Nvidia’s ambitious plan to pour up to $100 billion into OpenAI has stalled, casting fresh uncertainty over one of the largest proposed bets in AI infrastructure to date.
The nonbinding agreement was meant to finance roughly 10 GW of data center capacity over several years, powering OpenAI’s next generation of frontier models on Nvidia GPUs.
But internal doubts at Nvidia and rising competitive pressures in AI compute are now forcing both sides back to the drawing board.
What Changed Inside Nvidia
According to reports, some Nvidia executives questioned the sheer scale, capital intensity, and concentration risk of a $100 billion commitment to a single AI customer, even one as strategically important as OpenAI.
Over the next eight years, OpenAI has identified infrastructure requirements totaling $1.4 trillion, or roughly 28 gigawatts (GW) of power capacity.
This is a startling sum for a company with a relatively short operating history.
Implications For OpenAI And AI Infrastructure
For OpenAI, the pause complicates funding plans for the multiyear “Stargate” build-out, a program expected to require around $500 billion in total data center investment across partners like Nvidia and Oracle.
However, OpenAI is already pursuing a massive equity round at a reported $500 billion valuation and remains a magnet for Big Tech and institutional capital.
Future Outlook
The stalled megadeal highlights a maturing phase in the AI boom where capital discipline starts to catch up with exuberant infrastructure plans.
Nvidia is signaling that even in a market where its GPUs are effectively sold out, it prefers a portfolio of large, flexible partnerships rather than a single $100 billion bet.
OpenAI, meanwhile, is likely to secure substantial funding but may have to piece together its data center roadmap across more partners, technologies and geographies.
