Thomas Hayes — Called The Tech Wipeout, Now Reveals The Next Great Rotation (David Lin, Aug 2026)

Ticker	Name	Research	View	What he said	At
DEO	Diageo	QT · SA · STK · FA	Positive	His defensive-consumer pick and an AI beneficiary, not a victim: down 50% from 2021 highs yet still generating $3B free cash flow, 13.4% ROIC in FY2026, growing double digits everywhere but the US, where a tequila problem is being fixed by "drastic Dave" Lewis (ex-Tesco/Unilever). "It's trading like it's going out of business" — good business, great price, with a large margin of safety.	46:40
DIS	The Walt Disney Company	QT · SA · STK · FA	Positive	His offensive-consumer pick: new CEO Josh D'Amaro comes from the experiences business, which has had the highest ROIC for years and is where capital is now going (new cruise ships, park investment). Streaming has flipped from drag to FCF-positive and growing, ESPN is being re-valued, box office is crushing it; Euro Disney was packed. "This happens every 15, 20 years — the stock sells off 50%, everyone says it's over, then they realize they have the greatest IP library in the history of mankind."	51:11
XRAY	DENTSPLY Sirona	QT · SA · STK · FA	Positive	A defensive healthcare turnaround he owns ("one company we own"): largest dental-supply business since the late 1800s, a GDP-growing business whose stock was hit by bad management. Its tell — it used a good slug of its tariff refund to buy back stock "in the hole," which he rates far above management talk ("watch what they do").	27:49
TSM	Taiwan Semiconductor Manufacturing	QT · SA · STK · FA	Neutral	Cited as the bull case he rejects: TSMC saying demand is "insatiable" is exactly the sort of datapoint retail is chasing — "the answer is because that's known. The market is a discounting mechanism." Semis and memory remain the most crowded trade in the world.	12:19
MU	Micron Technology	QT · SA · STK · FA	Neutral	Named with TSMC/CoreWeave/Nebius as the "known" good news powering the bounce. Memory is where the crowding is worst, and he thinks semis/memory weaken again in coming months — none of them has made a new high on this rip.	12:19
CRWV	CoreWeave	QT · SA · STK · FA	Neutral	Its strong earnings are the standard objection to his sell call — and his answer is that a discounting market has already priced it: "that's known." Cited as evidence of the recency bias driving retail back in, not as a view on the company.	12:19
NBIS	Nebius Group	QT · SA · STK · FA	Neutral	Same bucket as CoreWeave: a strong AI-infrastructure print being used to justify chasing the bounce, which he dismisses as already discounted.	12:19
APO	Apollo Global Management	QT · SA · STK · FA	Neutral	One of the six firms in Nvidia's $500B third-party compute-financing consortium. Hayes' read: they get paid either way — "we're just going to collect fees brokering the securization and laying it off to institutions and retail investors" — while the credit risk lands on insurers and retail buyers of the paper.	33:56
BLK	BlackRock	QT · SA · STK · FA	Neutral	Named in the $500B Nvidia compute-financing consortium. That "the five guys at the table" flew in from the Hamptons in mid-August tells him how urgent the financing gap is — the AI thesis "has now run out of financing."	33:56
BX	Blackstone	QT · SA · STK · FA	Neutral	Consortium member in the $500B AI-infrastructure financing platform. Same framing: fee income now, with the securitized GPU collateral risk distributed to end buyers.	33:56
BN	Brookfield Corporation	QT · SA · STK · FA	Neutral	Named among the long-term capital providers underwriting the $500B of AI infrastructure financing announced by Nvidia two days before the interview.	33:56
GS	Goldman Sachs	QT · SA · STK · FA	Neutral	Consortium member. His analogy is 2007-08 mortgage securitization: the arrangers "won't get hurt" because they earn fees for structuring and distributing the paper — the loss sits with whoever holds it when the collateral turns out to be obsolete chips.	33:56
KKR	KKR & Co.	QT · SA · STK · FA	Neutral	Named in the $500B Nvidia compute-financing consortium — one of the arrangers whose success depends on getting rating agencies to bless securitized GPU exposure as investment grade.	33:56
CMCSA	Comcast (Universal Parks)	QT · SA · STK · FA	Neutral	Cited only as the read-across everyone got wrong: "everyone thought Disney was going to do poorly on earnings cuz Universal's parks were crap. But guess what? Universal was crap. Disney was up."	51:35
UL	Unilever	QT · SA · STK · FA	Neutral	Named only as part of the Diageo jockey's pedigree — Dave Lewis' cost-discipline and brand-management track record came from Unilever and Tesco before he took over Diageo.	47:10
TSCO.L	Tesco	STK	Neutral	Named only as where "drastic Dave" Lewis ran his best-known turnaround before Diageo — the evidence behind Hayes' "new jockey, which is critical in a turnaround" test.	47:10
NVDA	Nvidia	QT · SA · STK · FA	Negative	Negative on the financing scheme, not the chips: the $500B third-party compute platform only works if the rating agencies bless securitized GPUs as investment grade, which is why Jensen must argue a >5-year useful life. "You're going to take a semiconductor chip that is 5 years outdated… if you don't get paid back, you're basically not secured by anything useful." A mañana problem people only figure out in 3-5 years.	36:28
GOOGL	Alphabet	QT · SA · STK · FA	Negative	Its $85B upsized equity raise — the largest in US corporate history — to fund AI infrastructure is his supply-overwhelms-demand exhibit: "the ducks are quacking so they're going to feed them." Asked whether the $85B is productive: "the results in the short term have shown you that it's a bad use of capital." Record equity issuance eventually overwhelms demand.	40:41
INTC	Intel	QT · SA · STK · FA	Negative	Cited as the dilution datapoint: with hyperscaler free cash flow exhausted, the AI complex is raising debt and equity — "Intel diluted their shareholders by 20 some odd percent." Part of the record-issuance wave he expects to weigh on the whole trade.	15:18
OpenAI	OpenAI (private)	—	Negative	Half of the single point of failure: "a lot of this AI trade is predicated on the two frontier models succeeding, OpenAI and Anthropic" — yet there are $1.5 trillion of unfunded commitments tied to the two, and "those companies don't have the income to fund the commitments of capex that they've put out there which the entire finance industry is relying upon." Open-weight models at a tenth to a hundredth of the cost undercut them.	31:07
Anthropic	Anthropic (private)	—	Negative	The other frontier lab the trade depends on. Enterprises shocked by their token bills are moving to open-weight models at a tenth or a hundredth of the cost — which undermines the revenue assumed behind the $1.5T of unfunded commitments the finance industry is relying on.	31:07

