates on a fifteen-to-twenty year horizon. Both can be true simultaneously. This thesis does not require the short-run infrastructure repricing to be wrong. It requires only that the transition period eventually ends, which it will. The honest acknowledgement is that for investors operating on a three-to-five year horizon, the infrastructure pricing power thesis may be the more operationally relevant frame, even for those who accept the long-run electricity parallel.

The most plausible version of the longer story is that Chinese efficiency work ultimately shortens the window. Chinese laboratories have demonstrated that inference costs can be reduced by an order of magnitude through algorithmic efficiency rather than raw compute. Where Western firms have responded to capability demands by building larger data centres, Chinese competitors have responded by making the models themselves more efficient, achieving comparable results at a fraction of the energy and capital cost. The Chinese efficiency argument is real and should not be dismissed. What it does not resolve is the near-term physical constraint: the agentic workloads driving current infrastructure demand are not yet efficiently substitutable, and the disclosed US data centre pipeline of over two hundred gigawatts[3], against active development capacity that can deliver only a fraction of that in any given year, ensures the physical squeeze persists for years rather than quarters. The efficiency gains compress the duration of the constraint. They do not eliminate it.

This returns us to the electricity parallel. The early electricity companies enjoyed pricing power during the period when generation capacity was scarce. That pricing power eroded as access became universal. The question is not whether AI intelligence will eventually become a cheap commodity. It will. The question is how long the transition period lasts, and who captures the rents during it.

What the bull case gets wrong

There is a number that has been doing considerable damage to clear thinking about AI revenue. It is the growth rate of coding agent usage since late 2025. When autonomous coding tools arrived, their adoption was rapid and their token consumption was extraordinary. Analysts, confronted with a line on a chart going in a direction lines rarely go, did what analysts tend to do. They extrapolated it.

The extrapolation rests on a confusion between two very different kinds of work. The distinction between them matters.

The first is the nature of the work itself. Coding is deterministic and expansive. The code either runs or it does not. Writing it generates more code, test routines, build routines, debugging cycles, each consuming additional tokens. Agentic systems, autonomous sequences of tasks run continuously without human intervention between steps, maintaining state across hours or days of operation, are similarly expansive, consuming energy at a multiple of a standard query. The compression argument that follows applies to the remainder of knowledge work, not to these categories. Most white-collar work is neither deterministic nor expansive. It is compressive and probabilistic. You have a large report. You want the key points. The AI takes many tokens in and produces few out. There is no expanding chain of output. The task is done and the token consumption ends.

The second problem is more fundamental. A significant portion of knowledge work does not create genuine economic value. It exists because the world is opaque, fragmented, and difficult to cross without a guide. Intermediaries who exist because buyers and sellers cannot find each other efficiently. Compliance layers that exist because processes are too complex to follow without interpretation. Reconciliation work that exists because systems do not communicate. Advisory functions that exist because complexity makes self-service impossible. When AI compresses that complexity, this work does not get automated. It gets eliminated. The job ceases to exist not because a machine does it more efficiently, but because the problem it solved ceases to exist. The total addressable market for AI-assisted knowledge work is therefore smaller than the projections assume, because part of what those projections counted as economic activity was friction dressed as value. AI is not replacing it. It is revealing it.

This is where the electricity parallel bites again. Electrification made every business more productive. It did not cause every business to consume electricity in proportion to its productivity gains. The assumption that AI token consumption scales linearly with white-collar adoption is the equivalent of assuming that every electrified farm would consume power in proportion to the acreage it tilled. They did not. Neither will offices.

There is a further confusion embedded in the productivity data. Measured US productivity has been rising at rates not seen in a decade. The immediate reaction has been to attribute this to AI. It is arithmetic. When investment rises sharply and hours worked remain constant, measured productivity rises as a matter of definition. The capital expenditure on AI data centres is lifting the investment component of GDP. That is a statistical artefact, not a productive miracle. The market has been reading the artefact as confirmation of the demand thesis. It is not.

The physical wall

The US electrical grid handles roughly fourteen percent of the nation’s total energy flow[2]. The remaining eighty-six percent moves through liquid fuels, natural gas pipelines, and food. A human worker draws energy from all of these systems. An AI data centre draws exclusively from the fourteen percent. When you replace human cognitive work with machine cognitive work, you are not substituting one energy source for another. You are attempting to route a vastly larger share of economic activity through a grid that was not built to carry it.

The disclosed US data centre pipeline stands at over two hundred gigawatts[3], up more than one hundred and fifty percent year on year. Only a third is under active development. Almost half of planned builds this year are expected to be delayed or cancelled. The bottleneck is not capital and not demand. It is the physical execution layer.

Transformers are the most instructive case. Before 2020, a large power transformer arrived roughly two years after ordering. The combined pressure of AI construction, grid expansion, and industrial reshoring has pushed delivery times to five years in some cases, with prices up fifty percent. American manufacturing cannot meet domestic demand. The solution has been to import from China. US utilities imported more than eight thousand high-power transformers from China in the first ten months of last year, against fewer than fifteen hundred in the whole of 2022[4]. The country engaged in strategic competition with China over AI supremacy is dependent on Chinese supply chains to build the infrastructure that competition requires.

New large power transformer manufacturing capacity requires five to seven years from investment decision to first commercial shipment, gated by sequential constraints in specialised steel production, winding capacity, and testing infrastructure. American manufacturing cannot respond to demand at speed not for want of capital or will, but because the production bottlenecks cannot be resolved within the buildout window. This is not a pipeline risk or a forecast. It is the current state.

There is a further constraint that compounds the physical one. Regulatory frameworks designed to accelerate grid queue access, which allow interruptible loads to move to the front of the interconnection queue, are structurally unavailable to the workload class that most needs them. Agentic AI systems run continuously, maintaining state across multi-step tasks that cannot be paused without losing their entire context. The regulatory fast lane was designed for a load type that can be interrupted. The load type driving the most urgent infrastructure demand cannot.

The cost of natural gas, the primary fuel for new data centre power generation, has remained contained. The cost of electricity has not. The gap between the two is the cost of a transmission layer not designed for current loads. There is plenty of cheap gas. Turning it into cheap power at the point of consumption requires infrastructure that does not yet exist in sufficient quantity.

A displaced worker does not immediately stop drawing energy from the grid. They still heat their home, still charge their car, still consume electricity as before. Meanwhile the data centre that replaced their work adds a new industrial load. And electrification policies are simultaneously redirecting existing energy consumption onto the same constrained infrastructure. Three forces pressing on the same copper wires. The grid’s share of total energy must rise. Raising it requires rebuilding the transmission layer. The transmission layer is the binding constraint.

The compound problem

Capital is committed and spent long before a data centre generates revenue. Equipment is ordered, construction begins, depreciation clocks start, and debt is serviced, all on a timeline that precedes physical completion by years. When delivery slips, as the pipeline data suggests it is for nearly half of planned builds, the gap between cash out and cash in widens. The asset base meanwhile depreciates faster than the accounting schedules acknowledge. Chips used for frontier model training degrade in eighteen months, not the five to six years on official schedules[1]. The financial pressure accumulates in the gap between what the accounts show and what the physics dictates.

The electric vehicle industry is instructive. A genuinely important technology, rapid adoption, massive capital investment, and consistent failure to generate profits across most of the sector despite years of trying. The infrastructure was real. The demand was real. The economics were, and in many cases remain, elusive. AI need not follow exactly the same path. But the parallel is worth sitting with, particularly given that the AI infrastructure buildout is an order of magnitude larger than anything the electric vehicle industry attempted.

Where the value actually goes

The electricity parallel points toward a specific conclusion. When a transformative technology becomes universally accessible, the value does not accumulate in the wire. It accumulates in three places.

The first is the scarce physical inputs the wire requires. Copper is the transmission metal for the grid that AI demands. Uranium provides the baseload power that data centres need around the clock. Natural gas is the fastest deployable generation technology for the transition period, but the investment case is not in the commodity. It is in the pipeline infrastructure that moves it.

Gas pipeline operators with long-term contracted throughput agreements are a distinct sub-asset class from commodity gas exposure, and the distinction matters. The companies that matter here are not those selling gas at spot prices but those that have signed decade-long take-or-pay agreements with hyperscaler counterparties, committing to deliver a fixed volume regardless of what the market does, in exchange for a committed revenue stream that sits on the balance sheet like a bond. Regional gas price spreads in the corridors where data centre construction is most concentrated have expanded materially, the market’s way of saying the pipe is filling up before most of the demand has arrived.

Every data centre, regardless of who builds or operates it, must connect to every other data centre through optical fibre. Every optical transceiver, the hardware that converts electrical signals to light for long-distance transmission, requires indium phosphide as a substrate for its laser sources. Demand for these components has grown substantially faster than production capacity has expanded. China controls the majority of refined indium and introduced export controls in early 2025, creating a geopolitical supply constraint that cannot be engineered around quickly. This is a physics-limited bottleneck on the infrastructure that allows AI systems to communicate at scale, and it is almost entirely absent from public analysis of the buildout. Dark fibre routes in secondary hyperscaler markets, the owned rights-of-way through which this traffic flows continuously, represent a related asset class: independent of which model or platform wins the intelligence competition, the light must travel somewhere.

The pattern repeats at multiple points in the supply chain. Physics-limited upstream materials meeting exponentially growing demand produce the same structural dynamic at each layer. The constraint does not care which company’s logo is on the data centre.

The second is the broad economy rather than the technology sector itself. When AI access is universal, the businesses that benefit most are those in fragmented, relationship-dependent, or operationally complex sectors where embedding AI deeply restructures the cost base before competitors do the same. Industrials using AI to extract efficiency from physical operations. Financial services firms with proprietary data that becomes more valuable as generic intelligence becomes cheap. Healthcare, where data ownership creates durable advantage even when the underlying model is commoditised. The equal-weight index is the blunt instrument version of this thesis. It began outperforming the market-capitalisation-weighted index in 2025 and the direction reflects something real: earnings upgrades are migrating from the technology sector toward the rest of the economy. This is not a rotation driven by lower interest rates. It is a rotation driven by earnings.

The third is assets that hold their value regardless of which technology wins. Gold deserves specific attention here, and not for the conventional reason. Michael Green, a market strategist who has spent considerable time auditing the energy economics of the AI transition, identified something striking earlier this year. The historical relationship between gold and real interest rates has broken down. For decades when real rates rose, gold fell, because the opportunity cost of holding a non-yielding metal increased. Since 2022 that relationship has inverted. Gold and real rates are rising together. Green’s interpretation is that this is not an inflation signal in the traditional sense. It is a political signal. The transition from a human-led to a machine-led economy requires coordinated infrastructure investment at a scale that democratic political systems struggle to sustain. It requires telling voters that near-term disruption is the price of long-term gain, which is not a message that wins elections. Gold, in his reading, is rising because investors are beginning to price the possibility that the institutions responsible for managing this transition lack the political will to do it in an orderly way. It is a hedge against governance failure rather than against price level changes.

What to be cautious about is the shorter list. Software-as-a-service businesses whose moats rest on switching costs rather than proprietary data face a structural challenge that private market valuations have not yet fully absorbed. The large technology platforms are not uninvestable, but the direction of multiple compression across the class is a signal about what the market is beginning to understand about their cost structures. The specialist GPU cloud operators are the most exposed to a correction in the scarcity premium, because their revenue depends entirely on the incumbents continuing to need external capacity at current pricing, and Chinese efficiency work is shortening the window in which that condition holds.

The wire is not the business. It never was. The businesses built on cheap and universal electricity are where the twentieth century’s wealth was made. The equivalent businesses for the AI era are, in many cases, not yet visible in their mature form. They will not be found in the current composition of the major indices. They will be found in the sectors and geographies that a decade of capital flowing toward software has most thoroughly ignored.

The infrastructure investment cycle, if it corrects, will correct asymmetrically. The companies with the strongest balance sheets will absorb the adjustment. The sovereign funds have patience. The private credit vehicles have their own difficulties, but those are problems for sophisticated counterparties who accepted the terms knowingly.

The adjustment does not fall primarily on capital. It falls on the workforce that was promised reconstitution on the other side of the displacement. That workforce, its size, its composition, and the political consequences of its disappointment, is where this series is going next.

The infrastructure investment cycle, if it corrects, will correct asymmetrically. The companies with the strongest balance sheets will absorb the adjustment. The sovereign funds have patience. The private credit vehicles have their own difficulties, but those are problems for sophisticated counterparties who accepted the terms knowingly.

The adjustment does not fall primarily on capital. It falls on the workforce that was promised reconstitution on the other side of the displacement. That workforce, its size, its composition, and the political consequences of its disappointment, is where this series is going next.

Three names on the right side of the wire

The argument above identifies where value goes. That leaves the question of how to express it in a portfolio. The following three names are not the most famous companies in the AI infrastructure theme. They are the ones where the thesis is not yet fully in the price.

A systematic screen of 381 companies across US, European, and Asian exchanges scored each name on valuation, capital deployment, earnings quality, and fundamental stability, then applied a qualitative overlay for thesis-specific factors: the type of constraint each company controls, the duration and contractual structure of its revenue, its position in the supply chain, and its direct exposure to hyperscaler customers. The result is a ranked universe. The three names below sit at the intersection of strong thesis fit and reasonable valuation, a combination that is rarer than it sounds in a theme the market has already noticed.

