Nvidia invests $100B in OpenAI in circular deal critics are calling the 'Lucent moment' of AI
Sep 23, 2025
Key Points
- Nvidia invests $100 billion in OpenAI equity tied to GPU purchases, effectively pricing chips at 30% equity rather than Nvidia's typical 50-60% gross margins.
- The circular deal structure mirrors Lucent's failed vendor financing to WinStar in 1998, though OpenAI frames compute constraints as blocking life-saving applications and argues revenue growth will justify the valuation.
- Success hinges on whether ChatGPT's growth can sustain a multi-hundred-billion-dollar valuation and unlock new use cases at scale, mirroring ASML's 2012 customer co-investment that proved transformative.
Summary
Nvidia is investing $100 billion in OpenAI to build out AI computing infrastructure over several years. The structure ties investment to GPU purchases: for every $35 billion in GPUs that OpenAI buys from Nvidia, Nvidia invests $10 billion in equity. This prices Nvidia's GPU sales at roughly 30% equity, higher than Nvidia's typical 50-60% gross margins.
The deal draws comparisons to circular vendor financing from the dot-com era. Lucent extended $2 billion in vendor financing to WinStar Communications in 1998 to build a wireless network on Lucent hardware. WinStar went bankrupt and Lucent took a $244 million loss. Not all circular deals fail. In 2012, ASML's customers—Intel, TSMC, and Samsung—co-invested $6.8 billion in the chipmaker's R&D and equity to fund extreme ultraviolet lithography development. That bet succeeded spectacularly: ASML traded near $70 per share then and above $900 now.
The outcome hinges on whether the underlying technology scales and generates returns. Sam Altman rests his confidence in OpenAI's growth curve on three points. Stripe grew by taking a small percentage of e-commerce revenue across the internet without hitting a ceiling, and ChatGPT can similarly draw from search, knowledge services, and e-commerce infrastructure. The internet economy has proven resilient across recessions and pandemics. Distribution advantages compound when built on an already-installed base as massive as the internet, making adoption much faster than building new infrastructure from scratch.
OpenAI says it needs this capacity to meet demand it cannot currently satisfy. With limited compute, the company would have to choose between using AI to cure cancer or providing free tutoring to every student on Earth. Altman argues neither tradeoff should be necessary.
OpenAI's stated goal is to build a factory producing one gigawatt of new infrastructure per week. Altman's essay signals further innovation across chips, power, building, and robotics, and hints at "new interesting ideas" for financing the ongoing buildout to be revealed later in the year. He emphasizes speed: other countries are building chip fabs and energy production faster than the U.S., and OpenAI wants to help reverse that.
Nvidia popped on the announcement but is down 3% as of broadcast. Analysts and social media observers have flagged the circularity risk. Nvidia is essentially printing equity to fund the GPUs it manufactures and OpenAI buys. Whether this becomes a failure or a success depends entirely on whether ChatGPT's revenue growth can sustain a multi-hundred-billion-dollar valuation and whether the compute actually unlocks new use cases at scale.