1:23 CPI cools — and the hike chorus is fading

Same-day CPI: 3.4% headline / 2.5% core, both cooler; futures up ~20bp S&P, ~50bp NASDAQ, gold +1%, yields slightly down. Markets price 62% hold / 38% hike, no cut odds at all.
In June the consensus was hikes. Jobs and inflation now say "the chorus of hikes… is now diminishing, and that usually precedes the chorus of cuts accelerating." A hike before year-end is "dramatically diminished."

3:00 A cut before a hike becomes material

The dual mandate points the same way: tame inflation despite elevated energy, weakening jobs. Into fall/pre-election seasonality, "the possibility of cuts will be back on the table" — no one sees it on August 12th, just as no one saw the no-hike outcome in June.
Whether the cut lands before the calendar turns or in Q1 is "immaterial" — the direction is what matters.

4:50 The long end, Bessent, and Warsh's liquidity drain

If the 30-year gets out of control, expect Bessent to get involved — an Operation Twist is on the option list that's been trailed in recent articles.
Warsh's ability to not-hike (or cut) is predicated on draining liquidity from the balance sheet first; that hasn't begun in earnest yet. Each increment of drain buys him room to cut — and the incentive is to cut to refinance the debt and spur housing (~18% of GDP).

7:06 "This is all entertaining talk" — the actual process

Rate forecasting is entertainment: "we just like to buy good businesses at great prices. When the price is dislocated, they're out of favor — and sell them when everyone wants them."
The mandate: doubles and triples over a few years. Getting rates or currencies right "doesn't have a dramatic impact on how we make money."

