원문 보기
He bet 9.2 million dollars against AI. Then he burned his own fund down. By the end of this, you will understand what Michael Burry actually wagered, why he closed a billion-dollar hedge fund to hold a single position, and what the architecture around that trade tells you about your own financial life, which is the part that matters whether or not he is right.
The media error that got the number wrong by a factor of a hundred, the valuations he is betting against, the capex figure that has since quadrupled, why being early has destroyed better investors than him, and the one structural thing he has that almost nobody watching this does. I'm Adam Feather, a dentist in the north of England, and I spent two years researching international tax residency and the machinery of wealth, partly to plan my own way out.
I'm not telling anyone to copy this trade. The maximum loss in a position like this is total. It's not a retail strategy, and the rest of this is analysis, not a playbook. None of this is financial advice. In November 2025, Burry's fund, Scion Asset Management, filed the disclosure that sent the tech market into a brief screaming panic.
The headlines were unanimous. Michael Burry bets 912 million dollars against Palantir. The big short investor wages nearly a billion against the AI darlings. CNBC, The Journal, The Financial Times, Bloomberg, all of them reported that he had staked the better part of a billion dollars against one of the AI boom's biggest winner.
Palantir dropped 8% in a day. The Nasdaq fell three. Palantir's chief executive, Alex Karp, went on television and called the bet super weird and memorably bat crazy. The two companies he is shorting, Karp protested, are the ones making all the money. The idea that chips and ontology is what you want to short is bat crazy.
Then Burry broke a near two-year silence with a single post that quietly corrected the entire financial press. He hadn't spent 912 million dollars. He had spent 9.2 million. The media had confused a notional value, the theoretical worth of all the underlying shares his options covered, with the actual cost of the contract.
They were off by roughly a factor of 100, which is an impressive error margin even for financial journalism. And the mechanics of that correction are the whole story, so they're worth getting right. A put option is the right, but not the obligation, to sell a share at a set price by a set date. Burry had bought puts on Palantir at $50 strike prices.
50,000 contracts, 100 shares each, at $1.84 a piece. That's $9.2 million in premium, and it's the most that he can lose. If Palantir sits above $50 when they expire, they're worthless and the $9 million is gone. If it falls below, they pay the difference and the further it falls, the more they pay. Capped downside, enormous upside, and the entire bet riding on one direction over roughly 2 years.
After Palantir's Q2 earnings report, its share price hit $162, so things aren't looking great for Burry, who has, according to rumor, rolled his puts at least twice already. At first disclosure, Burry's position wasn't a casual contrarian dabble. It was a conviction of a kind he has shown exactly once before.
His Q3 filing last year put the total reported book at around $1.38 billion notion. The Palantir puts were about $912 million of that, roughly 2/3. The Nvidia puts added another $187 million, about 13 and a half percent. Together, the two AI shorts were close to 80% of everything Scion held. The rest was scattered defensive scraps, a few calls, some old economy names, nothing in technology, nothing that hedged the thesis even slightly.
No protective longs, no collars, no offsetting spreads, pure unhedged directional conviction that the AI names would break before the contracts expired. We've seen this shape from him once before. During the 2007 housing crisis, his credit default swaps against mortgage bonds dominated his book the same way.
almost everything on a single reading of the arithmetic and endured nearly 2 years of losses while investors tried to pull their money and colleagues openly questioned his sanity and then made hundreds of millions when the thing collapsed precisely as he said that it would. Same conviction, same willingness to bleed in public while everyone calls him deranged.
And since that first filing, he has not sat still. He's rolled the Palantir puts out, so he now holds a $50 strike running to June 2027 alongside a strike expiring in December 2026. Although actually, I think he may have sold those just yesterday. He has added to the Nvidia puts, taken positions against Oracle, the semiconductor index, and the Nasdaq, and even rotated some money long into beaten-down software names.
The book is broader than it was. The direction hasn't changed an inch. Here is the move that gives this video its title, and it's the most important detail in the whole story. He wound down Scion entirely. He gave up the management fees, the performance fees, the entire apparatus of a working hedge fund, the 27th of October 2025, he wrote to his investors to tell them so.
The letter is short and polite. With a heavy heart, he wrote, he would liquidate the funds and return the capital by year end. And then the line that matters, "My estimation of value in securities is not now and has not been for some time in sync with the market." Read that twice. Not now and has not been for some time.
