Nvidia and OpenAI’s Fates Get Even More Intertwined

Larry Ellison might have thought his firm’s 4.5 gigawatt cloud deal with OpenAI was unmatched.

That is, until he woke up on Monday morning to see OpenAI CEO Sam Altman and Nvidia CEO Jensen Huang announcing something much bigger: data center plans that are more than double the size of Oracle’s. 

The numbers are quite staggering. If OpenAI and Nvidia follow through on the deal—which, to be clear, is currently only outlined in a letter of intent—Nvidia would invest up to $100 billion in cash in OpenAI to help the startup build up to 10 gigawatts of data center capacity. The important part is it would allow OpenAI to own graphics processing units and put them in its own data centers, according to someone with knowledge of the deal.

That would help it shift away from primarily renting its GPUs from cloud providers such as Oracle, Microsoft, Google or CoreWeave. OpenAI executives have previously talked about their ambitions to build their own data centers. To date, Nvidia hasn’t gotten this directly involved in a firm’s data center build out.