“Big Short” investor Steve Eisman has issued a stark warning about the structural fragility sitting at the heart of the AI investment boom.
Eisman, speaking on CNBC’s “Fast Money” on August 13, said OpenAI and Anthropic account for roughly 70% of AI-related revenue at Microsoft, Amazon, Google, and Oracle.
He added that the two labs represent as much as 25% to 35% of those companies’ total cloud revenue, a level of dependency he described as alarming.
“The futures of these massive companies, in a sense, are a bet that OpenAI and Anthropic are going to succeed,” Eisman said, calling the dynamic the “Achilles’ heel” of the AI trade.
He warned that if either lab stumbles, cheaper Chinese open-weight models could ignite a price war that ripples across the entire industry.
Eisman said he is not yet shorting the trade, telling viewers he wants financial data once the labs go public before taking a position.
NVIDIA Corporation (NASDAQ: NVDA) and Palantir Technologies Inc. (NASDAQ: PLTR) sit on opposite ends of that concentration risk, representing two distinct bets on how the AI landscape resolves.
Nvidia supplies the chips that every AI lab requires regardless of which platform ultimately dominates, giving it an infrastructure-layer position that is insulated from the labs’ individual fortunes.
Palantir, by contrast, is positioning itself as the solution for enterprises that want to reduce their dependence on any single AI lab, a pitch that has resonated strongly with customers.
Palantir’s most recent quarterly results support that narrative, with revenue surging 93% to $1.94 billion, beating analyst expectations of $1.80 billion by a meaningful margin.
U.S. commercial revenue grew 149% to $764 million, and Palantir raised its full-year guidance to between $8.15 billion and $8.158 billion, which the company described as its largest-ever guidance raise.
CEO Alex Karp captured the company’s positioning bluntly, writing that “our customers have declined to become vassal states of the language labs,” a line that underscores Palantir’s sovereignty pitch to enterprise clients.
Karp also joined other technology leaders in urging the government not to restrict open-weight AI models, a stance that would widen customer choice well beyond the two labs Eisman flagged.
Nvidia’s bull case rests on the fact that every hyperscaler, including OpenAI, Anthropic, and any future challenger, still depends on its hardware, meaning Eisman’s concern targets the labs’ finances rather than Nvidia’s chip business directly.
Nvidia has also moved to spread its financial exposure, signing deals with Apollo, Blackstone, BlackRock, Brookfield, Goldman Sachs, and KKR to mobilize over $500 billion in third-party AI financing, with CEO Jensen Huang capping Nvidia’s own exposure at roughly 25% per project.
The company generated close to $48 billion in free cash flow last quarter, giving it a substantial buffer to absorb any slowdown among individual customers.
The bear case, however, is more pointed, with Michael Burry arguing that Nvidia’s reliance on a small group of hyperscale customers including Microsoft, Meta, Amazon, and Alphabet creates its own version of Eisman’s concentration problem.
A Bloomberg analysis cited by Burry traces roughly $46 billion in equity stakes and $879 billion in purchase commitments linking Microsoft, Oracle, Amazon, Google, Meta, OpenAI, Anthropic, xAI, CoreWeave, Nvidia, and AMD, including a $300 billion Oracle-OpenAI commitment.
Nvidia’s five-year credit default swap spread has roughly doubled over two months, a market signal that reflects growing unease about those entanglements.
The connection between Eisman’s warning and Nvidia’s outlook is direct: a Chinese-driven price war hammering OpenAI and Anthropic’s economics could push hyperscalers to pull back the capital expenditure that funds Nvidia chip purchases.
Palantir carries its own risks, with U.S. government revenue hitting $809 million last quarter, nearly matching commercial revenue and tying a large portion of results to federal budget decisions.
Karp himself acknowledged the challenge, describing his 18-month growth target as “a very high goal,” a phrase that signals even management recognizes the difficulty of sustaining the company’s current growth trajectory.
Nvidia attempted to address the vendor financing concerns directly, sending analysts a memo denying the practice, first reported by Barron’s, though Burry and other short sellers have publicly rejected that explanation.
Hedge fund interest in both companies continues to climb, with Nvidia held by 275 funds as of Q1 2026, up from 264, while Palantir was held by 96 funds, up from 89.