That's a man saying that the entire pricing system has detached from anything he recognizes as value, and that he is removing himself from the structure that forces him to keep explaining himself to the people who disagree. Because that is what winding down the fund actually buys him. No fund means no quarterly disclosure obligation, no investors able to yank their money at the worst possible moment, no need to justify a position that bleeds for 18 months before it finally pays.
He moved the whole thing into private capital and put the same positions on for himself, same strikes, same direction. And here is where your draft assumption, and frankly the assumption that most people make, turns out to be wrong. He didn't then go quiet, he started a Substack and has spent the months since as one of the loudest voices in the market, posting 10,000 word bear cases, confirming he is still holding, rolling the trade in public.
He removed himself from the machine that could force him out and kept the megaphone he chose to keep. That distinction between disclosure you are compelled to make and commentary you choose to make is the entire architecture of the trade. Burry isn't betting against AI as a technology. He's betting against valuations that price in perfection with no margin for a single error.
When he placed the bet, Palantir traded around $170, at a price to sales ratio north of 150. That means investors were paying more than $150 for every single dollar of revenue, not profit, not earnings, revenue, and the top line before a penny of cost. Set that beside Nvidia, which around the same time traded at roughly 30 times sales.
This is Nvidia, the company with the dominant chips, the actual monopoly characteristic, the hardware every AI deployment on earth runs on. Burry's own line on Nvidia, which is both accurate and funny, is that they are one of the two luckiest companies on the planet. They built a graphics chip. It happened to be right for crypto mining a decade ago and right for AI today, the correct product landing in front of them twice in a row, which is more luck than most companies get once.
Palantir, he says, is the other one. They built government software, put an AI cover on it when the wave arrived, and rebranded. His point isn't that these are bad businesses, it's that the market priced them as if they'd built the whole thing from first principles, when what they actually did was stand in the right place when it hit.
History is unkind to numbers like these. Before the dot-com crash, the biggest tech names peaked at 30 to 40 times sales, and that range has served as a ceiling throughout market history. Cisco hit 40 times sales in March 2000 and then fell roughly 90% and did not see that price again for about 17 years, while remaining profitable and dominant the entire time.
Amazon peaked nearly 30 times sales, fell about 95% and took 15 years just to get shareholders back to even, while becoming one of the great companies in history. Oracle hit 35 times sales, collapsed around 80% and needed a decade to recover. These were not failures, they were extraordinary businesses whose share prices had simply run so far past the fundamentals that even spectacular success could not rescue the shareholder.
The lesson was never that the businesses were bad, it was that the valuations were. Palantir at 150 times sales sits beyond anything a company that size has ever sustained, pricing in years of flawless growth, zero competition, zero stumble, every assumption correct at once. And on Burry's own podcast comparison, IBM does roughly the same work with a bigger data and analytics arm growing about as fast at a fraction of the multiple.
The gap between those two valuations is the entire bet. Underneath the valuations sits the thing Burry actually watches, which is the gap between what is being spent on AI and what is being earned from it. And the spending has gone somewhere your older notes will not have caught. The four biggest hyperscalers, Microsoft, Alphabet, Amazon, and Meta, spent roughly $410 billion on capital expenditure in 2025.
In 2026, they have guided around 700 billion, something like 77% in a single year. That money flows to Nvidia for chips, to the software firms, to the data center builders. The spending is real, enormous, and accelerating. The problem is the other side of the ledger. The returns are not arriving at anything like the same pace.
Amazon's free cash flow has fallen about 95% on a trailing basis to a little over a billion dollars, and it is forecast to turn negative this year. Because if you pour every spare dollar into data centers, there is no spare dollar left. Microsoft has admitted that 25 billion of its capex this year is simply component prices rising.
Oracle extended the assumed life of its servers from 6 years to 10, which on its own inflated reported earnings by cutting the depreciation charge. That's not AI generating returns, that is accounting making the spend look survivable. This is textbook late cycle behavior, and Burry's timing argument is more precise than something bad happens eventually.
He plots net investment, capex minus depreciation, against the size of the economy across previous bubbles, and the pattern is consistent. market peak almost always arrives before the capex cycle is even halfway through. Cisco grew revenue 55% in the year of the dot-com peak, and the stock topped that much anyway, while the spending ran on for another 18 months and shareholders lost 90%.
On his reading, we are somewhere mid-ramp, not at the end. The Oracle pattern is the tell. A company announces a few hundred billion of future bookings, and the market hands it $3 of value for every dollar merely promised. Larry Ellison was briefly the richest man on earth on the strength of an announcement on money not yet spent generating revenue not yet earned.