9:00 Yen intervention = indirect yield control

Lin's framing (which Hayes endorses 100%): buying the yen stops the Japanese from dumping Treasuries, so it's yield control by another route — this administration does not want higher rates.
Play a weaker dollar and stronger yen over the coming couple of years, "despite the short-term counter trend move." The carry trade is already being mitigated (not yet unwound) — part of why the "Mag 7 has become the lag 7."

11:03 Canaries in the coal mine: leverage and July's hedge-fund carnage

AI-related exposure inside leveraged ETFs has jumped from 26% to 58% in a couple of years.
One month of losses at "not amateur hour" funds: Whale Rock −21.7%, Lone Pine −27%, Altimeter −11%, Situational Awareness −67%, Vera −44% — in July alone, and those are just the canaries.

12:19 "That's known" — why good semi earnings don't save it

CoreWeave, Nebius, TSMC's "insatiable demand," Micron — all already discounted. Revenue-growth-and-earnings expectations haven't been this high since Q4 2021.
The 2022 rhyme: right after peak expectations came a tech cooling — VC blew up, ARK blew up, the sector corrected.

13:37 Anatomy of a dead-cat bounce

Across ~50 charts he sent the host, none of the semis has made a new high on this rip — "it's a deadcat bounce," running to maybe the third or fourth week of August on normal seasonality.
Retail is chasing a 20-30% short-covering move in names that fell 30-60% in July, with one-day options and calls creating a little gamma squeeze. Waterfalls like 2000 and 2022 persist "for multiple weeks, if not multiple months" and dead-cat bounces get sold the whole way.

17:07 What the June rotation call looked like in advance

Positioning first: retail call buying and concentrated call buying in semiconductors was at record all-time highs — the most crowded trade in the world, and still is.
Euphoria plus record crowding is the same pattern he sees on this bounce, "which is also going to wind up punishing."

18:07 The redemption chain — why the selling is mechanical

"First thing that they got on August 1st was redemption requests." Institutions must create liquidity in coming months — and unlike retail they don't sell in the hole; they wait for the dead-cat bounce to sell into.
Difference vs June: then, dealer hedging of retail call buying created a parabolic crescendo. Now the smart money is using retail's exuberance to get out and hold cash for lower levels in the fall.

20:50 The rotation: a consumer barbell

Get out of semis, into defensives — but the twist is a barbell: consumer staples (defense) and consumer discretionary (offense), because consumer confidence is bouncing off multi-decade and in some cases all-time lows.
"You never want to bet against the US consumer when they have a job" — 4% unemployment, slowing but positive job growth, energy in a $70-90 range.

21:59 25-year and 10-year relative lows — "backing up the truck"

Defensives relative to the S&P are at roughly a 25-year generational extreme; consumer discretionary relative to the S&P is at about a 10-year low. Both groups are washed out.
He also holds healthcare for the same reason — a Bloomberg piece the day before framed "long healthcare is the new short AI." Utilities are excluded: they got caught in the AI-trade backdraft.

25:53 Why the consumer is turning: insurance and rents roll over

The two stickiest post-COVID inflation components — insurance premiums and owners' equivalent rent — are softening "against all consensus that it ever would."
Consumer sentiment tracks forward inflation expectations; the oil plunge on a perceived deal taught the consumer there's a limit ("the taco trade"), and 5-year breakevens stayed contained despite Iran. Back-to-school should beat because "the bar is so low."

27:06 Tariff refunds — watch what they do, not what they say

Turnarounds normally hoard capital for the turnaround, so a turnaround buying back stock is a loud signal. DENTSPLY Sirona used a good slug of its refund "for buying back stock in the hole."
The mirror image in AI: more insider selling than buying, and buybacks collapsing at companies that were historically the premier buyers of their own stock — they're free-cash-flow negative and can't afford it until late 2027/early 2028.

28:53 Productivity growth is the only AI-ROI indicator that counts

"Everything else other than an increase in the productivity rate is noise." If monthly productivity isn't rising, there's no return on the invested capital yet — and there are no signs of it.
He's not bearish on AI — mid-innings, and he thinks the answer is yes in another 18-24 months. But if productivity doesn't come, "it's all for not," and you get a 2000 outcome: consumers get the benefit, the payers (Global Crossing, MCI WorldCom) don't.

