Skip to content Skip to sidebar Skip to footer
Western-style illustration of a sheriff holding a selloff warning as townspeople flee a chaotic frontier town, symbolizing the AI stock market downturn.

Michael Burry deserves a place in the bullish case, not merely a footnote beneath it. The financier made famous by his bet against the U.S. housing market before the 2008 financial crisis has been questioning whether the economics of artificial intelligence can justify the extraordinary valuations and spending surrounding it. By the close on Friday, November 20, 2026, the market had supplied meaningful support for his caution, even if it had not delivered a final verdict on his broader thesis. The widely publicized disclosure of his bearish AI positions arrived on November 3, 2025, when Scion Asset Management filed its third-quarter holdings report with the Securities and Exchange Commission. The filing showed put options referencing 1 million Nvidia (NASDAQ: NVDA) shares and 5 million Palantir Technologies (NASDAQ: PLTR) shares, positions held as of September 30, 2025. The filing established that the positions existed by quarter-end, not their exact purchase dates. One important correction belongs alongside the familiar “billion-dollar bet” description. The roughly $1.1 billion figure associated with those positions represented the reported value of the underlying shares, not the cash Burry paid for the options. SEC reporting rules distinguish between the underlying securities and the options themselves. The filing therefore cannot establish his actual capital at risk, purchase prices or eventual trading profits. Wall Street enjoys a dramatic headline, but an options premium and a notional exposure are not interchangeable accessories. That disclosure concerned Nvidia and Palantir, not a simultaneous, publicly documented three-stock wager against Nvidia, CoreWeave (NASDAQ: CRWV) and Nebius Group (NASDAQ: NBIS). The distinction matters when evaluating what Burry warned about, when he warned and what the subsequent price action actually proves.

How Far Nvidia, CoreWeave and Nebius Have Fallen

Using the November 3, 2025, closing prices as a consistent benchmark for the widely publicized Scion disclosure, all three AI infrastructure stocks were substantially lower by Friday, November 20, 2026. Historical daily closing-price data retrieved through Realtime Finance Data show:

  • Nvidia (NASDAQ: NVDA) declined from $206.88 to $125.53, a 39.3% decrease.
  • CoreWeave (NASDAQ: CRWV) declined from $126.32 to $39.51, a 68.7% decrease.
  • Nebius Group (NASDAQ: NBIS) declined from $120.47 to $81.31, a 32.5% decrease.

These are closing-price changes, excluding dividends, rather than declines from intraday highs. For CoreWeave and Nebius, November 3 is a shared comparison date, not evidence that Burry disclosed bearish positions in those companies that day. Friday alone was punishing. Nvidia fell 14.0%, CoreWeave dropped 12.0% and Nebius lost 16.1%, based on their November 19 and November 20 closing prices from the same historical-price dataset. Those figures support a carefully framed conclusion: Burry has been directionally correct about the vulnerability of these share prices. They do not establish the profitability of his options, which would depend on purchase prices, strike prices, expiration dates and any subsequent trades. A stock can eventually validate a warning without validating every trade used to express it.

His Company-Specific Warnings Became Sharper in 2026

Burry’s criticism developed in stages rather than arriving as one perfectly timestamped call against the entire AI ecosystem. His more recent Nvidia warning came on November 11, 2026, when he urged shareholders to exit the stock. On November 12, he extended his concerns to Nvidia-linked cloud operators, including CoreWeave and Nebius. His argument was that trouble at the central supplier could spread through businesses building and financing infrastructure around its technology. The subsequent closing-price performance has been consistent with that near-term caution:

  • Nvidia fell from $173.73 at the November 11 close to $125.53 on November 20, a 27.7% decline.
  • CoreWeave fell from $56.60 at the November 12 close to $39.51 on November 20, a 30.2% decline.
  • Nebius fell from $128.64 at the November 12 close to $81.31 on November 20, a 36.8% decline.

These comparisons use the warning dates’ closing prices as transparent benchmarks. They are not executable returns measured from the precise publication time of each comment. By November 19, Burry was publicly arguing that he had provided ample reason for caution. His message also distinguished between warning people away from an exposure and recommending that they establish short positions. That is a useful distinction: avoiding a drawdown and profiting from one are different financial exercises.

Why the “Big Short” Background Still Matters

Burry’s reputation rests on recognizing that apparently dependable mortgage assets contained risks that prevailing prices and conventional assumptions did not adequately reflect. His housing-market wager became central to Michael Lewis’s book The Big Short and its film adaptation. The relevance to AI is methodological, not a claim that graphics processors are mortgage bonds wearing better marketing. Burry’s recurring question is whether the assumptions supporting a financial structure remain credible when conditions become less favorable. In housing, that meant questioning the durability of mortgage credit. In AI, his criticism centers on equipment economics, capital spending, reported profitability and valuations. His earlier success gives those questions an audience. It does not make every subsequent forecast correct, nor does it establish that a major technological advance must produce poor long-term returns. The useful response is to examine the assumptions, not outsource the conclusion to the résumé.

