원문 보기
In November 2025, the man who predicted the 2008 housing crash placed a bet against the two biggest AI stocks on the planet. Then, he shut down his entire hedge fund. The media called it a billion-dollar short. Headlines screamed that Michael Burry had wagered 912 million dollars against Palantir alone.
But here's what nobody told you. He didn't spend 912 million. He spent 9.2 million in option premium. A small, calculated wager on a catastrophic collapse. And as of right now, that bet is working. Palantir is down more than 30% from its peak. Oracle, another stock he shorted, has collapsed over 55% and just days ago, Burry posted something that sent Palantir tumbling another 6% in a single session.
Today, I'm going to break down exactly what Burry sees, why he believes the AI bubble is, in his own words, "too big to save," and what this means for your money. Because if he's right, what comes next could reshape the portfolios of millions of regular investors. And if he's wrong, you still need to understand why.
Let's get into it. Let me start with what actually happened, because the media got this story completely wrong. On November 3rd, 2025, Scion Asset Management filed its final 13F with the SEC. That's the mandatory disclosure hedge funds have to make every quarter showing their holdings. And what it revealed stopped people in their tracks.
The filing showed five million Palantir put options with a notional exposure of 912 million dollars and 1 million Nvidia put options worth another 186 million. Together, these positions made up roughly 80% of Scion's entire disclosed portfolio by notional value. Every financial outlet ran with the same headline.
"Burry bets a billion dollars against AI." But here's the thing. That 912 million number is the notional value, meaning it's calculated by multiplying the number of shares those options control by the current market price. It's not what Burry actually risked. 10 days later, Burry himself went on X and corrected the record.
He wrote, "I spent 9,200,000 dollars, not 912 million," tagging CNBC, The Wall Street Journal, and The Financial Times. So what did he actually buy? 50,000 put option contracts at $1.84 each. Each contract controls 100 shares. That gives him the right to sell Palantir at $50 per share before January 2027.
At the time, Palantir was trading above 180. So this is a deeply out of the money bet. His maximum loss is that 9.2 million in premium. But if Palantir falls to zero, the theoretical maximum payoff is 250 million dollars. That's a 27:1 risk-reward ratio. This is not a reckless gamble. This is a precisely structured asymmetric trade.
The exact same kind of structure he used against the housing market in 2007. One week after that filing dropped, Burry did something that made this story 10 times more interesting. He shut down Scion Asset Management entirely. The SEC database listed the fund's registration status as terminated on November 10th, 2025.
In his letter to investors dated October 27th, he wrote, "My assessment of value in securities is currently not aligned with the markets, nor has it been for quite a while." Now, think about what this means. In 2008, Burry bet against the housing market through credit default swaps on mortgage-backed securities.
When that trade played out, he earned 100 million dollars personally and 700 million for his investors. And then, he dissolved Scion Capital. The identical pattern, massive short, then fund closure. He placed the bet, he walked away, and now he waits. By deregistering, Burry no longer has to file 13F disclosures.
His current portfolio is invisible to the public. We have no idea whether he's maintained, expanded, or closed any of his positions since November. The last window we had into his moves is sealed shut. So why is Burry so convinced that AI is a bubble? It comes down to one word, math. Not opinions, not feelings, the actual numbers behind the AI infrastructure buildout.
Here's the picture. The four biggest cloud companies, Alphabet, Amazon, Meta, and Microsoft, have committed a combined 635 to 665 billion dollars in capital expenditure for 2026 alone. That's a 67 to 74% increase from 2025. Amazon alone is spending roughly 200 billion. Microsoft, around 145 billion.
These are not small bets. These are the largest capital expenditure commitments in the history of corporate America. Now, here's the question Burry is asking. Where is the revenue to justify this? Because right now, the numbers don't add up. OpenAI, the company at the center of this entire revolution, generated 13.1 billion in revenue in 2025.
Sounds impressive, but it burned through 8 billion in the same period. For every dollar of revenue, it was spending $1.69. And its revised projections show cumulative cash burn through 2030 reaching 665 billion dollars with profitability not expected until 2029 at the earliest. Burry's former associate, Phil Clifton, who he described as the most prodigious thinker he's ever met, built the analytical backbone of this thesis.
Clifton's central argument is blunt. The investment community is anticipating far more economic significance from this technology than is likely to materialize. He isn't saying AI is useless. That's an important distinction. He's saying the financial returns on AI infrastructure investment are insufficient relative to the capital being deployed.
