Jensen Huang Admitted He Was Facing an "Avalanche" of Excess Wafers and Memory

A report describes a sharp internal dispute at Nvidia in August 2026 over the company’s financial commitments to AI customers and its growing stockpile of chip supplies. According to the account, CEO Jensen Huang said major cloud providers were shifting toward custom chips and reducing dependence on Nvidia, while prepaid TSMC wafers and memory were arriving in quantities he likened to an “avalanche.” CFO Colette Kress worried that Nvidia was taking on too much risk by helping cloud providers and AI firms finance purchases of its products. She reportedly reduced the proposed cap on guarantees for OpenAI computing capacity from about 8 gigawatts, valued at up to $250 billion, to 4.25 gigawatts, or up to $125 billion. Huang allegedly rebuked her for challenging the arrangements.

The account presents these remarks as evidence that the GPU and memory crunch may reflect overcommitment and inflated expectations, not simply AI demand consuming all available supply. Nvidia reportedly approached TSMC about winding down chip orders, while the company’s financing commitments were intended to keep customers buying its GPUs and absorb excess inventory. It also cites an estimate that roughly $200 billion in Nvidia AI chips may be sitting unused in warehouses because cloud firms lack sufficient power infrastructure or immediate demand. The broader argument is that aggressive AI investment and fear of missing out led companies to order capacity ahead of practical deployment, leaving Nvidia exposed to excess silicon and financial risk.