저자: MichaelBurry · 외국인
요약 보기
버리가 엔비디아·팔란티어·테슬라·오라클·반도체지수(SOXX)·QQQ에 풋/공매도 포지션을 잡았다고 설명. 근거로 2026년 5월 시장 밸류에이션(실러 CAPE 40.1, 1999년 이후 2번째 고점), SOX 연초 대비 65% 상승, 빅테크의 GPU 감가상각 연장(약 1,760억 달러 과소계상) 주장 제시. 버리의 발언을 2000년 닷컴 버블 말기에 비유. 반대 시각으로 엔비디아 실적 기반과 GPU 추론 수요 가능성도 언급. 영상은 투자 판단 대신 포트폴리오 집중도 점검을 권고.
다듬은 글
Michael Burry just massed almost 80% of his entire portfolio against the biggest AI stocks in America. He's shorting Nvidia. He's shorting Palantir. He's shorting Tesla. He's shorting Oracle. He's shorting the entire semiconductor index. He bought puts on the Nasdaq 100. And on May 8th, 2026, he posted on X that the market right now feels like the last months of the 1999 to 2000.com bubble.
The same man who spent three years studying every mortgage contract in America before betting against the housing market and making $800 million is now telling you the AI trade is about to collapse. But, here's what nobody is telling you. This isn't a billion-dollar bet. Every headline you've seen about Burry is wrong.
The media reported he made a $912 million bet against Palantir. Burry himself went on X and corrected them. His exact words, "I bought 50,000 of these things for $1.84. Each of those things is 100 doodads. So, I spent $9.2 million. Not 912 million." The media confused the cost of his bet with the value of the stock he's betting against.
And that distinction changes everything about what Burry is actually doing. Understanding the difference will teach you more about how Wall Street really works than anything you learned in school. Let me show you exactly what happened. On May 4th, 2026, Burry's fund, Scion Asset Management, disclosed that it sold every single share of GameStop.
1.7 million shares gone. The stock that literally made Michael Burry famous, the stock that launched the meme stock revolution, the stock that became the movie The Big Short, he dumped all of it. And the reason he sold is a story in itself. Ryan Cohen, GameStop's CEO, made an unsolicited offer to buy eBay for $56 billion.
$56 billion. GameStop, a company that most people associate with used video games and mall kiosk, tried to acquire one of the largest e-commerce platforms on the internet. Burry had originally bought GameStop because he believed Cohen could use the company's massive cash pile to make smart acquisitions and turn GameStop into something resembling Berkshire Hathaway.
He called it the instant Berkshire thesis, but the eBay bid changed everything. At the proposed valuation, the deal would push GameStop's leverage to roughly 7.7 times debt to EBITDA. That's borderline distressed territory. Burry posted one sentence on X, "Never confuse debt for creativity." And here's the part that tells you everything about Ryan Cohen.
After announcing the bid, Cohen posted on X saying he had come up with a creative way to help fund it. He was selling stuff on eBay to pay for eBay. Baseball cards, video games, old GameStop merchandise. He listed it all on his personal seller account. The top bids totaled about $138,000 toward a $56 billion acquisition.
Then eBay banned his account. You can't make this up. The CEO of a public company trying to buy eBay got banned from eBay while selling merchandise to fund the purchase. Burry saw all of this and sold every share the next day. At the same time, he exited Salesforce and Autodesk. He cleared the deck, and then he loaded up on one trade, the biggest concentrated bet of his career since 2008.
Here are the numbers. Puts on 1 million shares of Nvidia, strike price $110, expiring January 2027. Nvidia closed on May 8th at $215. That means Burry is betting Nvidia drops 49% in the next 8 months. Puts on Palantir at a $100 strike expiring December 2026 and $50 strike expiring June 2027. He also opened direct short positions on Palantir stock.
Puts on Oracle, which has already fallen 51% from its peak in late 2025. Puts on the SOXX semiconductor ETF, strike 2027. And puts on QQQ, the Nasdaq 100 ETF, strike $525, expiring March 2027. He also has direct short positions open on Tesla. In total, roughly 9 and 1/2% of his portfolio is positioned to the downside against the AI ecosystem through put options and direct short.