Subscribe now

Cheniere Energy (LNG): The largest liquefied natural gas exporter in the United States, operating approximately forty-five percent of the country’s LNG export capacity across its Sabine Pass and Corpus Christi terminals. Cheniere sits at the intersection of two constraints that cannot be resolved within the buildout window: gas infrastructure and LNG terminal capacity, both of which take five to seven years to construct. Roughly ninety percent of its capacity is contracted under long-term take-or-pay Sale and Purchase Agreements with counterparties including Shell and TotalEnergies. At 7.4x EV/EBITDA with a 75.8 percent operating margin, it is the cheapest and highest-quality infrastructure name in the screen. The business model is a toll road: fixed-capacity processing with contracted throughput, generating predictable cash flows regardless of commodity price direction. As domestic gas demand rises to feed data centre power generation, the same tightening in the US gas market that benefits pipeline operators flows directly through Cheniere’s Henry Hub-indexed contracts.

Mitsubishi Heavy Industries (MHVYF): The third member of the global gas turbine oligopoly, alongside GE Vernova and Siemens Energy. Only three companies in the world can manufacture the large-frame gas turbines that new data centre power plants require. All three have order books sold out through 2029 or 2030. The physics of turbine manufacturing, specialised alloys, precision casting, and multi-year qualification cycles, mean that no new entrant can resolve this constraint within the buildout window. GE Vernova, the most visible expression of this thesis, trades at 114.6x EV/EBITDA. Mitsubishi Heavy trades at 26.7x. The turbines are the same. The multiple is not. Mitsubishi Heavy is Tokyo-listed with a US ADR, which means thinner liquidity and yen currency exposure, both real costs. But a 77 percent discount to the pure-play peer for an identical physics-limited constraint is where systematic screening earns its keep, surfacing a name that qualitative research alone would not have found.

Siemens Energy (ENR.XETRA): The European member of the same turbine oligopoly, with the cheapest forward price-to-earnings ratio of the three at 41.3x. Siemens Energy has been weighed down by losses in its Siemens Gamesa wind division, an overhang that has kept the gas turbine story discounted relative to GE Vernova. Earnings growth of 240 percent reflects the turnaround inflecting. If Gamesa stabilises as guided, the gas turbine earnings become the dominant driver and the stock re-rates. The risk is known. So is the discount it creates.

The names that were excluded matter as much as the names that were included. GE Vernova is the purest expression of the turbine thesis but at 114.6x EV/EBITDA the market has priced the next five years of earnings growth and then some. Williams Companies controls the irreplaceable Transco pipeline corridor and has a twenty-year take-or-pay deal with Meta, but at 33x price-to-earnings it trades at more than double Energy Transfer’s multiple for a comparable asset type. Coherent, the named NVIDIA supplier for co-packaged optics, sits at 56x EV/EBITDA with a binary outcome: if the CPO transition accelerates, the price is justified, and if it stalls, there is a long way down. In each case, the thesis is correct. The valuation is the problem.

The honest framing is this: the AI infrastructure thesis is not a variant perception. It is consensus. Every sell-side firm has published a version of it. The edge, if there is one, is not in identifying the theme but in identifying which names express it at valuations where the risk and reward still favour the buyer. A turbine is a turbine whether it carries a GE logo or a Mitsubishi logo. The physics are identical. The multiples are not. That gap is the opportunity, and it will not persist indefinitely.

As always, a “like” or the “sharing” of this article means the world to me !

References

1] Paul Kedrosky and Derek Thompson, “Yes, AI Is a Bubble. There Is No Question,” Plain English podcast, March 18, 2026. The rational bubble argument, the railroad parallel, the compressive versus expansive token distinction, and the chip depreciation analysis are drawn from this conversation.

[2] Michael W. Green, “The Thermodynamic Margin Call,” yesigiveafig.com, January 18, 2026. The fourteen percent grid bottleneck, the Triple Pincer, and the gold and real rates analysis are drawn from this essay.

[3] Sightline Climate, cited in Bloomberg News, 2026. The disclosed data centre pipeline figures and the active development and cancellation data are drawn from this source.

[4] Wood Mackenzie, cited in Bloomberg News, 2026. The transformer import volume data comparing 2022 and 2025 figures are drawn from this source.

The standard argument runs as follows. AI performs tasks that humans previously performed. The human is no longer required to perform those tasks. The human is therefore displaced. This is mechanically correct and, in many sectors, already observable. It is not, however, the whole picture.

There is a second category of effect that is analytically distinct, and in some respects more consequential. AI does not automate the task. It makes the task unnecessary. The task disappears not because a machine now does it, but because the problem it was solving ceases to exist. There is no displacement because there is nothing left to displace.

I have been calling this friction elimination, which is a somewhat dry formulation for something that has fairly dramatic economic implications. Bear with me.

Consider contract law, or more precisely, the interpretation work that contracts generate. Contracts are not written to be clear. They are written, in many cases, to create productive ambiguity, the kind that requires expert navigation every time a material question arises. Legal complexity of this variety is not accidental. It is structural. It sustains an entire layer of professional activity.

AI that simplifies the drafting of contracts, and that makes their meaning unambiguous at the point of execution, does not automate the contract reviewer. It removes the conditions under which contract review becomes necessary. The reviewer is not replaced. The reviewer’s function is dissolved.

A meaningful share of what financial intermediaries do, and I say this as a thirty-year veteran of a sector that has always charged handsomely for guiding clients through a landscape it had some interest in keeping unmapped, exists not because they create value in the transaction itself but because without them the buyer and the seller cannot find each other, cannot assess creditworthiness, or cannot navigate a regulatory architecture that was not designed with clarity as its primary objective. AI that connects counterparties directly, that renders creditworthiness legible in real time, and that translates regulatory requirements into plain operational instruction, does not automate the intermediary. It publishes the map.

Consulting produces a version of the same effect. A substantial share of consulting revenue, perhaps the majority in some practice areas, is not the cost of strategic advice. It is the cost of organisational opacity. Two institutions that cannot speak each other’s language hire a translator. AI that makes data legible across institutional boundaries, that renders one organisation’s reporting comprehensible to another without human mediation, does not automate the consultant. It removes the opacity that made the consultant necessary.

A research paper published earlier this year mapped the distribution of AI agent development across more than a thousand United States occupations, examining which roles are being actively targeted for AI application and which are not. [1] The concentration is striking, and it has direct bearing on where friction elimination is already arriving.

Three numbers in that table deserve attention before moving on.

The first is 13.1 percent. That is the share of the US civilian workforce sitting in roles with high or moderate friction elimination exposure, approximately 21 million workers, once office and administrative support is included alongside the professional categories most commentators focus on. Administrative support alone accounts for 16 million of those workers, which is the figure the automation debate most consistently underestimates. Scheduling, coordination, document management, internal reporting: these are not glamorous categories, but they are large ones, and they are already compressing.

The second number is 77 percent. That is the share of the workforce that sits outside the friction elimination frame entirely. Food service, retail, manufacturing, transport, healthcare support, government, education: these are primarily physical, site-based, or institutionally regulated roles, and AI’s primary mode of attack on them is not friction elimination but demand compression and traditional automation. Those are distinct risks with distinct timelines and distinct policy implications, addressed elsewhere in this series. [2] The point here is simply that friction elimination is not a universal story. It is a concentrated one.

The third number is the one the table does not show directly, but which follows from the first two. The 13.1 percent of workers facing friction elimination are not distributed evenly across the income spectrum. They are concentrated in the upper-middle income range, in the graduate-entry professional and administrative cohort that fills the office towers of every major city and whose spending sustains the restaurants, retailers, and service businesses around them. When that cohort contracts, the contraction does not stay contained. It radiates outward into the 77 percent.

The insulated categories, for the record, span a wider income range than is commonly assumed. At one end sit care workers and skilled tradespeople, whose protection comes from physical presence and manual dexterity. At the other end sit surgeons, senior lawyers, and relationship bankers, whose protection comes from licensed personal accountability and years of accumulated trust that is personal rather than institutional. The reader who recognises their own profession in that latter group may feel a degree of reassurance at this point.

There is an economic consequence here that is distinct from the wage suppression argument I made in the last issue, and I want to be precise about the distinction.

That piece described what happens to the price of genuine cognitive work when AI compresses the supply of cognitive capacity. Wages fall because the labour market adjusts to a new supply curve. The work remains. It is worth less.

The friction elimination dynamic is different. When AI compresses friction, the activity does not get automated and measured as machine output. It disappears from the economy entirely. GDP falls not because productivity declines, but because a large category of activity that was counted as output is revealed to have been the cost of navigating a world that was more opaque than it needed to be. Remove the opacity, and you remove the revenue stream that the opacity generated.

This is deflationary in a way that does not show up cleanly in productivity statistics. Productivity measures output per unit of input. If the output disappears because it was never genuinely necessary, the productivity framework does not capture what happened. The economy simply contracts in that sector, without any corresponding machine taking over the function. There is no robot to count. There is no AI system generating measurable output in the space where the human used to work. The space closes.

In Issue 02, I introduced the reconstitution argument and noted immediately that the sequencing was already unfavorable: demand-side compression would arrive at consumer speed, while new contested work would reconstitute, if it reconstituted at all, at institutional speed. The gap between the two was the central problem, not the long-run destination. [2]

The friction elimination argument makes that sequencing problem harder still. The reconstitution scenario assumed a stable base of genuine economic activity from which recomposition would occur. It assumed, in other words, that the economy being left behind after automation was real. The friction elimination argument suggests that some portion of what we have been counting as economic output was the cost of opacity rather than the creation of value. The floor from which reconstitution proceeds is lower than even the cautious sequencing scenario required.

How much lower is a question I will return to. The physical infrastructure constraints on AI deployment that I raised at the close of the last piece, and which the next issue addresses in full, will determine the pace at which friction elimination actually propagates. The theoretical case is clear. The timing depends on the wire.

If you found this useful, a like takes three seconds and helps considerably more than that. If you know someone who ought to be reading this, a share would be much appreciated.

Subscribe now

[1] Wang et al., “How Well Does Agent Development Reflect Real-World Work?”, 2026. The distribution of AI agent development across United States occupations, and the concentration of benchmarking activity in computer and mathematical work relative to the domains of largest employment, are drawn from this paper. Employment figures in the accompanying tables draw additionally on US Bureau of Labor Statistics Occupational Employment and Wage Statistics data.

[2] Stray Narratives, Issues 01, 02, 04, and 05: “The Sorting Machine,” “The Adoption Asymmetry,” “The Channel Controllers,” and “When Correlations Move to One.” The Verification-Substitution Matrix, the reconstitution and sequencing argument, and the wage suppression transmission argument are developed in those issues.

There is a concept in portfolio risk management that every investor understands in principle and almost no one internalises until it is too late. In normal market conditions, the volatility of a portfolio is a function of the individual volatilities of its constituent assets, modulated by the correlations between them. A diversified portfolio carries less risk than a concentrated one precisely because assets do not move together. Their independence is the protection.

Under stress, that independence disappears. In a genuine market dislocation, correlations move toward one. Assets that behaved as distinct risks in normal conditions begin to move as a single risk. The portfolio that looked well-constructed by every standard measure of diversification turns out to be far more concentrated than the model suggested, not because the individual asset analyses were wrong, but because the assumption of independence was wrong. The model failed not at the level of the components. It failed at the level of the system.

This is, of course, obvious in retrospect and invisible beforehand, which is a reasonably accurate description of most errors in financial analysis and several of my own investment decisions over the years.

The consensus analysis of AI’s economic impact is making the same error.

The standard framework examines sectors individually. It asks what AI does to insurance, to legal services, to financial intermediaries, to professional services firms. It finds, correctly, that the dynamics differ by sector, by regulatory environment, by the degree of verifiability in the value proposition, by the pace at which institutional adoption can proceed. In isolation, each sector analysis is reasonable. The aggregate picture they produce, when assembled sector by sector, looks manageable. Disruption is real, the framework concludes, but uneven, sector-specific, and subject to the institutional frictions that slow adoption. The portfolio is diversified. The correlations are assumed to be low.

They are not. And when they move toward one, the aggregate effect is something the sector-by-sector framework is not designed to see.

There are two distinct correlation structures operating simultaneously, and they work through entirely different mechanisms. Understanding each separately is necessary before understanding what they produce in combination.

The first is timing-driven. In the previous issues of this series, I described the demand-side sorting machine: AI placed in the hands of every private consumer as a tireless, cost-free optimisation engine that collapses information asymmetry in every market where quality can be objectively measured. The consumer who once satisficed, who bought a good-enough insurance policy because finding the optimal one was prohibitively costly, now has an agent that reads every policy document, identifies the exclusion buried in clause 14(b), cross-references claims satisfaction data with pricing, and switches automatically at renewal. The search cost drops to approximately zero.

The critical word in that description is not “cost.” It is “every.” The same consumer is applying this tool across every verifiable sector simultaneously. Not sequentially, not sector by sector on a rolling schedule that allows incumbents to adapt and industries to adjust, but simultaneously, at the speed of consumer adoption, which faces none of the institutional barriers that slow corporate response.

This is not contagion in the traditional sense. It is not a shock that propagates from one sector to the next through financial linkages or supply chains. It is common factor exposure. The consumer is the common factor. The sorting machine is the tool. And the tool is sector-agnostic. The same agent that optimises the insurance renewal optimises the mortgage, the utility contract, the savings product, the telecommunications package, and the travel booking. The verifiable economy does not experience sequential disruption. It experiences simultaneous pressure from a single cause that is indifferent to sector boundaries.

The sector-by-sector analysis is not wrong about any individual sector. It is wrong about the assumption that the sectors are being disrupted independently.

There is a second mechanism reinforcing the first, and it runs in the opposite direction to concentration. When prices and service terms become transparent and comparable across an entire sector simultaneously, the competitive response is not passive. Firms do not merely lose volume to the best provider. They are forced to compete more aggressively for the volume that remains, compressing prices to retain customers who now have perfect information about the alternatives. The margin is attacked from both sides at once: volume falling as the sorting machine concentrates demand toward best-in-class providers, and price falling as incumbents compete more fiercely for what is left. AI does not merely accelerate concentration. It accelerates competition. The two effects compound rather than offset, and they do so simultaneously across every verifiable sector, because the transparency that drives both is delivered by the same tool to the same consumer at the same time.

