원문 보기
It's here. Michael Burry just made his biggest market crash call since 2008. The man who shorted housing before anyone else opened his Substack post this past weekend with two specific sentences. I am now a meme for the number of times I have called a crash. I have become the boy who cried wolf. Then, paragraphs later, he writes six more words he hasn't used in 17 years.
The market has jumped the shark. And what he wrote underneath those six words, the charts he pulled, the math he ran, the four leveraged short trades he placed in the last two weeks, is the most aggressive bearish stance Burry has taken since 2008. Let's break down exactly what he said, why he said it now, and what the data underneath actually shows.
Start with the track record, not the meme version, the real one. Burry left a partial neurology residency at Stanford in 2000 to start a hedge fund called Scion Capital, funded by an inheritance and family loans. His first major call was shorting Amazon in February 2000, right before the stock lost about 80% of its value in the dot-com crash.
His second major call, the one that made him famous, was shorting the subprime mortgage market between 2005 and 2008, which made his fund a profit of more than $700 million while the rest of Wall Street burned. In 2021, he called what he labeled the greatest speculative bubble of all time in all things, and the Nasdaq promptly lost roughly 34% in the first half of 2022.
In 2023, he called the regional banking crisis weeks before Silicon Valley Bank collapsed, comparing it to Enron-style accounting fraud. That's the actual scorecard, but the meme version is different because for every accurate call, the financial press also reported a dozen so-called crashes that Burry never actually called.
They came from reporters misreading his quarterly 13F regulatory filings. In November 2025, Burry tweeted the math directly, pointing out that one position the media reported as a $912 million bet against Palantir was actually $9.2 million in option premium. The headlines weren't reading his filings.
They were guessing. That distinction matters because this time he is not posting cryptic numbers. He is writing complete sentences and the language is unambiguous. The post is titled About That Boy Who Cried Wolf, published May 10th. For an investor who normally writes about specific stock tickers and price-to-sales ratios, language that direct is rare.
He hasn't used anything close to it since the housing market in 2007. What he pulls up next is a sequence of charts. The Nasdaq 100 today overlaid against the bubble peak of March 2000, the S&P 500 today overlaid against the run-up into October 2007, the Dow Jones into the September 1929 peak, the S&P 500 in 1972 right before the second wave of 1970s inflation crushed equities for the better part of a decade.
In each chart, he marks today's position with a yellow arrow and the same two words, "You are here." Each of those four moments, 1929, 1972, 2000, 2007, is one of the largest equity drawdowns in modern history and the chart pattern today matches all of them at once. The Philadelphia Semiconductor Index just went near vertical in early May.
The S&P 500 is up 48% the NASDAQ 100 is up 69% and the move is happening at the same moment US consumer sentiment hit the lowest reading on record. The University of Michigan Consumer Sentiment Index dropped to 48.2 in early May, lower than the depths of the 2008 financial crisis, lower than the COVID lockdowns, lower than the inflation panic of mid-2022.
The stock market has never been higher, the consumer has never been more pessimistic. That gap is what Burry is calling attention to. The technical detail that really got Burry's attention came from Jonathan Krinsky, the chief market technician at BTIG. Krinsky pulled a number that has only happened three times in modern market history.
When the S&P 500 trades more than 7% above its 50-day moving average, breath, meaning the percentage of individual stocks holding their own uptrends, has never come in below 55%. Historical average is 86%. Last week it came in at 52%. The other three dates in the last 30 years that came close were December 28th, 1998, November 19th, 1999, and March 27th, 2000.
Every single one within 20 trading days of the dot-com top. Krinsky then ran a different comparison. He looked at the top 10 performing stocks in the NASDAQ 100 over the trailing 12 months. In 1999, the top 10 averaged a 559% gain. In the 12 months leading to the March 2000 peak, the top 10 averaged 622%.
