Announced 29 Oct 2026 · Sources checked
The warning
Steve Eisman, the portfolio manager known for betting against mortgage debt before the 2008 crisis ('The Big Short'), told CNBC on 29 October that the artificial intelligence trade is near a 'tipping point.' He is not ready to bet against it outright, but he has made his portfolio 'much more conservative' and has shorted against the box a portion of his holdings.
Eisman's primary concern is with the frontier algorithm model makers. He believes they are exploiting a climate of fear around their technologies to set up competitive regulatory moats against cheaper products and to obscure faltering fundamentals. This framing treats AI regulation debates as partly a competitive strategy rather than purely a safety issue.
The Anthropic data
Eisman pointed to Anthropic's recently filed S-1, which left revenue figures for the latest quarter blank. He told CNBC it is 'very clear' that Anthropic's revenue growth has slowed dramatically and the missing numbers would show significant deceleration from the prior quarter. He also flagged a capital efficiency concern: the ratio of revenue to equipment expense fell from 2.08 in the second quarter to 1.7 in the third, a marked change in trajectory.
Anthropic's IPO filing was widely covered when it was published; Eisman's interview adds a skeptical investor's reading of the same documents, centered on unit economics rather than model capability.
The Google echo
Eisman cited recent Google commentary as additional evidence. Google executives said Wednesday they were running experiments to address weaker monetization in AI Mode, with chief business officer Philipp Schindler acknowledging that 'the strong growth in AI mode queries had an impact because of the lower monetization we are seeing in AI mode.' Eisman said 'the numbers from Google scared people.'
This is the business-model tension underneath the AI enthusiasm: users are increasingly getting answers through AI interfaces, but those interfaces generate less advertising revenue per query than traditional search results. If AI adoption continues to shift query volume away from high-monetization formats, the economics of the largest AI-adjacent businesses change.
Why not short it
Despite his skepticism, Eisman said he was scared to make a definitive bet against the AI trade because of how intertwined it is with the broader market. 'So much of the market is on the other side, and it's still reading tea leaves,' he said. He called it 'impossible' for the technology sector to be siloed away from the rest of the equity market and the economy.
That nuance matters: even the highest-profile bearish investor in AI is finding it hard to construct a short against a sector that has carried so much of the market's gains. It suggests the risks are clear but the timing of any repricing is genuinely uncertain.
Caveats
Eisman's views are one investor's interpretation, and the Anthropic revenue figures he cites are not official — the company left the quarter blank in its S-1. His revenue-to-equipment ratio relies on an analysis provided to him, not independently verified here. He explicitly declined to short the sector, so the warning is directional, not a confirmed market call.
The Google comments reflect the company's own framing of a monetization challenge; AI Mode advertising is early and Google's ad experiments have shown promising results, according to Schindler.
The broader investor mood
Eisman's cautious stance arrives at a moment when several high-profile voices across finance and tech are questioning AI's valuation. The 'death and consolidation' warning from Elon Musk about AI neoclouds, Huang's insistence that demand is 'off the charts,' and Eisman's tipping-point language all circle the same uncertainty: how much of the AI buildout is durable value versus speculative debt-fueled expansion.
The difference is that Eisman is primarily focused on the model labs themselves rather than the infrastructure layer. His read on Anthropic is that the company's economics — the ratio of revenue against equipment spend — are deteriorating even as its market profile rises, and that the blank revenue line in the S-1 is the tell. If a top-tier lab like Anthropic is facing unit-economics pressure, the argument goes, the froth may be broader than just the neoclouds.
Regulatory moat argument
Eisman's claim that frontier labs exploit fear to build regulatory moats is a distinctive and testable thesis. The idea: the more the leading labs warn about AI risk, the more policymakers push for oversight and safety regimes — and the harder it becomes for smaller, cheaper open-weight models to compete, since compliance costs scale more heavily for challengers.
If that framing is right, the safety debates that dominate AI policy discussions also serve industrial strategy. The counter-framing, of course, is that a frontier model is genuinely different from an open-weight model in capability and risk, and that regulation should differentiate based on actual risk rather than market position. Both interpretations are live in the policy conversation, and Eisman's interview brings an investor's version of the argument into the mainstream.
Common questions
Who is Steve Eisman?
The investor made famous by 'The Big Short' for betting against mortgage debt before the 2008 financial crisis.
What did he say about AI?
The AI trade is near a 'tipping point'; he's become more conservative and cites Anthropic's blank revenue figures, falling revenue-to-equipment ratios, and Google's AI Mode monetization concerns.
Is he shorting AI stocks?
Not outright — he says the sector is too intertwined with the broader market to make a clean short.
Sources & further reading
How this story was made
Written by Kristian Kostov with AI assistance and checked against the linked sources. Company performance claims are attributed to the company. Analysis reflects AiLookout’s interpretation; we have not independently tested the products discussed. The cover is an AI-generated editorial illustration, not a photograph or evidence of an actual event.
Our editorial standards