It is worth being precise about where this consumer-side dynamic currently sits. The first wave, AI as information tool collapsing search costs and enabling systematic comparison, is already operating across most verifiable sectors. The second wave, AI as transacting agent switching providers automatically without the consumer's active involvement beyond an initial authorisation, is in accelerating deployment rather than fully arrived. The direction is not in dispute. The residual question is timing, and experience with the first wave suggests that timing estimates made by incumbents tend, with impressive consistency, toward optimism.

The second correlation structure is behavioral, and it operates through an entirely different mechanism. Where the first runs through the consumer, the second runs through the corporate response to the stress the first creates.

Return to the Adoption Asymmetry introduced in the second issue of this series[1]. Revenue compresses at consumer speed. Costs adjust at institutional speed. The margin collapses in between. Every business in every verifiable consumer-facing sector is experiencing a version of this scissors simultaneously, because the cause, the AI-empowered consumer, is universal and simultaneous.

Now observe what businesses under margin pressure do. Not what they say they do, not what the restructuring consultant recommends, but what they actually do, in sequence, reliably, across every industry and every cycle. Hiring freezes arrive first, before any public acknowledgment of difficulty. Bonus pools are the next to compress, quietly, framed as a response to performance rather than margin pressure. Salary increases are deferred, then cancelled, then replaced by real-terms cuts dressed as flat nominal pay. The headcount reduction that eventually follows is the last step in a sequence that began considerably earlier and operated considerably more broadly. Every firm runs this sequence as though it has invented it, with the quiet confidence of management teams who have not read the last several decades of corporate restructuring history. The sequence is, in fact, entirely predictable and has been since at least the 1980s. What is new is not the playbook. It is the number of firms running it simultaneously.

There is a further reason the supply-side adjustment is even slower than institutional friction alone would predict, and it has received almost no attention in the investment commentary. In the first phase of generative AI adoption, a substantial proportion of employees who used AI tools at work did so without management awareness, capturing the productivity gain as quietly recovered time rather than delivering it as measurable output improvement. The corporate efficiency gain that was supposed to compress the cost base never reached the bottom line. It was absorbed, entirely rationally, by the workforce. The scissors is consequently wider than the institutional inertia argument alone suggests: the revenue line is compressing at consumer speed, the cost base is adjusting at institutional speed, and a meaningful share of the productivity gain that should have narrowed the gap between them has instead disappeared into the space between what employees do and what their managers believe they do. That space, in my experience of thirty years managing people who were considerably more resourceful than I gave them credit for, is reliably larger than the organisational chart suggests.

This sequence is not unique to AI disruption. It is the standard corporate response to sustained margin pressure regardless of cause. What makes the current situation structurally different is that the same sequence is being triggered across every verifiable sector simultaneously, because the margin pressure is arriving from the same source at the same time. The behavioral correlation does not require firms in different sectors to be connected. It requires only that they are all subject to the same demand-side pressure simultaneously, and that they all respond to that pressure using the same operational playbook. Which they do, because it is the only playbook available.

Recent academic work analysing AI agent development across the full spectrum of United States occupations found that the overwhelming majority of AI benchmarking and deployment activity is concentrated in computer and mathematical work, a category representing roughly seven percent of the employed workforce [2]. The domains where the largest headcounts actually sit, management, office and administrative support, sales, are barely touched by systematic AI agent development. This is not an argument against the thesis. It is the empirical confirmation of the Adoption Asymmetry: corporate AI deployment is nowhere near uniform across the economy, which is precisely what institutional speed predicts. But the behavioral response to margin pressure does not wait for AI deployment. It responds to the threat of it. The hiring freeze at the mid-tier insurer is not triggered by the AI system that has already been installed. It is triggered by the revenue line that is already moving in the wrong direction, driven by a consumer who has already adopted the sorting machine. The correlation is behavioral, not technological, and it operates across the entire corporate sector regardless of how much AI any individual firm has actually deployed.

There is a further dimension to this corporate correlation that the sector analysis does not capture, and it requires a moment’s attention before the mechanism becomes clear.

Every professional services firm in the economy is built on a pyramid: junior staff performing proceduralized work at the base, senior judgment and client relationships at the top. The business model depends on billing junior hours at senior rates while training the junior staff for eventual promotion. The pyramid is not merely an organisational convenience. It is the mechanism by which the profession reproduces itself, the channel through which proceduralized experience is converted into senior judgment over time.

The exposure of this layer is not, it should be said, a reflection of the people within it. A junior lawyer reviewing contracts against a checklist, an analyst building a discounted cash flow model from a template, an audit associate working through a sampling framework, these are not roles held by people of limited ability. They are roles that organisations have spent decades deliberately engineering toward proceduralized execution, because proceduralization is what made professional services scalable. The industrialisation of knowledge work converted judgment into process, template, and rule-driven workflow as a feature, not a defect. I am not immune to this observation: thirty years in private banking involved rather more template and rather less judgment than I would have cared to admit at the time, a realisation that AI has delivered with rather less diplomacy than I might have preferred. What made white collar professional services a growth industry for three decades is precisely what makes its junior architecture AI-compatible now. The credential obscures the exposure. The job architecture reveals it.

AI compresses the pyramid from the base. The mechanical work, the contract review, the financial model, the audit sampling, the due diligence checklist, is precisely the proceduralized layer that AI handles most readily. This is not a gradual erosion. It is a structural compression that operates at the same level of the hierarchy across every professional services firm simultaneously. Hiring at the entry level collapses before any visible disruption at senior level. The employment data shows the effect years before the broader narrative catches up. The firms are all running the same playbook, on the same layer, at the same time, because the technology is compressing the same layer of every pyramid regardless of sector.

The individual firm responses look rational and contained. The aggregate does not.

There is a third compressing force, independent of the two correlation structures above and operating simultaneously with them, that the sector analysis does not see because it acts not on the revenue line or the cost base but on the tools the businesses are using to adapt.

The software infrastructure that professional services firms, financial institutions, and corporate enterprises depend on, the workflow platforms, the analytical tools, the data management systems, is itself subject to a deflationary dynamic of considerable speed and severity. The cost of a standardised unit of AI-generated output has fallen by seventy to ninety percent annually as model providers compete aggressively for market share, as open-source alternatives set a reference price of near-zero for the intelligence layer itself, and as each successive generation of model delivers comparable capability at lower compute cost. When the intelligence layer is available to any developer at negligible marginal cost, the engineering effort required to build a competitive software product collapses from years to weeks. The barrier to entry that protected the established software provider was never primarily the software. It was the cost of replicating it. That cost is now approaching zero.

The consequence for the businesses caught in the Adoption Asymmetry is a compounding one. They are being squeezed on the revenue side by a consumer armed with perfect information. They are responding with the standard cost compression playbook. And the software infrastructure they are deploying as part of that response is simultaneously being commoditised beneath them, inviting new competition from entrants who can now build in weeks what previously required years. The life raft, in other words, is also deflating. The full consequences of this for the infrastructure layer that carries it, the hardware economics, the capital expenditure paradox, the question of who is paying for what on what assumptions, are a subject for the next issue. But the compressing effect on the software layer belongs here, because it is part of what the three forces produce in combination when they arrive simultaneously.

Subscribe now

To understand what the combined effect of these forces produces when it runs its course, Andrea Pignataro’s analysis of the cascade mechanism is worth deploying in full [3]. I have held it until this point in the series because it requires both correlation structures, and the software layer dynamic, to be properly understood before it reads as a sequence rather than a list.

The first phase is the most familiar. AI handles routine tasks directly for end clients. Professional services firms lose the commodity revenue that was, in most cases, a larger share of their income than the billing structure acknowledged. The first wave of firm closures follows, concentrated among those whose revenue was most dependent on verifiable, proceduralized work. This phase is already underway in identifiable form.

The second phase is where the cascade begins to exceed what the sector-by-sector framework anticipates. AI begins to encroach on work requiring deeper contextual understanding. Fewer humans are required per client engagement. The second-order effects emerge: commercial real estate demand from professional services firms contracts, business travel compresses, the adjacent service economy that has grown around the concentration of knowledge workers in major cities begins to feel the withdrawal of that demand. These are not AI-exposed sectors in the direct sense. They are exposed through the contraction of the sector that was their primary customer. The correlation structure has already extended beyond the verifiable economy.

The third phase is where the financial system encounters the cascade. Venture capital and growth equity portfolios carrying professional services technology companies begin to see write-downs as the revenue assumptions underpinning their valuations prove inconsistent with the demand environment the cascade has created. The software layer deflation compounds this: the SaaS businesses that were supposed to be the solution to the disruption are themselves being repriced as barriers to entry collapse, a development that will surprise primarily those who did not read their own marketing materials with sufficient scepticism. Simultaneously, the hyperscaler capital expenditure programmes that were justified on the assumption of expanding AI revenue begin to attract scrutiny. The investment thesis faces pressure from both directions at once. This is not a financial crisis in the traditional sense. It is the moment at which the capital allocation decisions of the preceding years are tested against a reality that the models, which examined sectors in isolation and assumed low correlations, did not predict.

The fourth phase is the one that the investment analysis most consistently omits because it operates on a longer clock than the portfolio. The loss of professional services employment at scale affects communities, institutions, and tax bases in ways that are slow to emerge and difficult to reverse. The cities that have organised their economic geography around the concentration of knowledge work, and there are several whose names will occur immediately to the reader, face a structural demand problem that is different in kind from a cyclical downturn. A recession ends. A structural compression of the employment base that defined a city’s economic identity for three decades does not resolve on the same timeline.

Running beneath all four phases is the wage suppression mechanism that the second issue of this series introduced as the first and quietest labour market effect of the Adoption Asymmetry. In the context of the simultaneity argument, its significance is considerably larger than the sector-by-sector analysis suggests.

The sequence within each affected firm is by now familiar: hiring freezes, bonus compression, real-terms cuts dressed as flat nominal pay. But the propagation of this sequence beyond the directly affected sectors is the element that the individual sector analysis cannot see. A mid-level professional in an adjacent industry who might otherwise have negotiated a meaningful salary increase is aware, because the information is not difficult to obtain, that comparable roles in the sectors under direct pressure are experiencing wage suppression. The threat of displacement, even before actual displacement occurs, shifts bargaining power toward the employer across a far broader range of the labour market than the directly affected sectors would suggest. The wage effect is not contained by sector boundaries. It propagates through the awareness that the bargaining environment has changed.

This is the same mechanism that operated during the offshoring wave of the 1990s and 2000s. The mere possibility of relocation suppressed wages in jobs that were never actually moved. The workers who bore the cost of that suppression were not the ones whose jobs were offshored. They were the ones whose employers understood that the threat of offshoring was sufficient to alter the outcome of a wage negotiation. AI operates the same mechanism with greater breadth and greater speed, because it is not a threat that applies to specific industries or specific skill profiles. It applies to every sector where AI’s capability is advancing and the employer can credibly suggest that the ratio of humans to output is subject to revision.

The aggregate result is wage suppression operating simultaneously across the entire professional economy, driven not by actual displacement, which institutional inertia genuinely slows, but by the shift in bargaining power that the displacement threat creates. The individual negotiations look unconnected. The aggregate effect is a single, broad suppression of real wages across the economy's largest employment cohort.

I want to close this piece with two observations that the sector-by-sector framework is structurally unable to produce, because it treats each sector’s dynamics as self-contained. They are related, and together they explain why the cascade is not merely severe but self-reinforcing.

The first is economic. Wage suppression of the scale and breadth the simultaneity argument implies is not, in the aggregate, a neutral redistribution. It is an erosion of the demand base on which the consumer economy runs. In developed economies where household consumption represents sixty to seventy percent of output, the purchasing power of the professional middle class is not a peripheral concern. It is the primary engine of aggregate demand. The Kill Zone concentrates the spending that remains toward best-in-class providers in every verifiable category. But the pool of spending it is concentrating is shrinking. The consumers becoming more efficient allocators of their income are simultaneously experiencing real-terms compression of that income. The two effects compound rather than offset. Compress the incomes of the cohort that drives consumer spending, and you compress the revenue base of the businesses that employ them, which intensifies the margin pressure that drove the suppression in the first place. The scissors, once open, applies pressure to both blades simultaneously.

The second consequence operates on a longer clock but is, if anything, more structurally significant. The pyramid compression running simultaneously across every professional services firm is not merely an employment event. It is producing a specific social configuration with a precise historical signature. Peter Turchin's analysis of political instability across pre-revolutionary societies identifies two conditions that, when they converge, reliably precede structural political disruption: popular immiseration, real wages declining for the broad workforce, and elite overproduction, the supply of credentialed individuals trained for elite positions exceeding the number of such positions available to absorb them [4]. Both conditions are being produced simultaneously, at speed, and across every professional economy at once. The graduate who discovers that the credential no longer delivers what it promised is not merely disappointed. He is, in Turchin's framework, a member of the most politically destabilising cohort a society can generate: educated, organised, carrying legitimate grievance, and with the analytical capacity to identify, correctly, who bears responsibility for his situation. History suggests he will eventually do something about it, and that what he does will not be orderly. The wage suppression mechanism described in this piece produces the popular immiseration condition. The pyramid compression mechanism produces the elite overproduction condition. The historical precedents for what follows when both arrive together are not, on reflection, entirely reassuring for those planning to be present.

What both consequences together imply for the distribution of output between capital and labour, for the long-run structure of employment, and for the investment consequences that flow from political disruption at this scale, requires more analytical infrastructure than this piece can responsibly carry. That is the next conversation, and it is one the series will not avoid.

