Steve Eisman, the investor known for his bet against subprime mortgages before 2008, says OpenAI and Anthropic together generate roughly 70% of the AI-related revenue flowing to major hyperscalers. He calls that concentration the AI trade's "Achilles' heel," and warns cheaper Chinese open-weight models could squeeze margins further.
The investor who called the 2008 housing crash now sees a different kind of dependency risk. Speaking on CNBC's "Fast Money," Eisman argued that Big Tech's AI revenue engine rests on just two private companies rather than a broad customer base.
Two labs, one shared dependency
OpenAI and Anthropic together account for roughly 70% of AI-related revenue flowing to Microsoft, Amazon, Alphabet, and Oracle, Eisman said. In his view, that concentration is the "Achilles' heel" of the AI trade.
That dependency runs deep into the balance sheet. AI-driven revenue represents approximately 25-35% of these companies' total cloud revenue, a figure that has underpinned premium valuations across the sector for years. Microsoft, Amazon, Google, and Oracle have poured hundreds of billions into AI infrastructure, yet the payoff hinges on continued spending by OpenAI and Anthropic specifically.
Chinese open-weight models add a new variable
Eisman also flagged cheaper Chinese open-weight AI models as a threat to pricing power across the industry. These models publish their parameters for others to build on, and Chinese labs have released competitive versions at significantly lower price points than the proprietary systems from OpenAI and Anthropic.
If those models reach a quality level where enterprise customers start switching, or use them as leverage in negotiations, margins could compress sharply. Hyperscalers have justified enormous capital spending on the assumption that AI workloads generate healthy long-term returns, and a price war would undercut that assumption.
Part of a broader pattern
Eisman's warning follows earlier comments in July 2026, when he flagged potential diminishing returns in the AI sector and disclosed that he had personally reduced his exposure to AI-related investments. His skepticism sits within a debate that has intensified through 2026, as tech companies keep announcing larger AI infrastructure budgets each quarter.
Source: Crypto Briefing
Trading involves risk.
Risk Disclaimer
