GMO's Inker Warns AI Stock Supply and Compute Glut Could Trigger 35-50% Market Drop
Ben Inker, co-head of asset allocation at Boston-based investment firm GMO, discussed his firm's third-quarter research report "A Catalyst for the AI Bubble Break" in a podcast interview on October 15. He argued that over the next 12 to 18 months, IPOs and lock-up expirations from SpaceX, Anthropic, and OpenAI could inject new equity supply equivalent to 6% to 7.5% of total U.S. stock market value—far exceeding the 2.5% to 4% seen during the dot-com bubble.
Citing historical patterns, Inker said that for every 1% increase in IPO supply as a share of market capitalization, annual stock returns tend to lag normal years by about 4 percentage points. If new supply reaches 6%, returns could trail by 24 points. He also highlighted that even with SpaceX's phased unlock structure, roughly $2 trillion in shares could become eligible for sale over the next 12 months, putting significant pressure on secondary-market liquidity.
On computing capacity, Inker believes AI compute is shifting from structural shortage to oversupply. With about half of current compute spent on training new models, a reduction in that share would release a flood of idle capacity, potentially leading to credit defaults, bankruptcies, and deteriorating returns for hyperscale cloud providers. In that scenario, he suggested the market could fall 35% to 50%, not just 18%. He pointed to CoreWeave's bonds falling below 80 cents and declines in Asian memory chip stocks as early signs of a crisis of confidence in AI cash flows.
Inker distinguished between companies' vulnerabilities: Microsoft's data center returns may disappoint, but it will not default; CoreWeave and Super Micro might not be so fortunate. He doubted governments can act as lenders of last resort, but noted the Fed still has room to cut rates. He also argued that if losses remain concentrated in equity markets rather than the banking system, a major drawdown could occur without systemic collapse. GMO's approach, he said, is to build portfolios with reasonable valuations and AI upside rather than making a pure bet on AI.
