Michael Burry is bearish again.
Every time the real-life inspiration behind The Big Short issues a warning, the market responds in two completely different ways.
Some immediately felt anxious, given that he had correctly predicted the housing market collapse before the subprime crisis; others dismissed it, thinking Burry has been bearish for so many years while the market has continued to rise time and again.
This time is no different.
Burry released a trading update this week in his paid report Cassandra Unchained, warning: "I still believe we may be nearing a major top or a crash similar to 1987; however, the S&P 500 reaching a new high is likely to attract new capital."
What is Bury really concerned about?
Burry's logic isn't just that "AI stocks are too expensive," but that high valuations, low volatility, and systematic trading are reinforcing each other.
He believes that when markets continue to rise and volatility declines, volatility-targeting funds, trend-following, and momentum strategies increase equity exposure and even raise leverage according to their models. This rally attracts more capital, while low volatility encourages greater risk-taking, creating a self-reinforcing cycle.
The risk is that this mechanism can also operate in reverse. Once the market suddenly declines and volatility rises, related strategies may be forced to reduce positions simultaneously. Selling drives up volatility, and rising volatility triggers even more selling, ultimately leading to a mechanical cascade.
Burry's analysis states that, compared to the period from 1985 to 2000, today's markets experience stronger short-term volatility shocks during sharp declines, but panic also dissipates more quickly. He is not concerned about a typical bear market, but rather about brief, severe imbalances caused by excessive concentration of positions.
This risk is not unfounded. Bury cited data from BTIG technical analyst Jonathan Krinsky, noting that the S&P 500 has only risen 5% and reached a new high over four trading days three times in the past several decades, with one instance occurring near the peak of the 2000 tech bubble.
Of course, the history of 1987 cannot be simply replicated. Today’s U.S. stock market has circuit breakers, and its market structure, trading mechanisms, and participant composition have changed. Even if risks materialize, it is unlikely to replay with a single-day drop of more than 20%.
So, more accurately, Burry is reminding the market that the surface calm may be built on increasingly crowded positions.
Record: Made a fortune in six months, lost it all in a few days?
Whether Burry's logic makes sense is one thing; whether the trade made money is another.
From publicly available information, he has been shorting the AI and semiconductor sectors in recent years, with particular attention on Palantir, NVIDIA, and the semiconductor ETF SOXX.
In 2025, Baur's fund disclosed put options on Palantir and NVIDIA in filings submitted to the U.S. Securities and Exchange Commission. However, such filings only reveal end-of-quarter positions and do not provide information on option premiums, strike prices, expiration dates, or whether the positions were later reduced or closed out.
Therefore, we cannot accurately calculate his actual earnings.
However, his subsequent public writings still reveal his trading approach.
His skepticism toward Palantir was most pronounced. In Bury’s view, the company’s biggest issue isn’t a lack of growth, but that its stock price has already priced in an overly perfect future. He questioned Palantir’s valuation, accounts receivable, equity compensation, and business replicability, arguing that the market’s price for the company far exceeds what its fundamentals can reasonably justify.
However, Palantir subsequently delivered an earnings report that gave short sellers a hard time. The company’s second-quarter revenue grew 93% year-over-year, with both its U.S. commercial and government businesses maintaining strong growth and raising its full-year guidance once again. After the earnings release, Palantir’s stock surged nearly 30% in a single day, quickly reversing most of its prior decline.
This is why many people say that the advantage Burry accumulated through months of shorting was wiped out by the market in just two days.
Although somewhat exaggerated, this statement highlights the challenge of shorting high-growth companies: you may be right about the valuation, but you might not live to see it revert.
NVIDIA is in a similar situation.
Burry is concerned that excessive AI capital expenditures are creating a self-reinforcing investment cycle among cloud computing companies, data centers, and chip manufacturers. If future AI revenues fail to cover these massive investments, the entire supply chain could face significant adjustments.
However, as of now, NVIDIA's performance remains strong. In the most recent quarter, the company's revenue increased by 85% year-over-year, data center revenue grew by 92%, and profits and cash flow significantly outperformed most high-valued tech stocks.
In other words, the risks Burry is concerned about may exist, but the market has not yet seen a turning point.
In comparison, his short positions on semiconductor-related assets such as SOXX performed better. The significant correction in the semiconductor sector in July at least suggests that Burry’s assessment of the sector being overheated and overly crowded was not entirely wrong.
He identified certain risks and achieved阶段性 profits on some trades, but the two most important AI standout stocks have not yet followed his script.
Additionally, according to the X account @MichaelJBurry__, in addition to SOXX, Palantir, and NVIDIA, Burry’s current short positions also include the following popular assets:
Oracle (ORCL): Entry price $144.63 (August 6, 2026), position size undisclosed.
Nebius (NBIS): Entry price of $211.77 (August 6, 2026).
Micron Technology (MU): Initial entry price approximately $1,051.87 (July 2, 2026), followed by additional position averaging at approximately $933.86 (July 24, 2026); position size undisclosed.
Tesla (TSLA): Entry price of $416.22 (as of June 30, 2026), position size undisclosed.
Invesco QQQ Trust (QQQ): puts, rolled over to February 2027, size undisclosed.
Wall Street hasn't caught up with the shorts yet.
The recent market cooldown has indeed provided some support for Baur's assessment.
Vincent Lin of Goldman Sachs Global Markets stated in a research report at the end of July that the high-beta momentum portfolio has retraced 32% from its peak, and the tech long-short momentum portfolio has retraced nearly 40%; the scale of selling by tech hedge funds has reached the highest level in Goldman Sachs’ data over the past decade. This indicates that the previously crowded AI trades are rapidly deleveraging, and Burry’s concerns about position vulnerability are not unfounded.
But Lin did not turn pessimistic as a result. He referred to this correction as a "healthy reset," believing that capital was being reallocated away from overconcentrated positions rather than abandoning the long-term logic behind AI; after some trades stabilized, Goldman Sachs also observed renewed buying interest.
The view of the UBS Global Wealth Management Chief Investment Office is similar.
UBS noted on July 22 that what truly matters going forward is not whether tech companies are still willing to invest in AI, but whether revenue growth can cover the depreciation of data centers, chips, and storage, as well as rising capital expenditures. UBS remains bullish on AI growth but advises investors to reduce concentration and diversify allocations into chips, hardware, tech giants, and relatively defensive sectors.
JPMorgan Asset Management believes that the trend of corporate adoption of AI is still ongoing, and industry fundamentals have not reversed; however, before it is proven that capital investment can generate sufficient returns, investors need to be more selective rather than treating the entire AI sector as a single trade.
This may be precisely the point where Burry diverges from the mainstream Wall Street view: both sides recognize the risks of high valuations, crowded trades, and massive capital expenditures, but Burry fears these cracks could trigger a crash, while mainstream institutions are more inclined to view recent volatility as an internal realignment.
As of August 6, Palantir is still trading at a price-to-earnings ratio of over 130, while NVIDIA’s P/E ratio is approximately 33. Although both companies are popular AI stocks, their valuations and fundamentals have clearly diverged.
Therefore, what needs to be observed next may not be the overly broad question of whether an “AI bubble” exists, but rather who can turn massive investments into sustained profits, and who can only maintain their stock price by relying on ever-rising expectations.
The market has not yet answered whether what Bery saw was merely a normal retreat of a crowded trade, or the first crack before a larger turning point.
Author: seed.eth
Twitter: https://twitter.com/BitpushNewsCN
BitPush Telegram community: https://t.me/BitPushCommunity
BitPush TG subscription: https://t.me/bitpush