Renowned bearish investor Michael Burry has stated that Anthropic's IPO prospectus has not changed his negative stance on the AI trade, and that even as technology advances, it cannot resolve his concerns about financing, infrastructure spending, and growth, according to a report.
This stands in sharp contrast to retail enthusiasm. On Stocktwits, retail sentiment around Anthropic (NASDAQ: ANT) is "extremely bullish" with "extremely high" message volume, while the broader market looks decidedly darker: sentiment for the SPDR S&P 500 ETF (SPY) and the SPDR Dow Jones Industrial Average ETF (DIA) is both "bearish" with "normal" message volume, and the Invesco QQQ Trust (QQQ) is "extremely bearish" with "high" message volume.
Good Technology Is Irrelevant to the Short Thesis
Burry made these remarks while responding to the prospectus in his Substack chat. "There is a lot here, but nothing changes my view," he said. "What is happening goes far beyond what this S-1 can cover."
The comments stemmed from a reader's question about whether advances in large language models (LLMs) had changed his view on the technology. The reader wrote that since 2024 and 2025 everyone has said LLMs would hit a wall or reach a plateau, "but they really haven't, and they are still improving exponentially," with the narrative remaining strong and continuing to fuel more building and investment.
"In short, my view is this: the technology is advancing, and that has nothing to do with any short thesis," Burry replied.
He then laid out the financial and industrial risks underpinning his position: "I look at the financing, the insurance shell leverage within it, the latest chips being unable to get power and thereby propping up demand for older chips, the DRAM capacity converted for HBM that has not disappeared and will come back with a vengeance, compression and small language model (SLM) diffusion on open-weight models, and the fundamentals behind not just the companies but the deals they are doing."
He also pointed to a mismatch between the tenor of the financing and the projects being financed, then returned to the distinction between technological progress and growth: "I never said LLMs would stop improving. I think they are already being used more efficiently, and that trend will continue — which is bad for their growth rates."
Understanding Burry's View
Burry's list of risks can be broken down into five points, each pointing to the same judgment: the pressure is not on model capability but on the funding and hardware structure supporting this buildout.
First, financing and "insurance shell leverage." The original text does not elaborate on the latter, but taken literally it points to leverage layered through insurance funds or special purpose vehicles in AI infrastructure financing. Burry places it alongside overall financing, indicating that his focus is not on any single company's balance sheet but on the way money enters the industry.
Second, chips that cannot get power. Because the latest generation of chips cannot be deployed due to power, data center, and grid connection constraints, demand for the previous generation is being "floated" — this both inflates the book demand for older chips and means the real ramp-up pace of new chips may be overestimated.
Third, the retaliatory return of memory. The DRAM capacity he converted for high-bandwidth memory (HBM) has not disappeared and will flow back once the demand structure changes — this directly hedges against the recent debate over when the memory chain's "price increase cycle" will peak.
Fourth, model compression and the spread of small models in the open-weight ecosystem point to a decline in cost per unit of compute, which in turn suppresses the pricing power of frontier models.
Fifth, the fundamentals of not just the companies but the deals themselves — that is, the mutually tied orders and investments among cloud providers, chipmakers, and model companies, whose quality needs to be examined separately.
What Burry Thinks Is Missing From the Prospectus
Regarding Anthropic itself, Burry said the prospectus does not reflect recent changes in the developer market. "Between the end of the second quarter and the third quarter, a lot changed in the developer landscape, and that is not in this prospectus," he said.
He also questioned how the document handles the retreat of "tokenmaxxing," the practice of maximizing AI token consumption. "The overall slowdown in growth has not been attributed to a decline in tokenmaxxing, but I did not expect it to mention that either," Burry said. "The Information reported that after tokenmaxxing was restricted, Meta's Claude Code users fell from about 60,000 to 30,000."
This contrasts with the prospectus's own statements. Anthropic explicitly warned in the filing: "We cannot assure you that (revenue) will continue to grow," while also saying it expects to achieve adjusted operating profit for a second consecutive quarter.
