Nvidia and OpenAI said they have struck a partnership where OpenAI will deploy at least 10 gigawatts of AI datacenters built on Nvidia's Vera Rubin GPUs. Nvidia will invest up to $100 billion in OpenAI as each gigawatt is deployed.
In other words, Nvidia is investing in $10 billion a gigawatt deployed. OpenAI will also be the first public reference for Nvidia's Vera Rubin platform when the first phase of the deal comes online in the second half of 2026.
In a press release, the two companies said they have inked a letter of intent for the partnership. OpenAI, which has noted it is chasing superintelligence, will use Nvidia to power the training of its next-gen ChatGPT models.
Nvidia CEO Jensen Huang said the two companies have "pushed each other for a decade, from the first DGX supercomputer to the breakthrough of ChatGPT."
- Nvidia invests $5 billion in Intel, aims to codevelop for data centers, PCs
- Nvidia delivers strong Q2, rides data center demand, but China a wild card
OpenAI CEO Sam Altman said that "everything starts with compute."
Key points of the partnership include:
- OpenAI will use Nvidia as its preferred compute and networking partner for AI factories.
- OpenAI and Nvidia will optimize roadmaps for models, infrastructure and software.
- The partnership will include a bevy of collaborators.
- The companies said they will finalize details of the partnership in the weeks ahead.
Constellation Research analyst Holger Mueller said:
"Good to see OpenAI will build its own data centers, and good to see plans for Nvidia's next platform, Vera Rubin. Someone please add up all the purchase commitments OpenAI has done for Oracle, Microsoft and now own data centers and see if Sam Altman might have to sell his shirt in 2026."