Over the last year, the top 10 in the NASDAQ 100 averaged a gain of 784% bigger than 1999. Bigger than the run into the March 2000 peak itself. And the cast of characters is eerily familiar. In 1999, bubble money rotated out of consumer internet stocks and into the hardware companies that supplied them.
Applied Materials, Samsung Electronics, JDS Uniphase. The logic was simple. Whichever software company won, the hardware suppliers would capture the revenue. JDS Uniphase ran up 4,636% into the March 2000 peak. Within 2 years, it had collapsed by about 95%. Applied Materials dropped nearly 40% in 2000 alone.
The wider NASDAQ lost roughly 78% peak to trough. Today, the same playbook is running. Samsung Electronics, Applied Materials, and Viavi Solutions, the successor to JDS Uniphase, are moving in the exact same vertical pattern they did into March 2000. Samsung Electronics was Burry's second largest position 1 year ago.
He has been trimming. The stock is up roughly fourfold in 12 months. That fourfold gain is the reason he is now writing about jumping the shark instead of writing routine trading posts. The next layer of Burry's argument is what he calls the tragic algebra. The basic claim is that the headline price-to-earnings ratio of the NASDAQ 100, the number CNBC quotes, is roughly 30 times forward earnings.
Burry argues the true number is closer to 43 times. The gap comes from a handful of accounting choices that, while technically legal, hide the real cost of running the modern tech business. The biggest single distortion is stock-based compensation. Wall Street's adjusted earnings metric adds it back, treating it as if shares given to employees are free.
They are not. They dilute existing shareholders by hundreds of billions of dollars a year across the largest tech firms. Strip out that adjustment, account for depreciation timing on data center build-outs, adjust for capital leases and ongoing M&A goodwill, and Burry calculates the largest growth companies in the Nasdaq 100 are overstating their real owner earnings by more than 50%.
That math is the difference between a 30 times multiple, which feels rich, and a 43 times multiple, which is dot-com territory. And then there is the inventory problem. In 2001, Cisco was the most valuable company on Earth, the backbone of the internet, locked into long-term supply contracts to secure the components it thought it needed.
When demand collapsed, Cisco was left holding components it could not use. The company recorded a $2.25 billion dollar inventory write-off in a single quarter. The stock dropped from $82 to $13. Today, Nvidia sits in roughly the same position. As of October 26th, 2025, Nvidia was carrying $50.3 billion in long-term inventory purchase and supply commitments to secure capacity at TSMC and other suppliers.
And the early stress is already showing. In the first quarter of fiscal 2026, Nvidia took a $4.5 billion charge charge against excess H2O chip inventory after export restrictions to China killed demand for one specific product line. That charge is what a Cisco-style write-down looks like in miniature.
A single product, a single quarter, $4.5 billion dollars vaporized. Burry's point is simple. If end demand for AI compute fails to track CapEx projections, which Morgan Stanley now estimates at roughly $1.1 trillion dollars total AI infrastructure spending by 2027. Those supply commitments turn from strategic assets into write-downs.
The single cleanest piece of evidence underneath all of this has nothing to do with charts or breadth or accounting. It is one ratio, the Buffett Indicator. Named after Warren Buffett, who once called it the best single measure of where valuation stand. The math is simple. Total United States stock market value divided by United States GDP.
If the value of every public company in America is growing far faster than the actual economy producing the underlying goods, prices are detaching from reality. The historical readings give you the calibration. Before the 2007 housing crash, the Buffett Indicator peaked at roughly 118%. Before the dot-com crash in 2000, it peaked at about 159%.
As of the most recent quarterly update, the Buffett Indicator sits at 230%. That is approximately 2.4 standard deviations above the long-term historical trendline. Statistically, that is the kind of reading you see roughly once every several decades. Practically, it categorizes the market as strongly overvalued by the indicator's own framework.
