NVIDIA invests up to $100B in OpenAI; deploying 10 gigawatts of Vera Rubin GPU infrastructure
NVIDIA and OpenAI announced a landmark infrastructure partnership where NVIDIA will invest up to $100 billion in OpenAI progressively as each gigawatt of compute is deployed. The partnership targets at least 10 gigawatts of NVIDIA Vera Rubin GPU systems, with the first gigawatt coming online in H2 2026. OpenAI will serve as NVIDIA's preferred strategic compute and networking partner, co-optimizing roadmaps between software and hardware.
The deployment represents the largest AI infrastructure project announced to date. OpenAI currently serves over 700 million weekly active users and faces training and inference demands from next-generation models supporting agentic AI and reasoning. NVIDIA CEO Jensen Huang positioned this as deploying "a billion times more computational power" than the initial DGX system shipped to OpenAI in 2016, underscoring the scale progression from early lab work to production inference.
For architects building AI infrastructure: the NVIDIA-OpenAI partnership crystallizes a trend of billion-dollar, multi-year vendor lock-in. The co-design model means OpenAI's entire roadmap is now optimized for NVIDIA's hardware stack (Vera Rubin GPUs, Grace CPUs, Spectrum-X networking). This mirrors Meta's recent full-stack Blackwell/Rubin commitment. Strategic implications: vendor diversity becomes harder; cost per token depends on infrastructure architecture, not just model weights.
Sources
- Primary source
- openai.com
“NVIDIA intends to invest up to $100 billion in OpenAI progressively as each gigawatt is deployed”
- blogs.nvidia.com
“This is the biggest AI infrastructure project in history. This partnership is about building an AI infrastructure that enables AI to go from the labs into the world.”
- openai.com
“OpenAI will work with NVIDIA as a preferred strategic compute and networking partner for its AI factory growth plans”