30:14 The token bill, open weights, and $1.5T of unfunded commitments

The free-then-charge playbook ("like every good drug dealer") ran into enterprise sticker shock on token expense — hence the rush to open-weight models at a tenth or a hundredth of the cost.
The trade is predicated on OpenAI and Anthropic succeeding, but there are $1.5T of unfunded commitments tied to two companies "that don't have the income to fund the commitments of capex… which the entire finance industry is relying upon."

33:56 Nvidia's $500B third-party compute-financing consortium

Announced two days earlier: independent compute-financing platforms with Apollo, BlackRock, Blackstone, Brookfield, Goldman Sachs and KKR to mobilize $500B+ of third-party capital, on Jensen's claim that "AI compute is revenue" — fungible, transferable, CUDA-improved.
Hayes' read on the timing: this is the AI thesis running out of financing and inventing a new source.

35:08 Securitization 101 — and what exactly do you repossess?

"Back to the future": music royalties, mortgages, autos, credit cards. The 2007-08 lesson is that the collateral has to be worth more than the loan when you take it back.
Here, the collateral is a five-year-old, heavily used GPU that may be obsolete in three years. "If you don't get paid back, you're basically not secured by anything useful." Jensen must argue >5-year useful life because the whole $500B depends on investment-grade ratings so it can be sold to insurance companies and retail.

38:13 The Hamptons-in-August tell and ratings-agency capture

Five principals leaving the Hamptons in mid-August for this meeting tells you how urgent it is.
The arrangers are structurally safe — "we're just going to collect fees brokering the securization and laying it off to institutions and retail investors" — provided the rating agencies play along, and their incentive is "insane amounts of fees by rating this quote unquote new asset class." Reluctant at first after the housing crisis, then pressured by fees. "This conversation clip will be very useful to refer back to 3 to 5 years from now."

39:35 Google's $85B — the largest raise in US corporate history

Upsized from $80B to $85B to fund AI infrastructure and the global compute buildout. "The ducks are quacking so they're going to feed them."
Asked how to judge whether $85B was productively spent: "the results in the short term have shown you that it's a bad use of capital." At some point, supply of equity overwhelms demand.

40:41 1995-2000: three 50% drawdowns on the way to being right

The SOX index had three separate 50% drawdowns between 1995 and 2000, even though the internet delivered everything promised times ten.
The error was paying in 2000 as if it had already been delivered — the NASDAQ then fell 80-90%. Today's buyers are again "paying as if the returns will definitely come."

42:37 Despondency is the buy signal

There will be a moment — weeks, months or a couple of years out — "where people will be despondent about the AI trade and we will only be halfway through." That's the drawdown to buy.
He expects the flush into the election: calls for S&P 8,000-9,000 get retail "back into the boat," then the trapdoor. That's where he'd be aggressively buying to build the long-term list.

45:20 Software goes in the "too hard" box

His day job now: find businesses that are AI beneficiaries, not businesses AI disintermediates. Software's carnage looks like generational value on price-to-free-cash-flow.
But he can't be certain it isn't fully disintermediated in five years — "I'm an investor, not a psychic" — so despite wanting to buy, capital-impairment risk keeps it in the too-hard box.

46:40 Diageo — defensive consumer, AI-proof

Spirits and Guinness don't get disintermediated "unless AI can figure out a way to help you forget all of your troubles that's better than having a cocktail with a friend"; if anything AI helps margins.
The setup: down 50% from 2021 highs (COVID pull-forward then the GLP-1 scare, though 85% of users are off within two years), still $3B FCF, 13.4% ROIC FY2026, double-digit growth everywhere but the US tequila drag — being fixed by "drastic Dave" Lewis. People are drinking less but drinking better, and Diageo is 1.4× the revenue of its nearest premium competitor.

50:27 Disney — offensive consumer, capital going to the best segment

Euro Disney packed, a whole new Frozen land in Paris while US parks add Avatar; the cruise business is "a monster."
New CEO Josh D'Amaro comes from experiences — the highest-ROIC segment — and that's where the money goes now. Streaming flipped to FCF-positive and growing, ESPN is being re-rated, box office is strong. Universal's weak parks made everyone expect a bad Disney print; Disney was up.

