Steve Eisman argues the artificial intelligence trade rests on a dangerous concentration of demand, a claim complicated by the massive scale of current hyperscaler spending. The former short-seller, known for betting against subprime mortgages, stated on his September 21 podcast that 70% of hyperscaler AI revenue comes from just two companies: OpenAI and Anthropic. He identified OpenAI as the potential single point of failure for the entire cycle, noting that these two entities represent roughly 25% to 35% of total cloud revenue for major providers.
The Concentration Risk
The risk is that the supply chain, led by NVIDIA and Microsoft, is built on the assumption that these two labs will continue to consume compute at current rates. NVIDIA, with a market cap of $5.52 trillion, reported fiscal Q2 revenue of $96.22 billion, up 105.8% year over year. Its CEO Jensen Huang described compute as "supply-constrained" and guided Q3 revenue to $108.0 billion. To support this, NVIDIA has made $279 billion in supply commitments, including $108.5 billion in guarantee obligations to AI cloud and data center partners. Microsoft, valued at $3.68 trillion, is the other critical node. It spent $115.95 billion on capital expenditures in fiscal 2026, and its commercial remaining performance obligations hit $678 billion. This spending intensity has already impacted cash flow, with Q4 free cash flow falling 23% year over year to $19.64 billion.
The Counterargument
The counterargument is that OpenAI is not a single point of failure but a diversified platform with strong margins. Six days before Eisman's warning, OpenAI CFO Sarah Friar told CNBC that the business consists of a "diversified set of revenue streams" with a diversified chip supply chain. Furthermore, CNBC reported that investors have approached OpenAI about a new funding round at a valuation as high as $1.5 trillion. Microsoft’s restructured deal with OpenAI, which gave it a stake valued around $135 billion in exchange for OpenAI contracting $250 billion in incremental Azure services, suggests institutional confidence in the lab’s ability to generate the necessary cash flow. Eisman offered no source for his 70% figure, leaving the precise degree of concentration unverified by public filings.
On balance, the market is already questioning the return on investment. Microsoft shares are down 3.62% over the past year, while NVIDIA is up 23% year to date. The line to watch is whether OpenAI’s next funding round prices near the $1.5 trillion mark and whether Microsoft’s late October earnings show Azure growth holding near the 45% constant-currency guide without another decline in free cash flow. If OpenAI cannot generate the cash to consume the $250 billion in promised Azure services, Microsoft’s capital expenditure looks less like a bet on a secular boom and more like credit exposure to two loss-making labs.