Think about it this way. A technology can change the world and still be a terrible investment. The internet changed everything, but most internet companies went to zero. The railroad transformed America, but most railroad investors lost their shirts. Clifton put it perfectly. "Just because a technology is good for society doesn't mean it's a good business proposition." Now, let's talk about what happened just days ago, because this is the freshest piece of the puzzle.
On April 7th, 2026, Burry posted on X that Anthropic is eating Palantir's lunch. He cited data from financial operations platform Ramp showing that Anthropic surged from 9 billion to 30 billion in annualized revenue in a matter of months. It took Palantir 20 years to reach 5 billion in total revenue.
And the data backing this up is genuinely striking. According to the Ramp AI Index from March 2026, Anthropic now captures 73% of all new enterprise AI spending among first-time buyers. Nearly one in four businesses on Ramp now pays for Anthropic. A year ago, it was one in 25. That's an explosion in market share that is coming directly at the expense of companies like Palantir.
Burry's argument is straightforward. Palantir's enterprise AI platform, the thing Wall Street is betting will drive its growth for the next decade, is being undercut by a faster, cheaper, more intuitive alternative. And it's not some garage startup. Anthropic is currently valued at 380 billion dollars post-money with over 300,000 business customers.
Palantir fell 6.2% the day after Burry's post. Now, here's where it gets interesting. Palantir's bulls will tell you the government business is the real moat. But Burry dismisses that, too, calling government contracts low-margin and small. And there's a new headwind nobody expected. Defense Secretary Pete Hegseth has directed an 8% annual Pentagon budget cut for the next five years.
Palantir derives over half its revenue from government contracts. That's not a tailwind. That's a direct threat to their largest revenue stream. I want to tell you a story that I think puts everything into perspective, because history has given us an almost perfect mirror for what's happening right now.
In March 2000, at the height of the dot-com bubble, a company called Cisco Systems became the most valuable publicly traded company on Earth. Cisco made the routers and switches that powered the internet. It was the picks and shovels play of the internet era, the AI equivalent of Nvidia today. Investors were convinced that as long as the internet kept growing, Cisco would keep printing money.
Then, the bubble burst. Cisco's stock collapsed by more than 80%. But here's the part that should genuinely concern every investor holding AI stocks right now. It took Cisco 25 years to reclaim its dot-com all-time high. 25 years. The stock finally passed its March 2000 peak in December 2025. A quarter century of waiting to break even.
And that's the price return. Adjusted for inflation, investors who bought at the top still haven't made their money back. Cisco didn't go bankrupt. It stayed profitable. It kept generating revenue. It made smart acquisitions. It grew earnings per share eightfold. But none of that mattered, because the price investors paid at the peak assumed perfection that reality could never deliver.
Now, consider this. Nvidia has already surged roughly 2,000% from its October 2022 lows. Its market cap briefly exceeded 5 trillion dollars, 10 times larger than Cisco's peak dot-com valuation. Palantir trades at a price-to-earnings ratio of 184 and a price-to-sales ratio above 80. To justify its peak valuation, analysts calculated that Palantir would need to sustain 35% annual revenue growth for 25 consecutive years.
Think about that. A quarter century of near-perfect execution with zero slowdown. That's not a valuation. That's a prayer. In March 2026, Burry published what he called his coiled spring warning. And this is where his thesis goes from being about individual stocks to being about the entire market. He argued that several forces are converging at once to create what he calls a market coiled spring primed for violent dislocation.
The first force is passive investing dominance. More than 60% of equity fund assets now sit in passive index funds. That sounds safe, but it means trillions of dollars are being invested with no regard for whether individual stocks are overpriced or underpriced. Money flows in automatically and buys everything.
This inflates prices across the board. But when flows reverse, there's no one making individual buy decisions. Everyone sells everything at once. The second force is demographic. Baby boomers are hitting required minimum distribution age, meaning they're legally required to start pulling money out of their retirement accounts.
That's structural selling pressure that didn't exist 5 years ago. The third force is the collapse of corporate buybacks. For years, big tech companies used their enormous cash flows to buy back their own stock, which supports prices. But now that cash is being redirected into AI capital expenditure, $635 billion this year alone.
That's money that used to prop up stock prices and is now being poured into data centers. Burry ran the math on valuations. The Shiller CAPE ratio, a measure of stock market valuation adjusted for inflation and earning cycles, hit 40.22 in January 2026. That was only the second time in 155 years it had crossed above 40.
