On Sept. 22 (U.S.), Nvidia and OpenAI announced a landmark strategic partnership that will deploy at least 10 gigawatts of Nvidia hardware across AI datacenters worldwide, backed by up to $100 billion in investments. Nvidia will serve as OpenAI's preferred strategic compute and networking partner, with the first datacenter expected to come online in the second half of 2026.

Under the agreement, OpenAI will construct multiple AI datacenters until the combined compute capacity reaches the 10GW target. Nvidia, in turn, will invest strategically in OpenAI, potentially through share purchases, while the two companies will "co-optimize" their respective roadmaps. OpenAI will focus on AI model and infrastructure software, and Nvidia will prioritize hardware and ecosystem development.

Nvidia CEO Jensen Huang commented, "From the first DGX supercomputers to breakthroughs like ChatGPT, Nvidia and OpenAI have built a decade-long partnership. This collaboration marks a new leap, empowering the next generation of AI with 10GW of compute." OpenAI CEO Sam Altman added, "Everything starts with compute. Compute infrastructure will be the foundation of the future economy. With Nvidia, we aim to deliver AI breakthroughs at scale and empower individuals and enterprises globally."

The partnership will leverage Nvidia's Vera Rubin platform and Rubin Ultra accelerators, featuring up to 76 TB of HBM4 memory, capable of FP4 inference at 3.6 exaflops and FP8 training at 1.2 exaflops. Rubin Ultra is projected to deliver 15 exaflops FP4 inference and 5 exaflops FP4 training, supported by 365 TB of HBM4e across 14 GB300 GPUs. The Rubin GPUs and Vera CPUs taped out in late Aug. (U.S.) and are being manufactured at TSMC.

The new collaboration builds on existing Nvidia-OpenAI ties and complements OpenAI's partnerships with Microsoft, Oracle, SoftBank, and the Stargate Alliance, further reinforcing their shared vision of creating the world's most advanced AI infrastructure. OpenAI currently serves over 700 million weekly active users, spanning enterprises, small businesses, and developers.

The agreement is expected to accelerate next-generation AI model development, support high-performance AI inference and training at scale, and facilitate innovations in AI-powered services across cloud platforms, datacenters, and enterprise solutions. Analysts note that the scale of investment and compute capacity could influence global AI hardware supply chains and set new benchmarks for hyperscale AI deployment.

Following the announcement, Nvidia shares rose nearly 4% in U.S. trading on Sept. 22 (U.S.), reflecting strong investor confidence in the strategic collaboration.