The market is pricing the press release, not the delivery. And that is exactly what the top of one of these things looks like. This is not the first time that Burry has placed a major short and looked spectacularly wrong before being proved right eventually. And the pattern is the point. In 2021, he bought Tesla puts, held them for months while the press mocked him.
And again, the headlines confused notional value with cost reporting that he'd bet 534 million against Tesla, off by two orders of magnitude in the now familiar house style. Tesla kept climbing, he looked catastrophically wrong, he closed the position in the October, and the press declared victory. Then Tesla peaked a month later and fell about 75% over the following year.
He was early and directionally right and made nothing because he couldn't psychologically endure watching the market prove him correct while the position bled. He did the same thing with Cathie Wood's ARK fund. Same mockery, same exit, same roughly 70% collapse arriving after he'd already gone. I could be wrong about this.
We don't know exactly when he sold, but this is my gut feeling. The pattern is ironclad. He spots a stretched valuation early, he bets, he holds through months of losses and ridicule, he gives up at the point of maximum pain, and then the market crashes 6 to 18 months after his exit, vindicating the thesis perfectly while he collects nothing.
This time, he has built a trade specifically so that that pattern cannot repeat. By winding down the fund and moving to private capital, he removes the one mechanism that has beaten him every previous time, the disclosure cycle and the investors who didn't want to be there. He cannot be forced out now.
He can simply wait. And that single structural change is the difference between this bet and everyone before it. So, where are we as this goes out? And the honest answer is messier than either side would like. Their thesis had a strong run in the spring. Palantir, which peaked at near $207, sold off hard into April, then came back up again after yesterday's Q2 earnings.
It trades about $165 now, down a little bit depending on which day you check. The wider software sector has fallen more than a third from its autumn peak, and it bottomed in April, and then it bounced hard. Software staged one of its strongest months in decades through May, and it's climbed most of the way back towards its high.
Oracle, one of Burry's targets, fell heavily into the spring and has since recovered to around $200, near where it started, and reports earnings as this goes out. So, that leg is not the clean win it briefly looked like it was going to be. Nvidia has largely traded sideways. Palantir, for its part, beat expectations again in Q1 and Q2 and raised its guidance to around $265 a share.
So, the scoreboard is not a victory lap, and it is not a defeat. It's a multi-year position with strikes running into late 2026 and the middle of 2027. In the middle of its life, Burry himself remains short, still holding the same two sets of Palantir puts, or at least one set of Palantir puts, still posting that he's not selling all of them, still calling the company a sandcastle held up by a narrative.
He may be proved correct, and he may be proved early, which for him has historically been the same thing, wearing a different coat. The thing to take from all of this, though, is not the scoreboard, it's the structure that let him survive long enough in order to find out. Here's the part that applies whether Burry is ultimately correct or not.
He can afford to be early and unpopular for 2 years because he engineered his life and his capital so that being early does not ruin him. Most working professionals cannot. They are locked inside structures, an employer pension, a single country tax exposure, savings in one currency, mobility on one passport, that quietly extract value every month regardless of whether they're right about anything.
The lesson isn't short Palantir, the lesson is structural. If your reading of where the financial system is heading is correct, about valuations, about pensions, about the arithmetic of an extractive tax state, you need your life arranged so that you can survive the gap between seeing it and being proved correct.
Otherwise, the cost of being right early is worse than the cost of being wrong. That's the real Burry lesson, and it's got nothing to do with options. He set his life up so that quarterly losses cannot force him to capitulate. He went private so reporting cannot drag him out of a position. He can hold something that bleeds for 18 months because his ability to eat doesn't depend on it paying off next quarter.
If he's right about Palantir, he makes a generational return once again. If he's wrong, he loses $9 million and carries on entirely unbothered. That asymmetry is what structure buys you, and that is precisely what most people watching this do not have, and exactly what the books and the framework on this channel are built to help you construct.
If you want to know which structural move is most urgent for your own situation, your income, work, family timeline. Honestly, rather than flatteringly. The nine free books are at the same link, and they go far deeper into the structural side of all of this than fits in a video. The Sovereign Stack waiting list is there, too, with founder pricing for everyone on the list when it opens.
Burry has placed his chips and arranged his life so that being early cannot break it. The exits are still open, some of them. Use them while you can. video.
Comments 0