Nvidia and OpenAI are locking arms on one of the most ambitious AI infrastructure buildouts ever attempted. The partnership, formalized through a letter of intent dated September 22, 2025, calls for OpenAI to deploy at least 10 gigawatts of Nvidia systems, representing millions of GPUs, to power the next generation of AI models.
What the deal actually looks like
The first phase of infrastructure is targeted to go live in the second half of 2026. It will be built on Nvidia’s Vera Rubin platform, the company’s next-generation chip architecture designed specifically for the kind of massive-scale AI workloads that OpenAI runs.
Nvidia is prepared to back the effort with up to $100 billion in investment, tied to phased milestones as those systems roll out. The investment structure is milestone-based, meaning Nvidia doesn’t write a single check. Instead, capital flows as systems are actually deployed and operational benchmarks are hit.
Why this matters beyond the AI sector