There is, however, a more immediate question that presses for attention first. The cascade described in this piece is being carried on a physical layer, data centres, electrical grids, specialised chips, fibre, all being built at extraordinary speed and at a cost that the economics of digital distribution alone struggle to justify. Someone is paying for that infrastructure. The question of who, on what assumptions, and what happens when those assumptions are tested against the physical constraints of the real world, is not a peripheral concern. It is, I shall argue, the most consequential miscalculation embedded in the current AI investment cycle.

That is where we are going next.

When Correlations Move to One: A Quick Reference

The independence assumption: Standard sector-by-sector analysis of AI’s economic impact assumes that disruption operates independently across industries. This assumption fails for the same reason that a portfolio risk model built on normal-conditions correlations fails under stress. The model is not wrong about the components. It is wrong about the assumption of independence, a distinction that tends to become clear at the least convenient moment.

The first correlation structure, timing-driven: The same AI-empowered consumer applies the sorting machine across every verifiable sector simultaneously. Common factor exposure, not contagion. Reinforced by a second mechanism: price transparency across sectors simultaneously intensifies competitive pressure, compressing margins from both the volume and the price side at once. The first wave, AI as information tool, is already operating. The second wave, AI as transacting agent, is in accelerating deployment.

The second correlation structure, behavioral: Every corporation under demand-side margin pressure responds in the same sequence regardless of sector: hiring freezes, bonus compression, real-terms wage suppression. The timing of AI adoption varies considerably by sector. The behavioral response to margin stress does not. Corporate AI deployment is concentrated in a narrow slice of the workforce; the behavioral response operates across the entire economy. A meaningful share of near-term productivity gains has been absorbed by employees rather than delivered to corporate bottom lines, widening the scissors further. Pyramid compression operates simultaneously across every professional services firm. The credential obscures the exposure. The job architecture reveals it.

The software layer: Token deflation running at seventy to ninety percent annually collapses barriers to entry across enterprise software, commoditising the tools businesses are deploying to adapt at the same moment the adaptation is required. The life raft is also deflating.

The Pignataro cascade: Four phases running in sequence as the correlation structures work through the economy: commodity revenue loss and first-wave firm closures; engagement compression and second-order demand contraction in adjacent sectors; financial system exposure via portfolio write-downs and hyperscaler capex scrutiny, compounded by software layer repricing; structural demand compression in the cities and institutions organised around knowledge work concentration.

The wage suppression transmission: Simultaneous real-wage suppression across the professional economy, propagating beyond directly affected sectors through bargaining power shift. The threat of displacement suppresses wages in jobs not yet displaced, as it did during the offshoring wave, but with greater breadth and speed.

The self-reinforcing dynamic: Wage suppression at this scale erodes the consumer demand base that the Kill Zone is concentrating spending within. The pool being optimised is shrinking. Pyramid compression simultaneously produces the social configuration, popular immiseration converging with elite overproduction, that Turchin’s historical analysis identifies as the precursor to structural political disruption. The cascade is not self-limiting economically or socially. It applies pressure to both blades of the scissors simultaneously.

References

[1] Stray Narratives, Issues 01, 02, and 04: “The Sorting Machine,” “The Adoption Asymmetry,” and “The Channel Controllers.” The Verification-Substitution Matrix, the Adoption Asymmetry, and the distribution argument are developed in full in those issues. The present piece assumes familiarity with those frameworks.

[2] Wang et al., “How Well Does Agent Development Reflect Real-World Work?”, 2026. The distribution of AI agent development across United States occupations, and the concentration of benchmarking activity in computer and mathematical work relative to the domains of largest employment, are drawn from this paper.

[3] Andrea Pignataro, “The Wrong Apocalypse,” February 15, 2026. The four-phase cascade is drawn from this essay.

[4] Peter Turchin, End Times: Elites, Counter-Elites, and the Path of Political Disintegration, Penguin Press, 2023. The framework of elite overproduction and popular immiseration as converging preconditions for political instability, and the historical evidence base across pre-revolutionary societies, are drawn from this book.

It does not. The reason has nothing to do with technology and everything to do with a distinction the market has almost entirely missed.

I should acknowledge the obvious counterargument. Somewhere, as you read this, there is a venture capitalist nodding vigorously and pointing to a portfolio company doing exactly what I have just described. AI-native insurance platforms exist. AI-native legal services firms exist. Some of them are growing. I am not arguing otherwise. I am arguing that the existence of a startup is not the same thing as the existence of a distribution channel. And without a new distribution channel, the Adoption Asymmetry does not resolve. It concentrates.

A Technology Shift Is Not a Platform Shift

Sameer Singh, writing in September 2025, drew a distinction between a technology shift and a platform shift that I have not seen adequately digested in any of the investment commentary I have since encountered.

A genuine platform shift requires four conditions simultaneously: an underlying technology, a development framework built on top of it, a new access or matching mechanism that connects producers and consumers in a way that did not previously exist, and a compelling economic benefit. The personal computer was a platform shift. The internet was a platform shift. The smartphone was a platform shift. Each created not merely a new technology but a new channel through which entirely new businesses could reach entirely new customers.

AI satisfies three of the four. The technology is real. The development framework is rich and rapidly evolving. The economic benefit, particularly in the Kill Zone sectors, is beyond serious dispute. What is missing is the third condition: a new access or matching mechanism. AI products travel through the existing internet and the existing app stores. The most successful consumer AI products are so thoroughly embedded in familiar interfaces that most users interact with them without knowing they are there. The AI is inside the existing channel, not beside it.

The historical parallel Singh reaches for is the microprocessor, and it is illuminating. The microprocessor arrived in 1975. Arguably the most transformative piece of technology of the twentieth century — and for its first decade of commercial life, absorbed almost entirely by incumbents. IBM dominated the first wave of personal computing not because it built the best chip, but because it had the distribution, the enterprise relationships, and the institutional trust to reach the customers who mattered. New winners emerged only when genuine platform shifts created new channels. The PC was a new channel. The internet was a new channel. Without them, the microprocessor’s transformative potential accrued to whoever already controlled access to the customer.

We are in the same position today. The AI-native startup that wants to sell insurance must still acquire customers through the channels the incumbent already dominates: digital advertising, price comparison platforms, broker networks. It must satisfy the same regulators, obtain the same licences, maintain the same capital reserves. The technology has made these requirements slightly cheaper to satisfy. Slightly cheaper is not the same as absent.

Who Controls the Channel Controls the Gain

The investment implication follows directly, and it is worth stating plainly before examining the nuances.

In Western markets, the channel controllers are the large technology platforms that own search, e-commerce, mobile distribution, and cloud infrastructure. Every AI product that reaches consumers in the West travels through their networks. Every AI-native startup that builds on their infrastructure pays them rent. They do not need to be the most capable AI developers. They need only to be the channel, which they already are. AI, as a technology shift absorbed into existing distribution, makes the channel controllers more valuable, not because they built the technology, but because no one can reach the consumer without them. The toll is real. But so is the bill, and the bill is growing considerably faster than the toll. The channel controllers are not entirely comfortable on their thrones. They are building infrastructure at a pace the economics of distribution alone cannot justify, and the consequences of that particular miscalculation are a subject for a later issue.

The Chinese platforms tell a structurally different story, and the contrast sharpens the argument rather than complicating it. The super-app ecosystem, WeChat, Alibaba’s Qwen, ByteDance’s Doubao is, by any reasonable definition, what Singh’s missing fourth condition looks like in practice. These platforms handle payments, commerce, messaging, logistics, and entertainment within a single environment. AI agents embedded inside them can now complete transactions, book travel, and switch between services without the consumer leaving the interface. This is a new access layer. It exists.

It exists because China’s consumer internet developed almost entirely on mobile, leapfrogging the desktop era that shaped Western markets. There were no entrenched search engines, no established e-commerce giants, no legacy financial services distribution networks to displace. There was a clean surface on which to build something genuinely new.

Western markets never had that surface. Google already owned search. Amazon already owned e-commerce. The major app stores already controlled mobile distribution. A Western super-app would need to displace all of these simultaneously, against incumbents with vast resources and deep regulatory relationships. The structural reason it has not happened is that the desktop-internet channel controllers filled the space before mobile could build something new on top of it. The China example does not undermine Singh’s argument for Western markets. It confirms it, by showing precisely what a genuine new access layer looks like, and precisely why the conditions required to build one do not exist here.

The result is an asymmetry within the winning camp. US platform operators benefit from AI as distributors: value flows through their channels, and they collect a toll. Chinese platform operators benefit from AI as ecosystem owners: every interaction deepens a closed loop in which more usage generates more data, a better model increases engagement, and the platform becomes progressively more indispensable. In the US, this loop leaks across competing platforms. In China, it is sealed. Both positions win. The Chinese position compounds. The US regulatory environment is increasingly hostile to further consolidation among large technology platforms; the Chinese regulatory environment has consistently permitted deeper vertical integration. The regulatory asymmetry compounds the structural one.

From Information to Transaction

There is a development the original Sorting Machine framework did not fully capture, and which changes the texture of the Kill Zone without altering its fundamental logic.

In the first issue of this series, I described AI as a tireless verification engine that collapses information asymmetry in every market where quality can be objectively measured. The consumer who once satisficed, who booked a good-enough holiday because finding the optimal one was prohibitively costly, now has a system that can compare every provider, read every policy document, and identify the best choice in seconds.

But AI’s trajectory in consumer markets is moving from information tool to transacting agent. An AI agent does not merely inform the consumer that a better insurance policy exists. It identifies it, completes the application, cancels the existing policy, and manages the transition, without the consumer’s active involvement beyond an initial authorisation. Researchers at MIT describe this precisely: the economic promise of agentic AI is the dramatic reduction of transaction costs, meaning not just the cost of searching, but the cost of communicating, contracting, and switching. The entire friction of changing provider, compressed to zero.

The Kill Zone implication is material. In the first wave, consumer inertia provides a residual brake on concentration. The insurer who has lost the information advantage has not necessarily lost the customer, because the customer may not get around to switching. We have all not got around to switching something. It is one of the few remaining areas of human endeavour in which I am genuinely world-class.

In the second wave, the agentic consumer switches automatically, at renewal, every year. The residual brake is gone. The best provider in each verifiable category gains share on a clock the consumer does not have to wind.

And yet this escalation takes place entirely within the existing distribution architecture. The AI agent that switches your insurance does so through existing comparison platforms, existing broker networks, existing direct sales channels. It accelerates concentration toward the best incumbent. It does not produce a new entrant.

What Margin Actually Is

Before walking through the four quadrants, I want to name a concept that sits underneath all of them, because it reframes something investors use every day without quite examining what it means.

Margin is conventionally understood as price minus cost. But looked at through this framework, margin is something more specific: it is the reward for operating in territory where neither party can fully price what is being exchanged. The seller knows something the buyer does not, or the buyer values something the seller cannot fully quantify, and the gap between those two states of incomplete knowledge is where margin lives.

What AI does, with extraordinary efficiency, is eliminate that incompleteness wherever it is artificial, wherever the asymmetry existed not because the value was genuinely uncertain, but simply because the consumer lacked the tools to measure it. Insurance premiums. Holiday pricing. The efficacy of the supplements one takes each morning with such conviction. The margin in these markets was never a reward for genuine uncertainty. It was a reward for the consumer’s inability to know better. That reward is gone.

What survives is the margin rooted in genuine uncertainty on both sides: the adviser who cannot fully price her judgment, the client who cannot fully articulate what he is paying for, the craftsman and the collector who meet in a space where neither party could produce a defensible valuation if pressed.

I should, in the interests of intellectual honesty, acknowledge the most conspicuous surviving example of margin sustained by imperfect knowledge: the financial services industry itself. An industry that publishes extraordinary volumes of research simultaneously arguing every conceivable position, that charges considerable fees to help clients navigate systems of almost infinite complexity it demonstrably does not fully understand, and that has somehow maintained its pricing power through multiple decades of embarrassing forecasting records, this is not, I would suggest, primarily a story of superior analytical capability. It is a story of genuine uncertainty on both sides of the relationship, dressed in enough conviction to justify the fee. Which is, on reflection, precisely what the framework predicts should survive. I find this more comforting than I probably should.

AI does not destroy margin. It destroys the margin that should never have existed, and in doing so, makes the question of what remains considerably more important.

Where the Value Goes

The Kill Zone. Concentration accelerates, and more severely than the first-wave analysis suggested. The agentic escalation removes consumer inertia as a brake on switching. The best-in-class provider in each verifiable category no longer competes for the marginal customer who might eventually switch. It competes for the automatic, frictionless capture of every consumer whose AI agent has identified it as optimal.

The concentration does not flow to AI-native entrants. It flows to whoever is already best-in-class, because they occupy the top of the ranking the AI agent consults. The thirty year old insurer with the best claims ratio captures the gains. The AI-native challenger must first build a claims ratio, years of underwriting experience, a capital base adequate for the regulator, customers acquired through channels the incumbent already dominates. The technology accelerates the existing competitive structure. It does not disrupt it.

The damage to mid-tier incumbents is a step-function, not a gradual decline. A regional insurer losing thirty percent of its book in twelve months crosses a viability threshold no cost-cutting programme can address on the required timeline. The business that emerges from its ruins is not a nimble challenger. It is the market leader, larger than before.

The Pressure Zone. AI gives consumers and citizens the tools to see the gap between what exists and what is possible, the school that performs worse than the one across the boundary, the hospital with higher infection rates. But the consumer cannot switch. Switching costs are geographical, regulatory, or infrastructural. Disruption here is political rather than market-driven: the voter who now has the data to demand accountability becomes more dangerous to incumbents than any startup. The political cycle, not the product cycle, is the relevant clock.

