Steve Eisman, the investor who made his name betting against subprime mortgages, is now warning that the biggest names in cloud computing are dangerously dependent on two AI companies. He says OpenAI and Anthropic account for roughly 70% of AI-related revenue at Microsoft, Amazon, Alphabet's Google, and Oracle. That concentration, he argues, leaves the sector exposed to a price war from cheaper Chinese open-source models.
The 70% concentration problem
Eisman's numbers paint a stark picture. OpenAI and Anthropic together drive about 70% of AI-related revenue at those four tech giants. When you look at cloud revenue specifically, the two companies still make up 25% to 35% of the total at Microsoft, Amazon, Google, and Oracle. That means a slowdown at either startup would hit the cloud businesses of the world's largest tech firms directly.
Eisman sees a key threat from Chinese open-source models that cost less and appear to be winning customers. If those models keep taking market share, he says, a price war across the AI sector could follow. That would squeeze margins at the very companies that have poured billions into AI infrastructure.
Why Eisman moved to cash
Eisman isn't just talking. Last month he sold his stake in Google's parent company, Alphabet, and moved the proceeds to cash. The move was designed to reduce his exposure to AI. He didn't say what he might buy next, but the decision to sit on cash suggests he sees more downside than upside in the near term.
His caution stands in contrast to the mood on parts of Wall Street. Tom Lee, a strategist often cited for his bullish calls, reads the fear around AI capital spending as a positive sign. Jim Cramer, meanwhile, claimed that the AI data center trade is regaining market leadership. Neither man addressed Eisman's specific warning about Chinese open-source competition.
Burry's bearish bets
Eisman isn't the only famous investor positioning for trouble. Michael Burry, who also gained fame for his subprime short, is now short on the iShares Semiconductor ETF (SOXX), Micron, Nvidia, Caterpillar, Palantir, Tesla, and Applied Materials. That's a broad bet against some of the most heavily traded names in the AI and tech complex.
Burry has also forecast that US stocks could suffer a 1987-type crash. He hasn't said when, but the short positions suggest he's not waiting for confirmation. The list includes both pure-play chipmakers and companies like Caterpillar, which is more tied to industrial demand than to AI directly.
What the bulls are saying
Tom Lee's view is that the market's fear over AI capital expenditure is actually bullish. He argues that the spending is a sign of confidence, not a bubble about to burst. Jim Cramer's claim that the AI data center trade is regaining leadership echoes that sentiment. Both are betting that the infrastructure buildout continues to pay off.
Eisman's counter is that the revenue base is too narrow. If Chinese open-source models undercut the incumbents on price, the big cloud providers could be forced to cut their own prices to compete. That would hit the very revenue streams that justify today's valuations.
The question now is whether the price war Eisman describes actually materializes. Chinese open-source models have already shown they can match performance at a fraction of the cost. The next few quarters will show whether customers are willing to switch.