What Burry Thinks Happens Next to AI

Burry’s central concern is not simply that AI share prices have risen too far. It is that the industry may be treating rapidly changing computing equipment as though its economic usefulness were more durable than it really is. In November 2025, he estimated that large cloud companies could understate depreciation by approximately $176 billion across 2026 through 2028. That was his estimate, not an SEC finding, a reported industry loss or proof of accounting misconduct. His contention was that lengthy depreciation schedules could make current earnings look stronger while postponing recognition of the economic cost of replacing equipment. His bearish scenario has several connected parts:

  1. New generations of AI chips deliver better performance and efficiency.
  2. Older equipment becomes less attractive economically, even if it remains functional.
  3. Rental prices, resale values or utilization disappoint relative to financing assumptions.
  4. Infrastructure operators face pressure from replacement spending and debt obligations.
  5. Customers become more selective about further capital expenditures, potentially slowing demand for new chips.

For Nvidia, the implication is that exceptional current profitability may not translate into equally exceptional growth and margins indefinitely. For CoreWeave and Nebius, the concern is more directly tied to the returns earned on expensive infrastructure and the obligations used to finance it. This is a scenario, not an established outcome. The opposing argument is that older processors can continue earning revenue through inference and less demanding workloads, extending their useful economic lives. Nvidia has pointed to continuing multiyear commitments for older A100 chips as evidence that a new generation does not automatically render the previous one worthless. The disagreement is therefore about how much cash the equipment can generate, for how long and at what replacement cost. That is a more consequential debate than whether AI can write an impressive paragraph.

Why Being Partly Right Does Not End the Bull Case

The selloff strengthens Burry’s valuation warning. It does not, by itself, prove that AI infrastructure demand is collapsing. Nvidia’s latest results still show substantial revenue, earnings and cash generation. CoreWeave continues to report expanding revenue and a large backlog, although its financing costs remain material. Nebius has reported stronger customer prepayment terms, while also acknowledging deployment delays and a significant capital-spending requirement. For many, the constructive opportunity begins with accepting both realities: the earlier share prices may have demanded too much, while the underlying businesses may still be building something valuable. The next test is not whether management teams can announce another ambitious capacity target. It is whether they can turn that capacity into durable cash returns after interest, depreciation, maintenance and replacement spending. Burry has earned credit for identifying vulnerabilities that the market subsequently repriced. The bulls must now earn their answer through execution. Fortunately for serious analysis, the income statement does not award points for either optimism or a particularly well-timed “I told you so.”

The Sources

  1. SEC.gov: Scion Asset Management Form 13F filing details, filed November 3, 2025, for holdings as of September 30, 2025.
  2. SEC.gov: Scion Asset Management complete November 3, 2025, filing, including the holdings information table.
  3. SEC.gov: Frequently Asked Questions About Form 13F, including reporting treatment for put and call options.
  4. Bloomberg: “Burry Reveals Nvidia and Palantir Puts After Bubble Warning,” November 4, 2025.
  5. Forbes: “The Burry Top? Legendary Investor Michael Burry Exits Amid AI Valuation Concerns,” November 16, 2025.
  6. Forbes: “Nvidia’s $500 Billion Bet To Make AI Compute Wall Street’s Next Asset Class,” August 10, 2026.
  7. Forbes: “GPU Debt Has Gone Investment Grade. Here’s Who Holds The Risk,” June 9, 2026.
  8. Forbes: “Nvidia’s Real Risk: Hardware That Ages Too Fast?” December 2, 2025.
  9. Nvidia: Third-quarter fiscal 2027 financial results, November 17, 2026.
  10. CoreWeave: Third-quarter 2026 financial results, November 10, 2026.
  11. SEC.gov: CoreWeave Form 10-Q for the quarter ended September 30, 2026.
  12. SEC.gov: Nebius third-quarter 2026 financial results, Exhibit 99.1.
  13. SEC.gov: Nebius third-quarter 2026 shareholder letter, Exhibit 99.2.
Disclosure: This article is for informational purposes only and should not be construed as investment advice, a recommendation, or an offer to buy or sell any security. Market data are subject to revision, and investors should conduct independent due diligence before making investment decisions. Investments may involve substantial risk, including the potential loss of the entire investment. Investors should conduct independent due diligence and consider their individual objectives and risk tolerance. See The Complete Disclosure via this link & at the top of the page.