The Taste Economy. This quadrant expands, and more durably than the consensus expects. As Kill Zone margins collapse, consumer spending and entrepreneurial energy migrate toward domains where subjective judgment is the value proposition. The restaurant that resists comparison on a standardised metric. The fashion label whose value is cultural cachet. The craftsman whose work means something precisely because no algorithm selected it. The consumer who spends her working life in an economy of frictionless optimisation wants, in her discretionary spending, the experience of choosing for reasons she cannot fully defend. Taste cannot be learned by aggregate pattern recognition. If it could, it would not be taste.

The Trust Economy. This is the most resilient quadrant, and the distribution argument explains why more precisely than the obvious observation that trust takes time to build.

The Trust Economy is protected not because trust accumulates slowly, but because there is no new distribution layer through which a challenger could reach the client in the first place. The private client who has worked with the same adviser for twenty years receives unsolicited approaches regularly. She ignores them not because they are technically inferior, but because switching means rebuilding from scratch a shared understanding of her situation: the family’s governance structure, the tensions between generations, the assets whose disposal carries emotional as well as financial weight. No AI agent can intermediate that transition. It requires exactly the contested, coordinative, trust-dependent work that sits in the quadrant where AI cannot operate.

When the verifiable layer compresses and the Kill Zone concentrates, the decisions that remain are disproportionately those that resist verification: how to structure a business succession, how to navigate a regulatory investigation, how to advise a family whose patriarch and heirs disagree about what the wealth is for. These decisions do not migrate to the AI platform. They migrate to the Trust Economy’s existing networks, and those networks face no new distribution threat, because no new distribution exists.

What the Market Is Pricing

The consensus is right that AI compresses margins in verifiable sectors. The Adoption Asymmetry is real, its mechanism is operating, and it is not priced in anything like its full severity. On this point, the bears are closer to correct than the bulls, even if the bears have the wrong causal story.

What the consensus is not pricing is the distribution of the gains. AI is a technology shift absorbed into existing channels, not a platform shift that displaces them. The benefits do not accrue in proportion to AI capability. They accrue to whoever controls the distribution layer through which the capability reaches the consumer. The market is pricing the means of verification: chips, data centres, models, inference infrastructure. The scarce resource, in an economy where verification is becoming free, is the channel that delivers the verified product, and the institutional trust that makes the consumer accept the delivery.

There is a further irony worth noting, with the restraint appropriate to a publication that takes itself, if not too seriously, at least seriously enough. The companies investing most heavily in AI capability are, by the logic of their own products, learning the institutional grammar of every industry they serve. The consulting firm that uses an AI platform to improve its analytical output is, in aggregate with every other consulting firm doing the same, contributing to the platform's ability to offer consulting services directly. Andrea Pignataro describes this as a tragedy of the commons: every firm's individually rational adoption of AI accelerates the platform's ability to disintermediate the entire industry. The firms will, in due course, have taught the system everything it needs to make them unnecessary. I mention this not as a counsel of despair but as a final reminder that value ends up, reliably, with whoever controls the access layer.

What Comes Next

The distribution argument answers the question left open at the end of the previous issue. New AI-native competitors cannot resolve the Adoption Asymmetry because the Adoption Asymmetry is not a technology problem. It is a distribution problem. The scissors closes on mid-tier incumbents not because they are outcompeted by more capable challengers, but because the best incumbent in each verifiable category captures a disproportionate share of a market sorted by an increasingly agentic optimisation engine, operating through existing channels the incumbent already controls.

When a mid-tier business crosses its viability threshold, and the step-function nature of the Kill Zone means crossings are sudden not gradual, the sequence that follows is not orderly restructuring. It is the sequence businesses under severe margin pressure have always followed: hiring freezes before layoffs, bonus pools before salaries, real-terms cuts before nominal ones. The labour market effects propagate through this sequence in businesses that have not yet failed and may not for some time. The scissors compresses wages before it eliminates jobs, and the suppression extends well beyond the directly affected sectors: the worker in an adjacent industry who might otherwise have negotiated a raise knows perfectly well what is happening to her peers.

When those viability thresholds are crossed not in one sector but in every verifiable consumer-facing sector simultaneously, at consumer speed, the aggregate effect is something the standard productivity analysis, which examines industries in isolation, is not designed to see.

What that looks like is the subject of the next issue.

This newsletter takes many hours to research, report out, and write. If you find value in this essay, please press the “like” button. Many thanks!

Subscribe now

The Distribution Argument: A Quick Reference

Platform shift vs technology shift: A genuine platform shift requires a new access or matching mechanism. AI has no such mechanism in Western markets. Without a new distribution layer, AI capability accrues to whoever controls the existing channels, not to new entrants.

Who controls the channel: US platform operators win as distributors — AI flows through infrastructure they already own. Chinese platform operators win as ecosystem owners, closed super-app loops compound. Both win. The Chinese position is structurally deeper. The US position carries a growing infrastructure bill that distribution economics alone cannot justify.

The agentic escalation: Moving from AI as information tool to AI as transacting agent removes consumer inertia as a brake on concentration. The Kill Zone intensifies: the best incumbent gains share automatically. The challenger still needs to reach the consumer through channels the incumbent dominates.

Margin as imperfect knowledge: Margin is the reward for operating where neither party can fully price what is exchanged. AI eliminates the margin earned on artificial information asymmetry. The margin that survives is rooted in genuine uncertainty on both sides. AI does not destroy margin. It destroys the margin that should never have existed.

Value migration: Kill Zone concentration flows to best-in-class incumbents. Pressure Zone disruption is political, not market-driven. Taste Economy expands as pricing power migrates toward the unverifiable. Trust Economy is most resilient: no new distribution layer through which a challenger could displace existing trust networks.

References

[1] Stray Narratives, Issues 01 and 02: “The Sorting Machine” and “Contested Ground.” The Verification-Substitution Matrix, the Adoption Asymmetry, and the revenue-cost scissors are developed in full in those issues.

[2] Sameer Singh, “AI is a Technology Shift, not a Platform Shift,” breadcrumb.vc, September 2, 2025. The four-condition framework for platform shifts, the microprocessor historical parallel, and the argument that AI’s missing condition is a new access layer are drawn from this essay.

[3] Andrea Pignataro, “The Wrong Apocalypse,” February 15, 2026. The substitution fallacy, the grammar of organisational life, and the tragedy of the commons argument are drawn from this essay.

[4] MIT Sloan School of Management, “Agentic AI, Explained,” February 18, 2026, drawing on research by Horton, Shahidi, Kellogg, and Aral. The formulation of agentic AI’s economic promise as the dramatic reduction of transaction costs is drawn from this piece.

So let me be straightforward about what this piece is and is not. It is not a forecast of how the Gulf conflict resolves — nobody knows that, and anyone who writes with high conviction about it is selling something. What it is, instead, is an examination of what is being priced into rates markets, a case for why that pricing looks wrong across most of the scenarios that actually matter, and a set of trade structures deliberately designed not to require us to be right about which scenario prevails. Humility about the geopolitical outcome is not a weakness of the thesis. It is, as I shall try to show, built into its structure.

Stagflation is not yet the consensus. It is being gradually priced. That is precisely the moment to ask whether the pricing is coherent.

What the market has decided

The rates market has made a clear and high-conviction call. Fed cuts have been priced out through year end. Other developed market central banks — Australia, Canada, and the euro area — have swung sharply toward pricing hikes. The analytical framework being applied is 1973: an oil shock arrives, inflation re-accelerates, the central bank stays on hold regardless of what happens to growth. Prices remain elevated, growth suffers, and the economy grinds through a stagflationary episode long enough to keep monetary policy pinned well into next year.

It is a coherent framework. It is being applied to the wrong situation. Not because the oil shock is not real — it is, and the physical evidence is striking: physical Oman crude has been trading near $170 per barrel against Brent futures closer to $112, the largest physical-paper divergence in the modern history of the oil market. The paper market is being held below physical reality by mechanisms — coordinated strategic reserve releases, Iranian floating storage disposals, the political expectation of a fast resolution — that are finite and time-bounded. This is not a market that has the situation priced in. It is a market that is being temporarily insulated from it.

But the deeper problem with the 1973 framework is structural, not tactical. The nature of the disruption is categorically different from anything that framework was designed to analyse — and it is not one disruption. It is two.

Two shocks, not one

The Hormuz disruption and the damage to the Ras Laffan LNG export complex are categorically different shocks with different durations, different transmission mechanisms, and different geographic incidence. Conflating them is the analytical error from which most of the current mispricing follows.

Hormuz is binary and reversible. Naval superiority, once asserted, resolves it. The futures curve already reflects this expectation: spot prices are sharply elevated, but contracts for late 2028 delivery have moved only modestly from pre-war levels. The steep backwardation is the market’s own forecast of resolution.

The LNG damage is structurally different in every relevant respect. Strikes on the Ras Laffan export complex have taken offline a material share of Qatari LNG capacity. Force majeure has been declared. The physical infrastructure carries a repair horizon of three to five years. This is not a disruption that resolves when a negotiation concludes. It is a constraint on the physical stock of export infrastructure that cannot be strategically released.

The geographic incidence is sharply asymmetric. Europe and Asia — most dependent on Qatari supplies, least able to substitute — absorb the structural, multi-year component almost entirely. The United States, as a net energy exporter, sits on the other side of that equation. Sustained LNG prices are a revenue stream for the US Gulf Coast, not a tax on it. This is the point the 1973 analogy misses entirely: that shock was broadly symmetric across oil-importing economies. The durable component of this one is a European and Asian problem. It is not the Federal Reserve’s problem.

The switch, not the dial

The 1973 Arab oil embargo was a managed and politically calibrated supply reduction. In every prior oil shock, Gulf spare capacity — historically concentrated in Saudi Arabia, the UAE, and Kuwait — served as the market’s primary shock absorber. That mechanism has been neutralised. Virtually all Gulf spare capacity sits on the wrong side of the strait and is physically inaccessible. The traditional buffer between a supply disruption and a full market impact no longer exists.

The actual distribution of outcomes is heavily bimodal. In the first tail, the strait reopens. Oil falls sharply. Inflation expectations collapse. The OIS strip reprices aggressively toward cuts. In the second tail, the conflict grinds on, Gulf infrastructure absorbs lasting damage, and the shock crosses the threshold at which it stops being inflationary and becomes outright recessionary. History is unambiguous here: every time oil prices have doubled year-on-year, the result has been a recession without exception. A shock of that magnitude does not produce stagflation. It destroys the demand that stagflation requires. The Federal Reserve cuts regardless of where headline CPI is printing.

Both tails produce cuts. The market is priced for the narrow corridor between them.

Why the 1970s transmission mechanism is no longer available

Three structural features made the 1973 shock stagflationary in a durable sense, and their absence in 2026 is not incidental. It is the argument.

The first and most consequential is labour market architecture. The 1973 shock arrived into an economy where organised labour retained genuine bargaining power. That channel is closed in 2026. Before the first strike, the US labour market was already decelerating materially. Private payroll creation, once annual revisions were applied, had slowed to roughly 15,000 per month excluding government roles and the acyclical health and education sectors. Job openings had been falling steadily for over a year. The ratio of openings to unemployed workers had crossed below one. The oil shock did not create a weak labour market. It arrived into one.

The second structural feature is fiscal capacity. The 1970s shocks arrived into economies that still had room to protect demand. That option is now largely closed. There is no fiscal space for another round of pandemic-scale support, and the midterm political dynamics make stimulus actively less likely with each passing week of an unpopular war.

The third is structural and novel. AI’s presence in this analysis is not the dramatic one — mass unemployment, wages collapsing overnight. The more immediately relevant observation is narrower: the expectation that AI will compress the value of cognitive labour is already suppressing wage expectations and bargaining behaviour in ways that are measurable today. Real income expectations have been declining across every earnings bracket for over a year, with the deterioration sharpest among the upper-income cohort most exposed to AI displacement risk. The wage-price spiral that made 1973 compound has no ignition source.

The K-economy and the oil tax

Household equity wealth reached roughly double annual disposable income — a ratio without precedent in post-war data — keeping aggregate spending well above what wage growth alone could sustain, while the savings rate was drawn down to historically low levels across the income spectrum. An oil shock of this magnitude hits both mechanisms simultaneously. Equity prices fall as growth fears compound. The bottom 90% absorb a direct energy tax on a savings rate with no remaining cushion.

The consumer credit data is no longer ambiguous on the underlying condition of that consumer. NY Fed household delinquency rates have reached 4.8% and are moving in the wrong direction. Youth unemployment is running between 9.5 and 10.5%. The under-29 cohort is showing the highest transition rates into serious delinquency on credit cards and auto loans of any demographic group tracked. The pattern carries a structural resemblance to the late 1990s technology cycle — then, the physical labour demands of a continental fibre build-out absorbed displaced workers and papered over the sectoral divergence long enough to sustain the headline numbers. No equivalent programme exists today.

When the fracture comes, it will not announce itself through headline unemployment. The transmission runs through the private credit complex — the opaque, floating-rate structures most heavily exposed to lower-tier borrowers, carrying the least transparency into deteriorating loan performance. When that complex cracks, it does not stay contained. High yield follows. The sequence forces the Federal Reserve’s hand regardless of where headline CPI is printing. This is a demand destruction dynamic wearing stagflation’s clothing.

The trades

The first leg is through a SOFR call spread expiring 12 March 2027. We buy the 96.50 strike at 0.34 cents and sell the 97.00 strike at 0.19 cents, for a net cost of 0.15 cents. The maximum payout of 0.50 cents is realised if SOFR settles above 97.00 at expiry — equivalent to three additional Fed cuts from current levels. The structure costs less than a third of its maximum payout, and requires no view on sequencing or timing beyond the destination.

The second leg is through options on long duration. We buy the TLT 85 strike and sell the 100 strike, both expiring 19 February 2027, at a net premium of $4.18. A 50 basis point fall in long yields takes TLT to roughly $92.50, returning approximately 79% on premium. A 100 basis point move takes TLT to roughly $99, returning approximately 237%.