There are reasonable criticisms. Multinational companies earn revenue overseas that doesn't show up in domestic GDP. Low interest rates over the last decade arguably justified some elevation. None of those caveats explain a reading that is 75% above the indicator's own historical trendline. They explain that the market can stay overvalued for longer than people expect.
They do not explain the size of the eventual reversion. And that is the point Burry is hammering. The reading is not we are due. It is we are at the most extreme valuation level ever recorded with weaker market breadth than 1999, with consumer sentiment lower than 2008, with hardware suppliers locked into Cisco-style supply commitments.
all of it at the same time. This is the part most people miss. Burry has not stopped at writing the warning. In real time, over the last 2 weeks, he has built one of the most aggressive bearish portfolios he has constructed since 2008. His trading posts from May 6th through May 8th document the trades.
He added to put options against SOXX, the semiconductor ETF, expiring January 2027 at a strike of $330. He added Nvidia puts expiring January 2027 at a March 2027 at $125. He added QQQ puts, those bet against the entire Nasdaq 100 at a $550 strike, also January 2027. Each of these is a deeply out-of-the-money put.
The structure is designed so that if the index moves modestly lower, the contracts expire worthless and Burry loses what he paid. But if the index drops 30% or more before expiration, those same contracts can multiply many times over. It is the same convexity structure he used in 2007 against the housing market.
At the same time, he is rotating fresh capital into stocks that look nothing like the AI trade. He sold his remaining smaller positions in Salesforce and Autodesk to buy a full position in Zoetis, the animal health leader, at roughly $86 per share. He bought a new full position in Mercado Libre, the Latin American e-commerce platform, in the $1,600 range.
He doubled his position in Temple and Webster, an Australian online retailer. He added to Tencent and Meituan in Hong Kong, and he started a new position in Higher Smart Home, a Chinese appliance manufacturer. The architecture is unmistakable. Leveraged shorts against the consensus trade, concentrated long positions in deep value international names.
It is exactly the architecture of his portfolio entering the dot-com crash. The question everyone asks at this point is, "What actually breaks it?" Burry's answer is the honest one. In 2000, there was no specific catalyst. The market just stopped going up. In 1929, no proximate cause. In 2007, the catalyst arrived late in the form of Bear Stearns and Lehman, long after the structural rot was already obvious.
He lists several candidates for this cycle. The first is the ongoing conflict involving Iran, which has pushed Brent crude back above $103 per barrel and elevated US gasoline prices roughly 40% above the start of the year. Energy inflation flowing through Q2 inflation prints could force the Federal Reserve to hold rates higher for longer than markets expect.
And the entire current valuation regime is priced on the assumption of rate cuts. The second is private credit. The data center build-out funding, AI is increasingly financed not by banks, but by private credit funds. Apollo and Blackstone are reportedly working on a $35 billion private credit facility for Broadcom.
If revenue from these data centers fails to support the debt service, the contagion path runs through opaque private balance sheets that regulators don't have full visibility into. The third, and this is the dark horse, is a Treasury action. Treasury Secretary Scott Bessent has floated radical ideas about extending the duration of US debt issuance.
A forced rollover of foreign-held Treasuries into 100-year bonds would violently reprice the long end of the yield curve and crush the long duration growth stocks at the heart of the AI trade overnight. Burry ends the May 10th post with one of his characteristic lines, "There is no telling, so I am telling." Below that, he places a Bloomberg chart of the Philadelphia Semiconductor Index in white, the 2000 peak, overlaid in yellow against the same index today.
The two lines are almost identical. The fable about the boy who cried wolf is a story most people remember wrong. They remember that the boy lied. What they forget is the ending. The wolf does eventually come. And by the time it does, nobody is listening. Burry is not saying the wolf is at the door tomorrow.
He is saying the wolf is now in the field. Whether the market drops in May, in October, in 2027, or starts a slower grind down from here, that part is unknowable. What is knowable is that the man who shorted the housing market just stopped hedging his language.
Comments 0