Why Michael Burry Has Got the AI Trade Wrong

Michael Burry has earned the right to be taken seriously. The problem is that his AI trade is starting to look like the classic mistake smart bears make after one legendary win: they spot real excess, then treat the entire market category as fake.

My view is simple: Burry is right that parts of the AI trade are stretched. He is right that semiconductor cycles can overbuild. He is right that investors are chasing anything with an AI label. But he is wrong if the conclusion is that the whole AI trade is another 2008-style illusion.

The reason: the best AI companies are not selling mystery paper to customers who cannot pay. Nvidia is printing real revenue, real margins, and real data-center demand. Palantir is showing real AI software adoption in U.S. commercial and government markets. There can be a correction and Burry can still be wrong about the core direction.

The first big public signal was Scion Asset Management's 13F for the quarter ended September 30, 2025. The filing showed put options covering 1 million Nvidia shares, valued at $186.58 million, and put options covering 5 million Palantir shares, valued at $912.1 million. The filing was signed by Michael J. Burry on November 3, 2025.

That was not a tiny footnote. MarketWatch described it as Burry challenging two of the market's favorite AI names, and the SEC information table confirms the Nvidia and Palantir puts. Then in July 2026, Burry widened the attack: fresh disclosed bearish bets against Nvidia, Applied Materials, Tesla, Caterpillar, and the iShares Semiconductor ETF, plus a reported short against Micron.

So this is not just one valuation quibble. Burry is aiming at the AI infrastructure trade: chips, memory, semiconductor equipment, power-adjacent industrial winners, software winners, and the psychology around all of it.

This is the part a lazy bull skips, and it matters. Burry is not wrong to smell froth. There is froth. There are AI names trading like every dollar of future demand is already locked in. There are second-derivative trades where investors are not buying earnings, they are buying proximity to the story.

If Burry's point is "do not buy every chart with an AI sticker," I agree. If his point is "a violent AI correction is possible," I agree. If his point is "some investors are confusing capex with profit," I agree.

But that is not the same as saying the AI trade is wrong. That is saying the trade needs a filter.

The AI trade is not one trade. It is a stack. Nvidia is not the same as a speculative software shell. Palantir is not the same as a no-revenue AI story. Micron is not the same as Caterpillar. SOXX is not the same as a single company.

That distinction is the whole game. Burry is attacking the category. The market is rewarding the companies that can actually convert the category into revenue.

Nvidia's latest official numbers are the cleanest counterargument. For Q1 fiscal 2027, Nvidia reported record revenue of $81.6 billion, up 85% from a year earlier. Data Center revenue was $75.2 billion, up 92%. GAAP gross margin was 74.9%. The company also guided Q2 fiscal 2027 revenue to about $91 billion, plus or minus 2%.

That is not dot-com vapor. That is a monster business selling the bottleneck product into the biggest infrastructure buildout on earth. You can argue the stock is expensive. You can argue customers are over-ordering. You can argue the cycle will cool. But calling the whole AI trade wrong while the main pick-and-shovel company is growing like that is fighting the scoreboard.

Burry's legend comes from seeing through housing. That short worked because the cash flows underneath the structure were rotten. Bad loans were packaged, rated, levered, and passed around until the system pretended weak borrowers were safe assets.

AI is different. The biggest winners are not hiding defaults inside securitized paper. They are selling chips, networking, software, compute, and workflow systems to customers with actual budgets.

That does not make AI stocks safe. It does make the analogy weaker. In housing, the core asset quality was fake. In AI, the core question is whether the current spending wave produces enough productivity and revenue to justify the capex. That is a very different debate.

The right comparison is not "2008 housing." The better comparison is "internet infrastructure after 1999." A lot of companies can blow up while the underlying technology still changes the world. The winners and losers separate over time. Shorting the whole wave because the worst parts are ugly is how bears miss the compounding machines.

Palantir is expensive. I will not pretend otherwise. But the bear case gets weaker when the company keeps turning AI demand into numbers. MarketWatch reported Palantir's Q1 2026 revenue rose 85% year over year to $1.63 billion, ahead of consensus, with adjusted earnings of 33 cents per share.