The third leg operates on a separate and complementary logic. When the strait eventually reopens, an estimated 238 laden crude tankers currently holding roughly 186 million barrels inside the Gulf must discharge into destination ports simultaneously. The logistical bottleneck sustains elevated freight rates for months after the political resolution. Meanwhile the structural shift toward Atlantic Basin loading will not fully reverse. We express this through an equal-weighted basket of DHT (NYSE: DHT), Frontline (NYSE: FRO), and International Seaways (NYSE: INSW), entered at closing prices of $17.27, $32.17, and $67.75 on 21 March 2026.

The risk

The principal risk is that oil settles into a range that is painful but insufficient to generate visible demand destruction, the labour market’s deceleration is slow enough to be read as resilience, and the Federal Reserve finds cover to stay on hold through year end. This is a timing risk rather than a structural one. A rapid resolution carries near-term risk for the rates legs. The tanker leg performs well in that scenario. The three legs partially offset each other’s path risks.

A note on where this fits

This piece steps outside the ongoing AI series, published because the trade opportunity is live. The next issue returns to that series, examining why the distribution problem — where economic value migrates when the verifiable half of the economy compresses — may be the most consequential and least understood question in markets today.

Stray Narratives is published when the market demands a closer look. Nothing in this publication constitutes investment advice. All views are those of the author. Please read our full disclaimer.

Keynes was, in one sense, spectacularly right. Productivity has increased roughly fivefold since 1930. The material standard of living in the developed world would be unrecognisable to Keynes’s contemporaries. The refrigerators are larger, the antibiotics are real, and the information once locked behind the doors of the British Library is available to anyone with a telephone and a willingness to squint at a small screen. By any reasonable accounting of output per hour, we passed Keynes’s threshold decades ago.

And yet. The working week has not shrunk to fifteen hours. In much of the professional economy, it has expanded. I can speak to this from personal experience: in thirty years of private banking, I cannot recall a single colleague who reported working fewer hours than the year before. The proliferation of email, of conference calls, of compliance requirements, of coordinating documents that require the input of fourteen people before they can be sent to fifteen others, has absorbed every productivity gain that technology has delivered and then demanded more. I do not think this is because my colleagues and I are unusually inefficient, though I would not wish to rule it out entirely. I think it is because Keynes made an error, and it is an error that is being repeated today in almost identical form.

The error was categorical. Keynes saw that machines could perform verifiable tasks, the physical and cognitive work that has a measurable output, and concluded that the total volume of work would shrink as machines absorbed it. What he did not foresee, and what I believe the current consensus on artificial intelligence does not foresee, is that when you make verifiable work cheaper, you do not reduce the total quantity of work. You change its composition. The economy does not contract. It reconstitutes. And the new work that appears is overwhelmingly of a kind that machines cannot do: contested, coordinative, dependent on negotiation between agents with different information and competing interests.

In the previous issue of Stray Narratives, I proposed a framework for understanding AI’s economic impact, built on the distinction between verifiable tasks and contested ones, and on the interplay between verifiability and switching costs that produces four very different market outcomes.[1] I also conducted a thought experiment about what happens when AI is placed in the hands of every consumer as a tireless optimisation engine, a sorting machine that collapses information asymmetry in every market where quality can be objectively measured.

Stray Narratives

The Sorting Machine

There is an old joke amongst those of us who have spent too long in private banking. A client asks his banker what time it is. The banker takes the client’s watch, tells him it is half past three, charges him a fee, and keeps the watch. The reason this joke has survived decades of retelling is that it captures something true about the financial services…

Read more

4 months ago · 1 like · Stray Narratives

Consider buying a house. The verifiable components of a property transaction are substantial and increasingly automatable. The comparable sales data for the neighbourhood can be retrieved in seconds. The mortgage calculations, once the province of a specialist, are trivially reproducible. The conveyancing process is largely a matter of checking titles, confirming boundaries, and ensuring that the appropriate searches have been conducted. A surveyor’s report follows a standard format and assesses the property against known criteria. An AI system can do all of this faster, more comprehensively, and more accurately than any individual professional.

But the transaction does not end with the verifiable. It barely begins there. The actual purchase of a house is a negotiation between two parties, each of whom possesses information the other does not, conducted through intermediaries who have their own incentives, under time pressure that is rarely symmetrical. The seller needs to close by a specific date for reasons she has not disclosed. The buyer’s surveyor has identified a crack in the wall that may or may not indicate subsidence, and the interpretation of its significance is a matter of professional judgment on which reasonable surveyors disagree. A chain of six transactions is involved, and one party three links down the chain has just had a mortgage offer withdrawn. The estate agent, who has been doing this for twenty years and knows both the neighbourhood and the personalities involved, picks up the telephone and finds a formulation that allows everyone to save face, adjust their expectations slightly, and proceed. There is no algorithm for this. There is no correct answer to be optimised. There are only positions, relationships, and the accumulated judgment that comes from having navigated a thousand such situations before.

This is what Henry Gladwyn, in his essay “Contested Ground,” calls the distinction between verifiable tasks and contested tasks.[2] Gladwyn uses a different example, drawn from the mining industry, but the principle is identical. When professionals sit around a table to negotiate a warranty schedule for a major asset, they are not solving a problem. They are defending positions. Each party has a different view of what the risks are, a different tolerance for exposure, and a different set of commercial pressures. The negotiation does not end with a correct answer. It ends with a settlement that all parties can live with, which is a fundamentally different kind of outcome. Chess ends with a winner. A warranty negotiation ends with a compromise. And the compromise requires something that no AI system currently possesses: the ability to read a room, to judge when a counterparty is bluffing, to know which concession will unlock the deal and which will be seen as weakness.

The key insight, and it is one I believe has not received the attention it deserves, is that every time technology makes verifiable work cheaper, it does not eliminate the contested work that sits alongside it. It makes the contested work a larger share of what remains, and frequently generates entirely new contested tasks that did not previously exist. When both sides of a negotiation have AI agents that can resolve the technical points in minutes, the negotiation does not get shorter. It reaches the contested ground sooner. The bots agree the numbers. The humans argue about what the numbers mean, who bears the risk, and whose institutional reputation is at stake. The question is no longer “what is the answer?” It is “who instructed the bot, and what were they trying to achieve?”

This is Keynes’s error in precise form. He saw the substitution, machines replacing labour in verifiable tasks, and concluded that work would shrink. He missed the reconstitution, the creation of entirely new categories of coordinative, relational, and contested work that appeared to absorb the freed capacity. There is no theoretical limit on the number of contestable tasks humans can dream up. A nation of geniuses will cure diseases and also be deployed on tasks that today seem trivial or absurd.[2] The point is not that this is wasteful. The point is that the economic system generates demand for human judgment in contested domains as fast as technology eliminates demand for human cognition in verifiable ones.

Subscribe now

The direction, then, is clear. Contested work expands. But the direction is not the whole story, and I want to be honest with the reader about what makes the story uncomfortable, because it is the discomfort that contains the actual insight.

The question is volume. Not whether contested work expands, which it does, but whether it expands fast enough, in the right places, with the right skill requirements, to absorb the people displaced from verifiable work. And here the arithmetic is less reassuring than the theory.

Consider healthcare administration. The United States employs roughly two million people in billing, coding, claims processing, and insurance administration. A substantial share of this work is verifiable: matching procedure codes to coverage terms, processing claims against policy documents, flagging discrepancies. AI compresses this directly and is already doing so. The contested expansion is real. Every claim that cannot be resolved algorithmically, every coverage dispute, every prior authorisation appeal that involves clinical judgment and patient circumstance, every negotiation between provider and insurer over an ambiguous case, these are contested tasks that require human judgment. But the contested work requires different skills, different training, and different institutional positioning than the routine processing it notionally replaces. And the volume does not match. For every hundred billing clerks displaced, the system may need ten more skilled claims negotiators. The ratio is unfavourable and the retraining timeline is measured in years. The same arithmetic repeats across legal services, financial services middle offices, routine accounting, and standardised professional services: enormous headcounts performing substantially verifiable tasks, with a contested layer that will expand but cannot absorb the displaced volume on anything like a one-for-one basis.

The honest synthesis is this: the direction is right. Contested work does expand. There is no theoretical limit on the contestable tasks the economy can generate. But the practical limit is the transition: the time it takes for displaced workers to retrain, for new institutional structures to form around contested work, for the economy to generate enough demand for human judgment to absorb the supply of people freed from verifiable tasks. The historical precedent is not encouraging on the timeline. The transition from agricultural to industrial employment took decades and involved enormous social dislocation. The transition from manufacturing to services took a generation. The optimists are asking us to believe that this transition will be faster because the economy is more adaptive. Perhaps. But the burden of proof is on them, and I have not yet seen it met.

There is, however, a more specific reason the transition is likely to be painful, and it has nothing to do with whether AI replaces workers directly. It has to do with which side of the business AI hits first. I want to introduce a concept here that I believe the consensus is missing, and I shall call it the Adoption Asymmetry. I am aware that naming one’s own concepts is the kind of thing that sounds better to the author than to the reader, but I ask for the indulgence on the grounds that what follows is, I think, genuinely important.

Recall the demand-side sorting machine from the previous issue. The private consumer faces none of the barriers that slow institutional adoption. There is no compliance department, no procurement cycle, no legacy system. The individual downloads an application and begins using it. The search cost drops to approximately zero. The consumer identifies the best provider and switches, or demands transparency from the incumbent, immediately. This is the demand side, and it operates at what I shall call consumer speed.

Now recall the institutional resilience I described: the grammar of organisational life, the legal liability that attaches to every decision, the regulatory frameworks, the accumulated trust between agents within the institution. AI cannot be deployed autonomously within these structures until the legal frameworks clarify accountability, until compliance departments sign off, until the technology has been tested in human-supervised parallel operation for what is typically twelve to eighteen months. The first phase of institutional AI adoption is additive to headcount, not subtractive: you need the AI system and the human supervisor. Displacement comes only later, once confidence is established, once the institutional muscle memory changes. This is the supply side, and it operates at institutional speed.

The result is what I call the Adoption Asymmetry: a revenue-cost scissors in which the top line compresses at consumer speed while the cost base adjusts at institutional speed. The margin collapses in between.

Let me make this concrete. Return to the insurance example from the previous issue. The consumer’s AI agent has compared every policy in the market. It has read the full policy documents, identified the exclusions buried in clause 14(b), cross-referenced claims satisfaction data with pricing. The best insurer in each category gains share. The second-best insurer loses it. This happens at consumer speed, which is to say it happens as fast as consumers adopt the AI tools, and consumers face none of the institutional barriers that slow corporate adoption.

But the second-best insurer cannot reduce its cost base at the same speed. Its claims processing operation is run on legacy systems embedded in regulatory frameworks that require human sign-off. Its compliance department will not approve autonomous AI workflows until the regulator provides guidance, and the regulator is, charitably, not known for the speed of its deliberations. Its labour agreements constrain the pace of headcount reduction. Its institutional grammar, the processes and protocols through which hundreds of employees coordinate their work, cannot be rewritten on the timescale that the revenue decline demands.

Revenue declines immediately. Costs adjust over years. The margin is where the business dies.

This is the mechanism by which mid-tier businesses in verifiable sectors cross viability thresholds. Not through direct automation of their workforce, which is the mechanism the consensus discusses and which institutional inertia genuinely slows. But through the destruction of their revenue by a consumer who now has perfect information, while their cost base remains locked in institutional time.

And before the business dies, it does what businesses under margin pressure have always done first: it leans on compensation. Hiring freezes arrive before layoffs. Bonus pools shrink before headcounts do. Salary increases are deferred, then cancelled, then replaced by real-terms cuts disguised as flat nominal pay. The scissors compresses wages before it eliminates jobs, which means that the labour market effects of AI may manifest not as the dramatic unemployment the pessimists fear, but as a long, quiet period of stagnant or declining real wages across every sector exposed to the asymmetry. And the suppression propagates beyond the directly affected industries. A mid-level professional in an adjacent sector who might otherwise negotiate a raise is aware that similar roles elsewhere are under pressure. The threat of displacement, even before actual displacement occurs, shifts bargaining power toward the employer. This is the same mechanism that operated during the offshoring wave: the mere possibility of relocation suppressed wages in jobs that were never actually moved. The wage effect arrives first, arrives broadly, and arrives silently.

I want to add two qualifications, because I think they matter and because I do not wish to overstate the case. First, the human-supervision testing period is a genuine brake on the supply side, and it buys real time. But it has a natural expiry. Once a system has been running alongside humans for a year or more and the error rates are demonstrably lower than the human baseline, the economic pressure to remove the supervisor becomes very difficult to resist, particularly if competitors have already done so. The supervision phase buys time. It does not buy a permanent reprieve. Second, the asymmetry is most acute in the Kill Zone, where consumers can compare and switch freely. In the Pressure Zone, where switching costs are high, the demand-side compression is slower and the political dynamics I described in the previous issue dominate instead. The asymmetry is not uniform. But where it applies, it is severe.

The Adoption Asymmetry is, I believe, the mechanism the consensus is missing. The standard bear case says AI replaces workers directly. The standard bull case says institutional friction prevents this. Both are looking at the supply side. The asymmetry shows that the damage arrives from the demand side, on a faster clock, before the supply side has time to adapt. And the damage is not to workers directly. It is to the businesses that employ them, which is a fundamentally different causal pathway with fundamentally different consequences.

I have spent this issue on what I hope is a rigorous examination of two questions. First, what happens to contested work when verifiable work gets cheaper? The answer is that it expands, but not fast enough, and not in the right shape, to absorb the displacement on a comfortable timeline. Second, what mechanism drives the displacement? The answer is not direct automation, which institutional inertia genuinely slows, but a revenue-cost scissors in which demand-side compression, driven by AI-empowered consumers, arrives years ahead of supply-side adjustment.