52:50 The wrap: price is what you pay, value is what you get

Be in consumer discretionary for offense, consumer staples and healthcare for defense — "when the fits and starts come and the AI trade backs up, these groups are going to do just fine and in many cases exceptional."
Great Hill runs money for ultra-high-net-worth clients and family offices; the free weekly blog at hedgefundtips.com covers two holdings in depth, plus the "Hedge Fund Tips with Tom Hayes" podcast. The mandate is unchanged: good businesses and turnarounds that can double or triple in 3-5 years.

Diageo owns the world's biggest collection of premium liquor brands plus Guinness. Hayes buys it for the simplest reason in his book: AI can't replace it. "It's not going to get disintermediated unless AI can figure out a way to help you forget all of your troubles better than having a cocktail with a friend." If anything, AI helps it — cheaper marketing and back-office work means fatter margins.

The stock is cut in half from its 2021 peak for two reasons that are both fading. First, COVID pulled drinking forward (people were paid to sit home and drink), so the comparison years were fake-good. Second, investors panicked that weight-loss drugs would kill alcohol demand — but Hayes notes 85% of people quit those drugs within two years. Meanwhile the business still throws off $3 billion of cash a year, earns 13.4% on the capital it employs (FY2026), and grows double digits in Africa, Europe and South America. The one real problem is the US tequila business, which the new CEO — Dave Lewis, nicknamed "drastic Dave" for the turnarounds he ran at Tesco and Unilever — is fixing.

His summary: it's "trading like it's going out of business" while generating billions, so the downside is protected ("a large enough margin of safety") even if people permanently drink less — because they're drinking better, and Diageo is the premium leader at 1.4× the revenue of its next competitor. When the AI bounce rolls over and money looks for somewhere to hide, he expects Diageo to take its next leg higher.

Disney is his "offensive" consumer bet — the one that wins if the American consumer keeps spending rather than the one that survives if they don't. The core of the argument is where the company now puts its money. New CEO Josh D'Amaro came up through the "experiences" side (parks, cruises, resorts), which has earned the highest return on invested capital in the company for years — meaning every dollar spent there produces more profit than a dollar spent anywhere else. Now that's the segment getting the capital: new cruise ships, new park investment.

The rest of the business has quietly flipped from problem to asset. Streaming, which burned enormous cash for years, is now free-cash-flow positive and growing. ESPN, which everyone had written off, is being re-valued as live sports rights get scarce. The box office is doing well. He watched Euro Disney in France come off a redeye and found it packed, with a brand-new Frozen land — while US parks are still selling Space Mountain.

His pattern recognition: "this happens every 15, 20 years — the stock sells off 50%, everyone says it's over, and then they realize they have the greatest IP library in the history of mankind" and find a new way to monetize it (theaters to VHS in the 80s, DVD to streaming now). One sanity check he liked: everyone assumed Disney's earnings would be bad because Universal's parks were weak, and Disney was up anyway.

DENTSPLY Sirona sells the equipment and supplies dentists use — it has been the largest supplier in that market since the late 1800s. That makes it a boring, defensive, roughly GDP-growth business inside healthcare, exactly the sort of thing Hayes wants while the AI trade unwinds. It's a "turnaround" because the stock was beaten down by bad management, and he owns it.

The reason he singles it out is a behavioral tell rather than a forecast. Companies in the middle of a turnaround almost never buy back their own stock — they need every dollar to fix the business. So when DENTSPLY took a chunk of its tariff refund and used it to repurchase shares while the price was still "in the hole," that was management putting real money behind a belief the stock was too cheap. His rule: "watch what people say, but watch what they do is more important." He contrasts it directly with AI companies, where insiders are selling and buybacks have dried up because the cash is gone.

TSMC makes most of the world's advanced chips, and it told the market demand is "insatiable." Hayes' point is not that this is untrue — it's that everyone already knows it. Markets price in what's known; the stock only moves on what's new. So when people ask how he can sell a rally backed by great chip earnings, his answer is "because that's known."

He adds a warning from history: earnings and revenue-growth expectations across the sector are the highest since late 2021, and 2022 is what happened immediately after the last time expectations peaked. Semiconductors and memory remain the most crowded trade in the world — which is why he expects them to weaken again over coming months rather than break out.