The Prospectus Itself: High Growth, Massive Non-Cash Losses, Huge Compute Commitments
Anthropic submitted its S-1 registration statement to the U.S. Securities and Exchange Commission on Monday, October 26, seeking to list on the Nasdaq Global Select Market under the ticker "ANT." The filing did not set the number of shares or a price range, with Goldman Sachs and Morgan Stanley leading 24 underwriters.
On the financial side, revenue in the first half of 2026 was $16.286 billion, compared with $1.163 billion a year earlier; revenue from 2023 to 2025 was $41 million, $386 million, and $4.585 billion, respectively. The net loss over the same period was $80.748 billion, of which about $72.388 billion came from fair value changes in financial instruments such as convertible notes and derivative liabilities, representing non-cash accounting losses. Operating loss narrowed to $693 million from $2.135 billion a year earlier, adjusted operating profit was $1.62 billion, and net cash flow from operating activities turned positive at about $3.9 billion. As of June 30, cash, cash equivalents, and short-term investments totaled about $131.65 billion. Second-quarter revenue alone was $11.56 billion, and the company recorded its first quarterly operating profit of $974 million.
The cost side is what matters most. Compute and infrastructure spending in the first half was about $8.945 billion, compared with about $2.3 billion a year earlier. The company has committed to future spending of about $518 billion on cloud services and compute, including an agreement with Amazon extending to April 2036 worth at least $110 billion and an agreement with Google through July 2033 worth at least $114.5 billion. On customer structure, in the 12 months through September 30, there were more than 11,900 customers paying over $100,000 and more than 1,900 paying over $1 million; net revenue retention was about 900%. Concentration is also striking: two unnamed customers together contributed about 23% of revenue in the first half.
On valuation, some investors have floated a valuation as high as $2 trillion, while others favor about $1.5 trillion; if it exceeds $2 trillion, it would surpass SpaceX's roughly $1.77 trillion listing in June of this year to become the largest IPO ever.
The prospectus devotes substantial space to risk warnings, including that advanced AI could pose "catastrophic or existential risks" to humanity and that liability from extreme harm could exceed the coverage of any insurance. Governance structure also merits attention: public investors can buy shares but cannot control the company — a single Class F share held by seven co-founders carries 50.1% of shareholder voting power, and a trust selects a majority of the board; a public-interest fund will hold a 5% equity stake in the company through non-voting shares, and each co-founder has committed to donating 80% of personal holdings to philanthropic causes. In addition, Amazon holds a $6 billion note convertible into non-voting stock, capped at 12% of the company — financing arrangements and equity structure are two sides of the same coin here.
This is exactly where Burry's "nothing to do with technology" lands: revenue has multiplied more than tenfold in just over a year and the company can already post quarterly operating profit, but what supports it are future compute commitments of more than $500 billion, deeply tied procurement agreements with major cloud providers, and a valuation still awaiting market pricing. Put another way, this prospectus places the AI trade's disagreements on the same table: on one side are the revenue curve, customer retention, and the first appearance of quarterly operating profit; on the other are committed spending, customer concentration, and financing structure. Burry is betting on the latter — he has never denied that models are getting better; he believes that getting better is itself suppressing the growth rate of this business.
Burry's Historical Stance
These remarks are a continuation of his existing views. Previously on X, Burry said Anthropic's valuation was enough to buy 78 profitable S&P 500 companies, including Domino's, Clorox, and Lululemon. He also wrote, "For humanity's sake, the market should fall hard and stop the OpenAI and Anthropic IPOs." In September of this year, he said he had moved up his timeline for an AI downturn and replaced direct short positions with put options. "If spending stops or slows, everything falls apart," he said at the time. Anthropic raised $65 billion in May at a valuation of $965 billion. Burry once compared that valuation with the largest pre-IPO valuation between 1990 and 2000 on an inflation-adjusted basis — United Parcel Service's $119 billion — noting that UPS was then 92 years old, traded at 26 times earnings, had a net margin of about 8.6%, and a price-to-sales ratio of 2.4 times.
Comments