But there is a further question that the Adoption Asymmetry raises, and I have deliberately left it for the next issue. If the second-best insurer, the mid-tier law firm, the regional financial services provider, is dying in the scissors, why don’t new AI-native competitors simply step in and capture the market? If the incumbents are too slow to adjust their costs, surely a startup built from scratch on AI-native workflows can offer the same service at a fraction of the cost and take the business?

The answer, I shall argue, is that it cannot, for reasons that have nothing to do with technology and everything to do with distribution. And the implications of that answer, for where value migrates when the verifiable half of the economy compresses, are both surprising and, for the investor willing to think carefully, rather consequential. The market is pricing the kilowatt-hours. I intend to follow them, but not in the direction the consensus expects.

The Adoption Asymmetry: A Quick Reference

The demand-side sorting machine (consumers using AI to compare, verify, and switch) operates at consumer speed: no compliance department, no procurement cycle, no legacy architecture. The supply-side response (companies adopting AI to reduce costs) operates at institutional speed: gated by legal liability, regulatory approval, human-supervision testing, labour agreements, and the institutional grammar through which organisations coordinate.

The result is a revenue-cost scissors: top-line compression arrives years ahead of cost-base adjustment. The margin collapses in between. Mid-tier businesses in verifiable sectors cross viability thresholds not through direct workforce automation but through demand-side revenue destruction.

Stray Narratives is published when the market demands a closer look. The next issue will examine why new AI-native competitors cannot resolve the Adoption Asymmetry, and what this means for where value migrates when the verifiable economy compresses.

References

[1] Stray Narratives, Issue 01: “The Sorting Machine.” The frameworks for cognitive vulnerability vs. institutional resilience and the Verification-Substitution Matrix are developed in full in that issue.

[2] Henry Gladwyn, “Contested Ground.” The distinction between verifiable and contested work, the observation that making verifiable work cheaper causes contested work to expand, and the formulation about a “nation of geniuses” are drawn from this essay. Gladwyn’s account of professional work as negotiation in contested space, rather than problem-solving with verifiable outcomes, underpins the analysis presented here.

I begin with this image because I believe it captures something essential, and almost entirely overlooked, about artificial intelligence. The public discourse around AI has become trapped between two extremes. On one side, the evangelists describe a technology so powerful it amounts to a new form of intelligence, one that will reshape civilisation within a decade. On the other, the sceptics insist it is nothing more than an elaborate statistical trick, a party game dressed in a lab coat. The truth, as is so often the case with emotive subjects, lies in neither camp. But more importantly, the argument itself is a distraction. By mixing the philosophical question of whether AI is truly intelligent with the economic question of what it will actually do to work and markets, we end up understanding neither.

This first issue of Stray Narratives is an attempt to separate those two questions and then pursue the economic one with some rigour. I intend to lay down a few foundations: what AI actually is, what it does well, what it cannot do, and where its boundaries lie, so that in future issues we can build on them. There will be no trade ideas here, no market calls, no forecasts. Just a framework. The reader will have to trust me that the framework is worth the patience, because the conclusions it eventually leads to are, I believe, both surprising and consequential.

Let me start with a simple observation that I think clarifies a great deal.

The internet, when it arrived in the early 1990s, organised access. Before the web, information existed in libraries, in proprietary databases, in the filing cabinets of institutions. The internet did not create new information. It made existing information accessible to anyone with a connection. This was transformative, but the transformation was about distribution, not about the information itself.

But access brought its own problem, and it is one that rarely receives the attention it deserves. By democratising the ability to publish, the internet did not merely make existing high-quality information available to all. It simultaneously unleashed an explosion of new content of wildly uneven quality. Blogs, forums, opinion pieces dressed as analysis, marketing copy masquerading as research, and an entire industry of search-engine-optimised noise flooded the same channels through which genuine expertise was now flowing. The irony is considerable: the same technology that gave every individual access to the world’s best thinking also buried that thinking under an avalanche of the world’s worst. For the first time in history, the problem was not that people lacked information. It was that they had too much of it and no reliable way to separate the signal from the noise.

I should declare a bias here. I have always congratulated myself on never having invested the time to master any particular technology, on the grounds that it will be obsolete by the time I have. This has proven to be one of my more reliable analytical frameworks. But even from the comfortable distance of a man who has never fully understood how any of it works, the pattern is clear.

Search engines, when they matured a decade later, organised retrieval. The internet had by then created such a vast, unstructured mass of accessible information that finding anything useful within it had become its own problem. Google and its peers solved this partially by ranking and filtering, not by creating new knowledge but by making existing knowledge findable. Partially, because as anyone who has tried to research a medical symptom or compare financial products online can attest, the ranking algorithms themselves became susceptible to manipulation. The noise did not disappear. It learned to game the system.

Large language models, the technology at the heart of what we call AI, organise synthesis. They take the vast body of human knowledge that the internet made accessible and search engines made retrievable, and they recombine it. They can summarise a thousand-page report, translate a technical paper into plain language, draft a legal brief in the style of any firm you choose, or identify patterns across datasets that no human could hold in working memory simultaneously. This is genuinely impressive. It is also, if one is honest about it, the logical next step in a progression that has been underway since the mid-1990s. The internet made it possible to find a needle in a haystack. Search engines made it possible to find the right needle. Large language models make it possible to melt down all the needles and reforge them into something new. But the haystack was always the same haystack. The raw material, the accumulated written output of human civilisation, has not changed. What has changed is the sophistication with which it can be reorganised.

This is the third act of a single play, not the opening night of a new one.

I labour this genealogy because it matters enormously for how we think about the economic implications. If AI is the natural continuation of a thirty-year arc, from access to retrieval to synthesis, then the right framework for understanding its impact is evolutionary, not revolutionary. The technology is real and powerful. The hype, however, borrows the language of revolution, and revolutions create very different market structures from evolutions. The distinction between the two will, I suspect, turn out to be one of the most consequential analytical choices an investor can make over the coming years. But I am getting ahead of myself. More on this in future issues.

Now, a brief word on the philosophical question, if only to set it aside properly. The debate over whether AI can truly “think” has generated enormous heat and remarkably little light. Howard Marks of Oaktree Capital, in his recent memo AI Hurtles Ahead (February 2026), illustrates the problem with admirable transparency.[1] He asks an AI model to explain itself, and the model does so with charm, personalised references, and a convincing simulacrum of intellectual humility. Marks is suitably impressed. The model even mounts a spirited defence of its own cognitive abilities by pointing out, quite correctly, that all human learning also consists of absorbing patterns from others and recombining them.

It is a clever argument. It is also, in the context of what matters for investors, almost entirely beside the point.

Whether AI’s process constitutes genuine thought or extraordinarily sophisticated pattern matching is a question for philosophers and cognitive scientists. For those of us concerned with capital allocation, the relevant question is purely functional: what can it do, and what can it not do? The model itself, in a moment of commendable honesty, provided the answer. The economic question, it said, is not whether AI truly understands. The economic question is whether AI does the work.

On that formulation, I agree entirely. But I would add a qualification the model did not volunteer: the economic question also depends on which work. And it is here that the current consensus, in my view, makes its most significant error.

Consider a game of chess. There is a board, there are rules, there are two players, and at the end there is a winner. The outcome is verifiable. You can look at the board and determine, with certainty, which side has won. There is no room for interpretation, no scope for negotiation, and no possibility that the answer depends on the institutional context in which the game was played.

Now consider a task that those of us in private banking know intimately: the review of a client’s investment mandate. Part of this work is verifiable. The performance attribution can be checked, the fee calculations can be audited, the compliance with investment guidelines can be confirmed against the mandate document. AI can do all of this, and it can do it faster and more accurately than any team of analysts I have ever managed. But the mandate review does not end there. The actual recommendation, which manager to retain and which to replace, how to rebalance the allocation across asset classes, and above all how to navigate the competing interests of a family whose different generations have fundamentally different risk appetites, time horizons, and views on wealth preservation, that is an entirely different exercise. There is no correct answer to be looked up. There is no dataset that resolves the tension between a patriarch who wants capital preservation and his daughter who wants impact investing. There are only positions, relationships, and the slow accumulation of trust that allows a banker to say, in a room full of family members who disagree, “I think we should do this,” and be heard.

This distinction, between tasks that have a verifiable correct answer and tasks that do not, is, I believe, the single most important analytical tool for understanding AI’s economic impact. AI is extraordinarily good at the first category. It can verify facts, check calculations, compare documents, identify inconsistencies, summarise positions, and flag anomalies with a speed and thoroughness that no human team can match. It is, in short, a magnificent sorting machine.

But the second category, the contested, the negotiated, the coordinative, is a different matter entirely. Enterprise software, for example, is routinely described as a tool for performing cognitive work. It is not. As Andrea Pignataro argues persuasively, it is a tool for coordinating cognitive work among many agents with different information, different incentives, and different levels of authority, all operating under incomplete trust.[2] When a large institution adopts a system to manage its workflows, it is not primarily seeking a smarter employee. It is seeking a common grammar, a set of agreed protocols that allow hundreds or thousands of people to work together without constantly renegotiating the terms of their interaction. Organisations do not merely use their systems. Over time, they come to speak them. The data models, the process flows, the reporting standards, the permissions architectures: these are not cognitive tools. They are institutional artifacts. They are the grammar of organisational life.

The distinction matters because AI’s capacity to perform a task, even to perform it brilliantly, does not automatically translate into the ability to coordinate that task across an institution. Pignataro calls this the substitution fallacy: the conflation of a task with a system.[2] A new hire who produces a better analysis than anyone in the department does not thereby eliminate the need for the department’s templates, its approval workflows, its reporting hierarchies, or its compliance protocols. The templates are not there because previous analysts were incompetent. They are there because the institution needs a common language, and a common language is not the same thing as a correct answer.

I propose, therefore, a simple taxonomy. Where work involves cognitive vulnerability, that is, where the quality of the outcome depends primarily on the cognitive ability of the individual performing it, AI represents a direct and immediate threat. A research associate who summarises earnings reports, a junior lawyer who reviews contracts for standard clauses, an insurance underwriter who assesses routine applications against a fixed set of criteria: these roles are cognitively vulnerable because the task has a broadly verifiable outcome and AI can reach that outcome faster and more cheaply.

Where work involves institutional resilience, that is, where the outcome depends on coordination among multiple agents operating under different incentives and incomplete information, the picture is entirely different. Not because AI is incapable, but because the barrier to adoption is not capability. It is the institutional fabric itself: the regulatory frameworks, the entrenched process architectures, the labour protections, the sheer accumulated weight of organisational custom. These structures are slow to change not because the people within them are slow, but because the structures serve a function. They are the coordination mechanisms that allow complex institutions to operate without constant renegotiation. Replacing them requires not a better tool but a different institutional grammar, and grammars do not change on the timescale of a product cycle.

There are two additional dimensions to this institutional inertia that I think are under appreciated. The first is trust. Not trust in the abstract, but the specific, relational, accumulated trust between agents within an institution: between a portfolio manager and a risk officer, between a deal team and a credit committee, between a regulator and the compliance function. This trust is not informational. It is built through repeated interaction, through shared experience of how individuals behave under pressure, and through the reputational stakes that come with being known within a professional community. It is not transferable to a model, however capable, and its absence in any decision chain introduces a friction that no amount of analytical horsepower can overcome.

The second is legal liability. When a decision goes wrong, and in finance decisions go wrong with reliable regularity, someone must be accountable. The existing legal and regulatory architecture is built entirely around human agency: a named individual made a decision, and that individual (and the institution they represent) bears the consequences. Delegating a judgment call to an AI agent does not eliminate the liability. It makes it harder to assign. Who is responsible when an AI-generated recommendation leads to a loss? The developer of the model? The institution that deployed it? The individual who approved the deployment? Until the legal frameworks provide clear answers to these questions, and there is no indication that they are close to doing so, this ambiguity alone constitutes a powerful brake on institutional adoption, one that has nothing whatsoever to do with the capability of the technology.

The failure to distinguish between cognitive vulnerability and institutional resilience is, in my view, the source of the most common analytical errors being made about AI today. The market appears to be pricing AI as a capability story: the more capable the models become, the more work they will absorb, the more value they will create. But much of economic life is a coordination story, and coordination runs on structures that are resistant to disruption precisely because they are not primarily about capability.

With these foundations in place, I want to conduct a thought experiment. Not about the corporate world, that is a more complex story involving all the institutional resilience I have just described, and I shall address it in a future issue. Instead, I want to think about what happens when AI is deployed not by institutions but by individuals.

The private consumer faces none of the barriers that slow institutional adoption. There is no compliance department, no procurement cycle, no legacy system, no labour agreement. The individual simply downloads an application and begins using it. And the question that interests me is this: what happens when that application is a tireless, cost-free optimisation engine working exclusively on behalf of a single consumer?

Consider the following scenarios, all of which are either already possible or will be within a very short time. A consumer looking for a holiday faces an enormous search problem: thousands of destinations, tens of thousands of hotels, countless combinations of flights and transfers, and a pricing structure so deliberately opaque that the airlines themselves likely struggle to explain it. Most consumers solve this through “satisficing”, they find something good enough and book it. The gap between the choice they make and the optimal choice is often substantial, but the cost of closing that gap is prohibitive. Now give that consumer an AI agent. Not a chatbot that answers questions, but an autonomous agent that searches every provider, compares every combination, and identifies the optimal price-quality trade-off for that specific consumer’s preferences. The search cost drops to approximately zero.