But, here's the part that will surprise you. While Burry is shorting Nvidia, Palantir, Tesla, and Oracle, he's actually buying one tech stock, Microsoft. In April 2026, Burry disclosed a new long position in Microsoft. He called it a company with fortress-like fundamentals, high-margin software, and a cloud business that dominates globally.
When someone asked why he wasn't shorting Microsoft, too, he said, "If I short Microsoft, I'm shorting a global office productivity SaaS Goliath." Think about what that tells you. Burry doesn't think tech is dead. He doesn't think AI is fake. He thinks specific companies are wildly overpriced and that the hype has disconnected stock prices from reality.
But, the companies with real diversified cash-generating businesses, he's buying those. This is not a blanket bet against technology. It's a surgical bet against the most speculative parts of the AI trade. Now, here's why the timing mattered. The Philadelphia Semiconductor Index, the SOX, is up 65% this year alone.
65% in 5 months. The SOXX ETF is currently trading 60% above its 200-day moving average. That is the widest gap between the price and its long-term average since the ETF was created in July 2001. Let me put that in perspective. That span includes the dot-com crash, the 2008 financial crisis, the 2018 trade war, the 2020 pandemic crash, the 2022 chip recession.
In none of those periods did semiconductor stocks get this far above their average. The only time anything similar happened was March 2000, the exact month the dot-com bubble peaked. And it's not just semiconductor. The Shiller cyclically adjusted price-to-earnings ratio for for entire S&P 500 stood at 40.1 as of May 8th, 2026.
That number has only been higher one time in the last 150 years, November 1999, when it hit 44.19. That was the all-time peak, the absolute top of the dot-com bubble. And after that peak, the S&P 500 fell 49%. The Nasdaq fell 78%. We are now at the second highest valuation reading in the entire history of the American stock market.
That doesn't mean a crash is coming tomorrow, but it means the market is priced as if nothing can go wrong. And when something eventually does, the fall from these heights is historically severe. And that's exactly what Burry said on CNBC on May 8th. He said the market today feels like the last months of the 1999 to 2000 bubble, not the middle of a bull market, the last month, the final push before everything reversed.
He added something else that most people missed. He said, "People mistakenly believe they understand the two letters AI. And it feels like watching the final month of the bubble transitioning from 1999 to 2000. In 1999, Qualcomm stock rose 2,619% in a single year. 12 other large-cap stocks rose over 1,000%.
The Nasdaq composite hit its peak in March 2000 and then fell 78% over the next 2 and 1/2 year. Three out of every $4 invested in tech stocks disappeared. Now, let's talk about why Burry specifically targeted Nvidia and Palantir. Because this is where it gets interesting. Nvidia is a $5.23 trillion company.
That makes it the most valuable company on Earth. Its PE ratio is 43, which actually sounds reasonable compared to the rest of the AI sector. But Burry isn't shorting Nvidia because of its valuation. He's shorting it because of what its customers are doing with their accounting. On May 8th, Burry posted this on X.
Understating depreciation by extending useful life of assets artificially boosts earning. One of the more common frauds of the modern era. Massively ramping capex through purchase of Nvidia chips on a two-to-three-year product cycle should not result in the extension of useful life. Let me translate that.
Meta, Amazon, Microsoft, Google, and Oracle are spending hundreds of billions of dollars buying Nvidia's AI chip. Those chips have a useful life of roughly two to three years before they become obsolete and need to be replaced. But these companies are recording them on their books as lasting five to six years.
Why does that matter? Because when you spread the cost of a chip over six years instead of three, you record half the expense each year. Lower expense means higher reported profit. Higher reported profits mean higher stock prices. And here's the proof that the product cycle really is two to three years.
Nvidia released the Hopper architecture in 2022, then Blackwell in 2024, then Rubin in 2026. Rubin Ultra is already announced for 2027. That's a new generation of chips every one to two years. Each generation is dramatically faster than the last, which means the chips companies bought in 2024 are already becoming second-tier hardware in 2026.