The first time was December 1999, right before the dot-com crash. To return to the post-1990 average, the S&P 500 would need to fall 32% to around 4,700. To reach long-term historical averages, it would need to fall 52%. Those aren't Burry's opinions, that's arithmetic. There's one more data point I need to show you because I think it tells you more about what's really happening than any headline or earnings report ever could.
Over the past 2 years, insiders at Nvidia and Palantir, the two companies at the very center of the AI boom, have sold a combined $9.65 billion in stock. That's not a typo. Nearly $10 billion in insider selling across the two most hyped AI stocks on Earth. Nvidia insiders sold $3.98 billion. Palantir insiders sold $5.67 billion, including a $290 million sale by co-founder Peter Thiel in March 2026.
At Nvidia, CEO Jensen Huang alone sold $2.9 billion in shares. And across both companies, the insider buying total during the same period was effectively zero. 15 Nvidia insiders sold, not one bought. That has been described as one of the most lopsided sell-to-buy ratios in the S&P 500. Meanwhile, Palantir CEO Alex Karp sold approximately $1.88 billion in shares since early 2024, while simultaneously going on CNBC and calling short sellers batshit crazy.
Think about that for a second. The man running the company is telling you the stock is a great investment while he's personally cashing out at the fastest rate in the company's history. That's not necessarily proof that the stock will crash, but it's a signal you cannot afford to ignore. Now, here's where I need to be honest with you because this story has two sides and you deserve both of them.
Burry has been wrong before, a lot. Since his legendary 2008 trade, he has made at least 12 major bearish predictions that didn't play out. He shorted Tesla in 2020 and it doubled. He predicted the mother of all crashes in February 2021 and the market kept climbing. In August 2023, he placed a $1.6 billion notional bet against the entire market and later admitted he was wrong.
As one analyst put it, the calls Burry has made since 2008 have not been good. And the bull case for these companies is not imaginary. Palantir reported Q4 2025 revenue of $1.41 billion, up 70% year over year. Its US commercial revenue surged 137%. It's guiding for $7.2 billion in 2026 revenue. These are genuinely impressive numbers.
Nvidia reported $68.3 billion in revenue in a single quarter, beating Wall Street estimates comfortably. Unlike the dot-com bubble, today's AI companies are generating real revenue, real profits, and real cash flow. The companies funding the AI buildout, Google, Amazon, Meta, Microsoft, are among the most profitable businesses ever created.
They're not pre-revenue startups burning venture capital. That's a meaningful difference. And Burry's actual position is more fragile than people realize. His Palantir puts have a strike price of $50 and expire in January 2027. With the stock currently around 140, those puts are roughly 65% out of the money with about 9 months left.
For them to expire profitably, Palantir would need to fall another 65% from where it is today. That's a massive move in a short window. The puts may have gained value as the stock dropped from 207, which is where the printing money narrative comes from, but they're still technically out of the money relative to the strike.
So, what does all of this mean for a regular investor? Let me be very clear about something. I'm not telling you Burry is right and I'm not telling you he's wrong. What I am telling you is that the data he's pointing to is real and you need to at least understand it before you make decisions with your own money.
Here's the pattern that matters. Every major bubble in market history, the railroads, the dot-com, the housing crisis, followed the same script. A genuinely transformative technology emerges. Investment pours in. Valuations detach from fundamentals. The people who point out the disconnect get called crazy and then math catches up.
Not always on the schedule the bears predict, but eventually. The question you need to ask yourself right now is not whether AI is real. It is. The question is whether the prices you're paying for AI stocks today already reflect a decade of perfect execution. Because if they do, and if anything goes even slightly wrong, the downside is enormous.
Burry's own defenders have a phrase they keep repeating. Being early is not the same as being wrong. His 2008 housing short appeared to be failing for nearly 2 years before the market proved him right. He was mocked, he was questioned by his own investors, and then the entire financial system collapsed exactly as he predicted.
Is this that moment? I don't know. Nobody does. But what I do know is that the man who saw 2008 coming has placed his bet, closed his fund, and walked away. The insiders at the two biggest AI companies on Earth are selling billions of dollars in stock. The Shiller CAPE ratio is at a level that has only been reached twice in 155 years.
And the AI industry's largest company just revised its cash burn projection to $665 billion. You can choose to ignore all of that. Or you can choose to at least look at your portfolio tonight and ask one simple question. If the market drops 30% tomorrow, am I prepared? Because the people who get rich through market crashes are not smarter than everyone else.
They're not luckier. They're just the ones who asked that question before the crash, not after.
Comments 0