Burry: Anthropic Prospectus Doesn't Change His Bearish AI Trade Thesis
nashnova research
"Big Short" investor Michael Burry says Anthropic's IPO prospectus has not changed his bearish view on AI trades, listing five specific risks — his core logic targets not whether AI technology is stalling, but whether the funding structure and hardware ramp are overestimated.
What exactly is Burry shorting?
He stated plainly: "Technology is advancing, and that has nothing to do with any short thesis."
This means → he is not betting against AI's usefulness. He is betting against the funding chains and hardware spending built around AI.
In plain terms = whether the tech works is one question; whether the money poured into it can be recouped is another — Burry is wagering the latter breaks down.
Where do the five risks point?
Risk one: the financing structure contains "insurance-shell leverage" — his focus is how money flows into AI, not any single company's balance sheet.
Risk two: the latest-generation chips cannot ramp because power and data-center capacity lag behind, which props up demand for older chips. The market may be overestimating new-chip shipment timelines.
Risk three: DRAM capacity retooled for HBM — high-bandwidth memory, a specialized chip memory that speeds up data reads — has not disappeared. If demand shifts, that capacity snaps back, directly offsetting the storage-chain price-hike narrative.
Risk four: model compression and small-model proliferation in the open-weight ecosystem are pushing down per-unit compute costs, eroding frontier models' pricing power.
Why does the fifth risk deserve its own section?
Risk five: cloud vendors, chipmakers, and model companies hold intertwined orders and investments. The fundamental quality of each deal needs separate scrutiny.
This means → headline "mega-orders" may be related parties endorsing each other, not genuine end-user demand.
Burry also flags a maturity mismatch between financing terms and project timelines — in plain terms = the debt is short, the projects are long, and any gap in between can trigger a blowup.
What does the prospectus itself leave out?
Burry noted: "A lot changed in the developer landscape in late Q2 and Q3, and none of that is in this filing."
He questioned how the prospectus handles the retreat of "tokenmaxxing" — maximizing AI token consumption. Per The Information, after restrictions were imposed, Meta's Claude Code users dropped from roughly 60,000 to 30,000, yet the prospectus does not attribute the growth slowdown to this.
Anthropic itself wrote in the filing: "We cannot assure you that [revenue] will continue to grow."
What do Anthropic's financials actually say?
H1 2026 revenue: $16.286 billion. Net loss in the same period: $80.748 billion, of which roughly $72.388 billion was non-cash fair-value adjustment losses.
The company has committed approximately $518 billion in future cloud and compute spending. Its Amazon agreement alone extends to April 2036 and is worth at least $110 billion.
This means → even stripping out non-cash items, the burn rate and locked-in long-term commitments are staggering — exactly the structural pressure Burry calls "far beyond what this S-1 can cover."
How wide is the gap between Burry's position and market sentiment?
Burry has pulled his AI-downturn timeline forward, switching to put options instead of outright shorts. He says: "If spending stops or slows, everything falls apart."
He has publicly noted Anthropic's valuation could buy 78 profitable S&P 500 companies, writing that "for the good of humanity, the market should crash hard enough to stop the OpenAI and Anthropic IPOs."
This reflects an extreme divergence: retail sentiment on Stocktwits rates Anthropic "extremely bullish", while Burry remains firmly bearish — the ultimate test will be post-listing revenue growth and the pace at which compute commitments are honored.
市场有风险，内容仅供研究参考，不构成投资建议。