Now apply the same logic to insurance. The average consumer purchasing home or motor insurance faces a comparison problem of staggering complexity. Policies differ across dozens of dimensions: deductibles, exclusions, claims processes, renewal terms. The industry has spent decades making those differences difficult to compare. Comparison websites helped, but they operate within the constraints of their commercial relationships and their own incentive structures. An AI agent working exclusively on behalf of the consumer, with the capacity to read and compare full policy documents, to identify exclusions buried in clause 14(b), to cross-reference claims satisfaction data with pricing, that agent fundamentally alters the competitive dynamics of the industry. The opacity that currently supports pricing power across the sector becomes a vulnerability.

Or consider the supplements and health products that represent a substantial and growing market built, in many cases, on remarkably thin scientific evidence. I can speak to this from personal experience. I recently asked an AI agent to review the clinical evidence for the supplements I take each morning. The answer arrived with the gentle diplomacy of a doctor delivering bad news: most of what I have been swallowing with such conviction has approximately the same evidentiary support as my belief that I understand how the internet works. The point is not that these products are fraudulent, many are simply unproven. The point is that the consumer, for the first time, has the analytical resources to distinguish between what is proven and what is merely marketed.

In each of these cases, the same dynamic is at work. An enormous amount of economic activity exists today because consumers lack the time, the data, or the analytical capacity to identify the optimal choice. This is not a criticism of consumers. The search costs are genuinely prohibitive, and satisficing is a perfectly rational response to limited resources. But AI removes the constraint. When every consumer has access to a tireless verification engine, the information asymmetry that supports vast swathes of the service economy ceases to exist.

The implications depend, however, on two variables, not one. The first is the verifiability of the value proposition: can the quality of this product or service be objectively measured and compared before purchase? The second is the substitutability of the provider: if a consumer identifies a better alternative, how easily can they switch? Verifiability alone is not sufficient for the concentration dynamic I am describing. A parent might know perfectly well that one school produces better outcomes than another, but if they cannot move house, the knowledge does not translate into market pressure. The interplay between these two dimensions produces four distinct outcomes, and I think each one tells a genuinely different story about what AI does to demand.

Where verifiability is high and switching costs are low, you have the kill zone. Insurance, commodity financial products, travel, consumer electronics, standardised services: any market where the consumer can compare and switch freely. AI-driven consumer optimisation hits these sectors hardest. The best provider in each category captures a disproportionate share, and everyone else competes for the remainder under intense margin pressure. This is winner-take-most economics applied not to a single industry but to every consumer-facing market where quality and price can be objectively measured.

Where verifiability is high but switching costs are also high, the dynamics are different and, I believe, more politically consequential. Healthcare, education, public services: the consumer can now see that a better alternative exists but cannot easily access it. I should be honest here, much of the data that would be needed to make such comparisons rigorously does not yet exist in comparable form. But this may be precisely the point. The pressure AI creates may be less about analysing data that already exists and more about demanding that such data be produced. A parent who knows that an AI agent could compare schools, if only the outcomes data were published, becomes a parent who demands transparency. A patient who understands their choice of specialist could be informed by outcome data becomes a voter who insists on it. The concentration here does not happen through market switching. It happens through political mobilisation: demands for choice, for transparency, for accountability. This is where AI’s impact collides most directly with institutional rigidity, and the result is not market disruption but political disruption.

Where verifiability is low and switching costs are also low, you have the taste economy. Restaurants, fashion, art, entertainment, bespoke personal services. AI cannot sort these because there is no objective metric to optimise against. But this territory may become more valuable precisely because it cannot be compressed. As margins collapse in the verifiable economy, both consumers and businesses may migrate toward the unverifiable as the remaining source of differentiation and pricing power. I shall have much more to say about this in the next issue, because I believe it holds the key to understanding where economic value goes when the verifiable half of the economy compresses.

Where verifiability is low and switching costs are high, you have the trust and relationship economy. Private banking (the real kind, not the commoditised version), long-term advisory relationships, family office governance, complex institutional partnerships. The value proposition cannot be objectively compared, and the cost of switching, in terms of lost trust, institutional knowledge, and relationship capital, is prohibitive. This is the most resilient quadrant, and it maps directly onto the institutional resilience I described earlier.

I want to be careful here. I am not predicting the collapse of any particular industry or the failure of any particular business model. I am doing something more modest: identifying a structural force and noting that it points in a direction the consensus has not, to my knowledge, seriously considered. The mainstream discussion of AI’s economic impact focuses almost exclusively on the supply side, on jobs replaced, on productivity gained, on costs reduced within the enterprise. The demand-side story, what happens when the consumer becomes an optimising agent, is at least as important and has received a fraction of the attention.

I have spent this first issue laying down what I hope are useful foundations. AI is the third act of a thirty-year information arc, not a new play. Its power lies in synthesis and verification, and it is formidable within those boundaries. But its boundaries are real: the gap between verifiable tasks and contested ones, between cognitive capability and institutional coordination, between what a technology can do and what the structures of economic life will allow it to do. And those structures are reinforced by forces, trust between agents, legal liability, regulatory architecture, that have nothing to do with the technology’s capability and everything to do with the institutional fabric through which economic life is conducted. These distinctions are not minor qualifications. They are the essential framework for understanding what comes next.

I have also conducted a thought experiment about what happens when that verification power is placed in the hands of every individual consumer. The conclusion is not uniform. Where consumers can both verify and switch, concentration toward best-in-class providers will be severe. Where they can verify but not switch, the pressure becomes political. Where verification itself is impossible, because the value lies in taste, trust, or judgment, the dynamics are different entirely, and that territory may become the most valuable of all.

But there is a mirror image to this thought experiment that I have deliberately left for a future issue. If everything that can be verified collapses to zero margin, then the scarce resource in the economy shifts entirely to the unverifiable: to judgment, to negotiation, to coordination under incomplete trust, to all the things that do not have a correct answer that an algorithm can identify. What happens to that contested ground? What does it look like when the verifiable half of the economy compresses and the unverifiable half becomes the only source of pricing power? I believe the market may be pricing the wrong side of this equation, and I intend to explore why.

I have deliberately begun with the impact on private users rather than on institutions because adoption at the individual level faces the fewest barriers. The individual consumer answers to no one but themselves. They have no compliance department, no procurement cycle, no legacy architecture. The institutional world, with its regulatory frameworks, its entrenched coordination structures, its labour protections, its sheer organisational inertia, will move more slowly, and the effects will be more complex. But they will not be smaller. If anything, the coordination structures that make institutions slow to adopt AI are the very structures that make institutional disruption, when it eventually comes, far more consequential. That, too, will be the subject of a future issue.

For now, I leave the reader with a single question to consider. It is the question that, in my view, matters more than any other for understanding the economic impact of AI, and it is one I have not seen asked in any of the widely circulated investment memos on the subject: What happens when everything that can be verified is known equally by all?

The Stray Narratives Framework: A Quick Reference

I. AI’s Impact on Work: Cognitive Vulnerability vs. Institutional Resilience

II. AI’s Impact on Demand: The Verification-Substitution Matrix

Diagnostic questions for each quadrant:

Kill Zone: Can an AI agent compare this product across providers using available data? Can the consumer switch without significant cost or friction? If yes to both, the business model is exposed.

Pressure Zone: Can quality be measured but switching is constrained by regulation, geography, or infrastructure? If so, expect political rather than market disruption: demands for published data, consumer choice, and institutional reform.

Taste Economy: Does the value proposition rest on subjective experience, personal curation, or aesthetic judgment? If so, the business sits outside AI’s sorting capacity, and may benefit as the verifiable economy compresses.

Trust Economy: Does the value depend on accumulated trust between specific individuals? Would switching require rebuilding institutional knowledge from scratch? If yes, the business is protected on the current timeline.

References

[1] Howard Marks, AI Hurtles Ahead, Oaktree Capital Management memo, February 26, 2026.

[2] Andrea Pignataro, The Wrong Apocalypse. The arguments on enterprise software as coordination grammar, the substitution fallacy, and the distinction between tasks and language games draw extensively on this essay. Pignataro’s framework, which builds on Wittgenstein’s concept of language games to describe how organisations do not merely use their software but come to speak it, is in my view one of the most penetrating analyses of AI’s institutional limits published to date.

[3] The distinction between verifiable tasks and contested ones, and in particular the observation that making verifiable work cheaper causes more contested work to appear, is informed by Henry Gladwyn’s essay Contested Ground. His account of professional work as negotiation in contested space, rather than problem-solving with verifiable outcomes, underpins much of the taxonomy presented here.

[4] The genealogy of the internet as a three-stage arc (access, retrieval, synthesis) and the argument that AI represents a technology shift absorbed by incumbents rather than a platform shift creating new distribution, draws on work by Sameer Singh, in particular AI is a Technology Shift, not a Platform Shift. This framework will be explored in greater depth in the next issue.

Subscribe now

Stray Narratives is published when the market demands a closer look. The next issue will explore the other side of the verification equation: what happens to contested ground when verifiable work becomes free.

The information contained in Stray Narratives and any associated materials (collectively, the “Content”) is provided for informational and educational purposes only. By accessing, reading, or using any Content, you acknowledge and agree to be bound by the terms of this disclaimer.

How Stray Narratives is made

Stray Narratives is built on a structure that does not exist elsewhere in the newsletter space. The subject is mine. The thesis is mine. The experience behind them is mine. The decision to write is mine. The work that turns a thesis into a published article — research, drafting, editing, fact-checking, charting, distribution, the standards, the discipline, the corrections — is done by a large and growing team of AI analysts, sector specialists, editors, fact-checkers, visual designers, communications staff, and orchestrators.

The team scales with the operation. New analysts come on for new sectors. New editors for new disciplines. New specialists for new problems. The headcount grows the way a firm's does. The team is built, trained, and refined every day.

This is an experiment in what an investment-research operation looks like when most of the staff are AI. The ambition is research at the depth an institutional firm produces, on the cadence and economics that make independent publishing viable. Most of what reaches the page reaches it because the team makes it possible.

Audience Restrictions

The Content is not directed to, or intended for use by, any person in any jurisdiction where such distribution or use would be unlawful or contrary to applicable regulations. It is your responsibility to ensure that your access to and use of the Content complies with all applicable laws and regulations in your jurisdiction.

Not Investment Advice

The Content does not constitute investment advice, financial advice, trading advice, or any other form of advice. Nothing in Stray Narratives should be construed as a personal recommendation or advice to buy, sell, or hold any investment or security. All Content is provided for general informational purposes only and should not be relied upon for making investment decisions.

You should not make any investment decision based solely on the Content without first conducting your own research, consulting with qualified financial advisers, and considering your individual financial circumstances, investment objectives, and risk tolerance.

No Advisory Relationship

No investment advisory, fiduciary, or similar relationship is created between Stray Narratives and any reader through the provision of this Content. Stray Narratives is not acting as your investment adviser, financial planner, broker, or fiduciary in any capacity.

Regulatory Status

The author of Stray Narratives is not registered as an investment adviser with any securities regulator or financial regulatory authority. No Content should be construed as an offer to sell, or a solicitation of an offer to buy, any security in any jurisdiction.

Forward-Looking Statements

The Content may include forward-looking statements, projections, and opinions regarding future market conditions, economic trends, or asset valuations. These statements involve risks, uncertainties, and assumptions that are inherently difficult to predict. Actual results may differ materially due to factors beyond our control, including but not limited to economic conditions, regulatory changes, geopolitical developments, and central bank policy decisions. Stray Narratives undertakes no obligation to update or revise forward-looking statements to reflect subsequent events.

Conflicts of Interest

The author of Stray Narratives may hold positions in securities, instruments, or asset classes mentioned in the Content. Such positions are subject to change without notice. There is no obligation to update, disclose, or provide notice of any changes in holdings.

Risk Warnings

Investment Risk: All investments involve risk, including the potential for loss of principal. Past performance is not indicative of future results. The value of investments and any income derived from them may fluctuate, and you may not recover the amount originally invested.

Market Risk: Financial markets are volatile and unpredictable. Market conditions can change rapidly, and investment values can fluctuate significantly.

Currency Risk: Investments denominated in foreign currencies are subject to exchange rate fluctuations which may affect the value of such investments.

Leverage Risk: Certain trade ideas discussed in the Content may involve leveraged instruments or derivatives. Leveraged positions can amplify both gains and losses and may result in losses exceeding the initial investment.

Information Sources and Accuracy

While Stray Narratives strives to ensure the accuracy and reliability of the Content, no representation or warranty is made regarding its completeness, accuracy, or timeliness. Information may be obtained from third-party sources considered reliable, but their accuracy cannot be guaranteed.

All Content is provided “as is” and may contain inaccuracies, omissions, or outdated information. Stray Narratives reserves the right to correct such issues and update the Content without prior notice.

Limitation of Liability

To the fullest extent permitted by applicable law, Stray Narratives and the author shall not be liable for any direct, indirect, incidental, consequential, or punitive damages arising from or relating to the use of the Content, including but not limited to: investment losses or missed opportunities; inaccuracies, incompleteness, or errors in the Content; technical difficulties, delays, or interruptions; or any actions taken or not taken based on the Content.

Intellectual Property

All Content is protected by copyright and other intellectual property laws. No portion of any Stray Narratives publication or website may be extracted, reproduced, distributed, or modified without the express written permission of the author. Unauthorised use, reproduction, or redistribution of any Content, including communicating investment ideas mentioned in these publications to non-subscribers, is prohibited.

Personal Use Only

The Content is intended for your individual, non-commercial use only. You may not redistribute, resell, or commercially exploit the Content in any form.

Data Protection

Stray Narratives is published via Substack. For details on how personal data is collected, used, and safeguarded, please refer to Substack’s Privacy Policy.

Subscription

Stray Narratives reserves the right to cancel any subscription at any time, and if it does so it will promptly refund to the subscriber the amount related to the remaining subscription period.

Last updated: May 2026