Depreciating them over six years is like saying your 2020 iPhone will still be competitive in 2026. It won't be, and everyone knows it. But the accounting says otherwise. Burry calculated that this accounting practice is hiding a hundred and seventy-six billion dollars in understated depreciation between 2026 and 2028.
That's a hundred and seventy-six billion dollars of profits that exist on paper, but don't reflect the actual cost of doing business. If those companies were forced to depreciate their AI hardware over the actual useful life, their earnings would drop significantly. And if their earnings drop, their stock prices follow.
And if the biggest buyers of Nvidia chips suddenly look less profitable, the market starts asking whether they'll keep spending at the same rate, which hits Nvidia's revenue, which hits Nvidia's stock. It's a chain reaction. And Burry is betting on the first domino. Now, look at Palantir. This is the stock Burry is most aggressively shorting.
He's using multiple expiration dates, multiple strike prices, and direct short position. And he didn't just tweet about it. He published a full valuation model on his Substack in February 2026. His base case puts Palantir's fair value at $46. His bear case puts it at $21. And even his most optimistic scenario, the absolute best case, puts it at $146.
The stock currently trades at $146, which means even in Burry's best-case scenario, there is zero upside from here. Every other scenario points down significantly. Palantir trades at 153 times trailing earning. For comparison, Apple trades at 33 times. Microsoft trades at 35 times. Even Nvidia at 43 times looks cheap next to Palantir.
And Palantir's market cap is roughly $260 billion on $7.6 billion in revenue. That's a price-to-sales ratio of about 34 times. During the dot-com bubble, Cisco peaked at a price-to-sales of roughly 28 times. Palantir is more expensive than Cisco at the peak of the dot-com bubble. And Cisco fell 86% from its high and never recovered.
25 years later, it's still below its 2000 peak. In April 2026, President Trump posted about Palantir's work with the US government, and the stock jumped. Billions of dollars in market cap added in a single day on one social media post. Burry responded by adding more to his short position. He went on CNBC and said he was still betting against it.
Most traders would have covered after a presidential endorsement. Burry doubled down. He told his Substack readers he's shorting the inherent unsustainability of its business model as well, not merely its valuation. Now, Palantir bulls will tell you the revenue growth justifies it, and the growth is real.
68% revenue growth in Q1 2026, that's not nothing. But, Burry's argument isn't about whether Palantir is growing. His argument is about what happens when growth slows even slightly. When a stock is priced at 153 times earnings, any deceleration in growth doesn't cause a small dip. It causes a collapse because the entire valuation is built on the assumption that growth continues at this rate for years.
If it doesn't, the math break. And Burry's track record on exactly this kind of trade is worth noting. He first opened his Palantir short when the stock was around $207. It's now at $137. That's a 34% decline. He hasn't taken profits. He's adding to the position. That's conviction. Now, here's where this connects to your money because this isn't just about hedge fund games on Wall Street.
These are stocks sitting in your retirement account right now. If you own an S&P 500 index fund, you own Nvidia. It's the largest holding. If you own QQQ, you own both Nvidia and Palantir. If you have a target date retirement fund through Vanguard, Fidelity, or Schwab, you almost certainly own semiconductor stocks through the index funds it hold.
Nvidia alone makes up over 7% of the S&P 500. One company, 7% of the entire index. If Nvidia drops 49% like Burry is betting, that single stock would drag the S&P 500 down roughly 3 and 1/2% by itself. That's before counting the impact on every other AI stock that would fall with it. And if you own Tesla in your portfolio, that's another stock Burry is directly shorting.
If you hold Oracle, same thing. These aren't obscure companies. They're the biggest names in your 401k. And one of the most famous investors alive is betting against all of them at the same time. And this isn't just Burry saying this. Paul Tudor Jones, another billionaire hedge fund manager, told CNBC on May 7th that the AI rally feels exactly like 1999.
But here's the twist. He said it still has another year or two to run. He's bullish and comparing it to a bubble at the same time. His argument is that bubbles can keep inflating long after they become irrational. In 1999, people knew tech was overvalued by July. The Nasdaq didn't peak until March 2000.