Micron makes memory chips, and it's in the same bucket as TSMC here: the good news is real but already reflected in the price. Hayes uses it as an example of the "recency bias" pulling retail investors back in — they see a strong print and assume the rally continues.

What he's watching instead is positioning: memory is where the crowding is most extreme, and across the roughly 50 semiconductor charts he sent the host, not one has made a new high on this bounce. A rally that can't make new highs after that much good news is a rally being sold into, not a new uptrend.

His problem isn't Nvidia's chips — it's the new way the buildout is being paid for. Nvidia just announced a plan with six giant investment firms to raise over $500 billion from outside investors to finance AI data centers. In plain terms: instead of customers buying chips with their own money, Wall Street will bundle the loans against those chips into bonds and sell them to insurance companies and ordinary investors. That's "securitization," the same machinery used for mortgages before 2008.

The test of any secured loan is simple: if the borrower stops paying, what do you get to take back? With a house, you get a house — it's still standing in fifteen years and probably worth something. Here, you get a five-year-old graphics chip that's been run hard and may be obsolete. "If you don't get paid back, you're basically not secured by anything useful." That's why Jensen Huang keeps publicly arguing these chips last more than five years — the entire $500 billion depends on convincing credit-rating agencies to stamp the bonds investment grade so they can be sold widely, and the rating agencies are paid handsomely to agree.

Hayes doesn't think this blows up soon — "people aren't going to figure this out until 5 years from now." He was explicit that the clip would be useless three months out and very useful in three to five years. He also reads the urgency as a signal: five senior principals left the Hamptons in mid-August for this meeting, which tells you the old sources of AI financing have run dry.

Alphabet just sold $85 billion of new stock — upsized from $80 billion, the largest equity raise in US corporate history — to fund AI data centers. Selling new shares means every existing shareholder owns a slightly smaller piece of the company, so it only makes sense if the money earns a good return. Asked what he'd look for in the filings to judge that, Hayes cut it short: "the results in the short term have shown you that it's a bad use of capital."

His broader worry is arithmetic. The big tech companies used to be the largest buyers of their own shares, which shrank the supply and pushed prices up. Now they've burned through their spare cash and are doing the opposite — issuing record amounts of stock and debt (Intel diluted holders about 20%). "The ducks are quacking so they're going to feed them." At some point that flood of new paper overwhelms the money available to buy it, and prices fall for reasons that have nothing to do with how good the businesses are.

Intel is his shorthand for how expensive the AI buildout has become for existing shareholders: it "diluted their shareholders by 20 some odd percent." Dilution means the company issued so many new shares that anyone who already owned it now owns roughly a fifth less of the same company without doing anything wrong.

He isn't analysing Intel's business here — he's counting it as one more entry in a record wave of equity issuance across the AI complex. When the companies that used to shrink their share count start expanding it aggressively, the supply of stock grows faster than the demand, and that weighs on the whole group.

Hayes' concern is concentration risk. "A lot of this AI trade is predicated on the two frontier models succeeding" — OpenAI and Anthropic — and roughly $1.5 trillion of spending commitments across the industry are tied to those two companies continuing to buy. But neither generates anywhere near the income needed to honour those commitments; they're funded by expectations. "Those companies don't have the income to fund the commitments of capex that they've put out there, which the entire finance industry is relying upon."

What makes it fragile is competition from below. Large companies got their first real AI bills and were shocked, which pushed them toward "open weight" models — free or near-free AI systems anyone can run — at a tenth or even a hundredth of the cost. If good-enough AI is that much cheaper, the revenue that was supposed to back $1.5 trillion of commitments never shows up.

Anthropic is the other half of the pair the AI trade rests on. The same logic applies: an enormous amount of data-center spending has been committed on the assumption that Anthropic and OpenAI will keep paying for compute at premium prices, yet enterprises shocked by their token bills are actively shopping for cheaper open-weight alternatives.

Hayes thinks the shift to cheaper models eventually improves returns for the companies using AI — but it does so by taking revenue away from the two frontier labs whose success the financing structure assumes. Two companies carrying the credit weight of the entire buildout is the single point of failure he flags.

Thomas Hayes — Called The Tech Wipeout, Now Reveals The Next Great Rotation

1. Stocks & names mentioned

2. Talking points