That matters because Palantir is not just using AI as a marketing word. Its pitch is operational AI: bring models, data, security, permissions, auditability, and decision workflows into institutions that cannot just plug sensitive systems into a random chatbot. Government, defense, manufacturing, healthcare, finance, logistics - those customers care about reliability, control, and speed.

Again, valuation can be too high. The stock can get cut in half and still have been a real business. But "too expensive" is not the same thing as "fake." Burry seems to be pressing the first idea into the second.

This is where I think most investors need to level up. The phrase "AI trade" is too sloppy. There are at least six different trades hiding inside it.

Burry is most likely right about the last bucket and maybe early on parts of memory, equipment, and power if capacity outruns demand. But if he is using that to short the best parts of the stack, the trade can be too blunt.

This is the sentence I want on the wall: valuation risk is not thesis risk.

A stock can be overpriced and the company can still be executing beautifully. A stock can fall 30% because expectations got too hot and still be a long-term winner. A bear can make money on a correction and still be wrong about the structural trend.

That is the trap with celebrity shorts. People see Burry short something and assume the business is broken. Sometimes the simpler explanation is that the stock is extended. Those are not the same trade.

If Nvidia drops because the market reprices AI multiples, that does not mean AI demand vanished. If Palantir drops because the multiple compressed, that does not mean customers stopped buying AIP. If Micron falls after a parabolic run, that does not mean the memory cycle was fake.

I am not dismissing the bear case. Here is what would make me take Burry's AI short much more seriously.

Those would be real warning signs. The problem is that the strongest current evidence is not pointing there yet. The current evidence says the trade has pockets of excess around a real demand center.

I would not blindly chase the AI leaders here. That is not the takeaway. The takeaway is to stop treating "AI bubble" as a complete investment thesis. It is a starting question, not an answer.

My playbook is simple: own or watch the companies where revenue proves the story, avoid the obvious AI cosplay, respect position size, and use pullbacks instead of FOMO candles.

That is more useful than copying a famous short. Burry can be early, right, wrong, or profitable for reasons that do not fit your account size or time horizon. Your job is not to become his shadow. Your job is to know what you own and why.

Michael Burry has got the AI trade wrong if he is treating the whole thing like a fraud cycle. The better read is that AI has both real platform winners and obvious bubble pockets.

I think he is right to attack the slop. I think he is right to warn about overvaluation. I think he may even make money if the hottest AI names get hit. But the leaders are not imaginary. Nvidia's revenue is not imaginary. Palantir's growth is not imaginary. Enterprise and government demand for AI systems is not imaginary.

The smarter stance is not blind bullishness. It is selective bullishness. Buy evidence, not hype. Respect valuation, not fear. And never confuse a violent correction with the end of a platform shift.

SEC Scion 13F primary document and SEC Scion 13F information table for the September 30, 2025 reporting period, Nvidia put position, Palantir put position, table totals, and Burry signature.

Nvidia official Q1 fiscal 2027 results for revenue, Data Center revenue, gross margin, share repurchase authorization, and Q2 revenue outlook.

MarketWatch Burry Nvidia and Palantir report for public reporting around the Scion Nvidia and Palantir put positions.

MarketWatch Palantir Q1 2026 earnings report for Palantir's Q1 revenue growth and adjusted earnings context.

MarketWatch Micron short report and Business Insider July 2026 Burry short report for Burry's newer reported shorts against Micron, Nvidia, Applied Materials, Tesla, Caterpillar, and SOXX.

Nasdaq NVDA quote page, Nasdaq PLTR quote page, and Nasdaq SOXX quote page for current market-reference pages checked before publication.

*Disclaimer: MentorSurge is not a financial advisor and this is not financial advice. This post is for educational and entertainment purposes only. Nothing here is a recommendation to buy, sell, short, or hold any security. AI stocks, semiconductor stocks, options, and short positions can be extremely volatile. Numbers cited were checked before publication and can change quickly. Always do your own research and consult a licensed professional before making decisions with real money.*