Eight more months of gains after everyone agreed it was a bubble. You could have shorted in July 1999, been completely right about the bubble, and still gone bankrupt waiting for it to pop. That's the risk Burry is taking. His Nvidia puts expire in January 2027. If Nvidia stays above $110 until then, Burry loses his entire bet, every dollar he put in.
The options expire worthless. And right now, Nvidia is at 215. It needs to drop nearly 50% in 8 months for Burry to profit. The last time Nvidia dropped that much in 8 months was the 2022 chip crack. It can happen, but it requires a major catalyst. So what could that catalyst be? Burry has pointed to three possibilities.
First, the depreciation issue. If even one major company is forced to restate earnings based on shorter GPU useful life, it triggers a domino effect across every hyperscaler. We already saw Amazon acknowledge the problem by shortening its depreciation schedule. If Meta or Google follow, the 176 billion dollar question becomes front page news overnight.
Second, AI revenue disappointment. If the company spending hundreds of billions on AI chips can't show proportional revenue growth from AI products, the spending stops. Goldman Sachs published a report in 2024 asking exactly this question. Who is paying for all this AI and where is the revenue? Two years later, the question still hasn't been answered at scale.
Third, a broader market correction. The Shiller Cape ratio at 40 is telling you the entire market is priced for perfection. If inflation stays sticky, if the Fed doesn't cut rates, if the economy slows, AI stocks fall the hardest because they're priced for perfection in a perfect environment. And the stocks with the highest multiples fall the fastest.
Now, before you take all of this and panic, consider the other side. Because this channel exists to help you think, not tell you what to do. Nvidia is not a dot com fantasy. It has real revenue. A hundred and thirty billion dollars in trailing 12-month sale. Real profit. Real products that every major tech company on Earth is buying.
The AI infrastructure buildout is the largest capital expenditure cycle in technology history. Even if there's a correction, the long-term demand for AI chips is likely to grow for decades. Shorting Nvidia in 2022 would have cost you everything. The stock went from 110 to 215 in the recovery. And the depreciation argument has a counterpoint.
Some analysts argue that AI GPUs don't become worthless after 2 years. They lose their edge for training the most advanced models, but they can still run inference workloads. The actual day-to-day processing of AI requests for years afterward. It's like a race car that can't compete in Formula 1 anymore, but still runs perfectly fine as a regular car.
If you account for this value cascade, a 5-year depreciation schedule might be closer to reality than Burry suggests. The truth is probably somewhere in between. The question is whether 3 years or 5 years is right. And that gap represents a hundred and seventy-six billion dollars in earning. And Burry has been wrong before.
He shorted Tesla in 2021 and lost money. He bet against the market multiple times between 2015 and 2019 and was early on all of them. Being early and being wrong look identical until history decide. The man himself says his timing is often off. The question is whether the structural analysis is correct, not whether it plays out on his specific timeline.
So, what should you consider doing? Not what should you do. That's for you and your financial advisor to decide. But, here are things worth understanding. First, look at your portfolio. Actually, look. If you own index funds, check how much Nvidia exposure you have. In the S&P 500, it's over 7%. In QQQ, it's even higher.
If you own individual stocks like Tesla, Palantir, or Oracle, know that one of the most famous investors alive is actively betting against all three. You might have more concentrated AI exposure than you realize. Second, understand the timeline. Burry's options expire between December 2026 and March 2027.
If something is going to happen, he's betting it happens in the next 6 to 10 months. That's the window to pay attention to. Third, consider concentration risk. This isn't about whether AI is real or fake. It's about whether having 7% of your retirement in one company is appropriate, regardless of how good that company is.
Diversification exists for a reason. Fourth, pay attention to earnings report. The next round of earnings from Meta, Amazon, Microsoft, and Google will be the first test of the depreciation thesis. If any of these companies changes how it accounts for GPU useful life, the market will react. And if none of them do, the clock keeps ticking on $176 billion in hidden costs that eventually have to be recognized.
This is the same pattern playing out across every video on this channel. If you want to understand what happens next with the AI trade, the SpaceX IPO, and what these billionaires are doing with their money, subscribe so you don't miss a single breakdown. I'll see you in the next one.
Comments 0