OpenAI and NVIDIA have unveiled a historic partnership to scale AI infrastructure on an unprecedented level, with multi-gigawatt data centers powered by millions of NVIDIA GPUs.
The announcement was made in an interview with CNBC’s Jon Fortt, featuring NVIDIA founder and CEO Jensen Huang, OpenAI CEO Sam Altman, and OpenAI president Greg Brockman.
“This is the biggest AI infrastructure project in history,” Huang said. “This partnership is about building an AI backbone that enables artificial intelligence to move from research labs into the real world.”
Scaling to Multi-Gigawatt Infrastructure
As part of the agreement, OpenAI will deploy at least 10 gigawatts of NVIDIA systems, including the NVIDIA Vera Rubin platform, to support its next-generation AI infrastructure. NVIDIA also plans to invest up to $100 billion in OpenAI, tied to each gigawatt milestone.
“There’s no partner but NVIDIA that can do this at this kind of scale, at this kind of speed,” Altman said. “The million-GPU factories created through this partnership will give us the capacity to train and serve the next frontier of AI models.”
Altman emphasized the central role of compute in OpenAI’s ambitions:
“Building this infrastructure is critical to everything we want to do. This is the fuel we need to drive improvement, build better models, grow revenue — drive everything.”
Meeting Surging AI Demand
Since launching ChatGPT in 2022 — the fastest-growing app in history to hit 100 million users — OpenAI has expanded to more than 700 million weekly active users. Its models now power agentic AI, advanced reasoning, multimodal inputs, and long-context capabilities.
To sustain that growth, OpenAI’s compute needs extend beyond training to the inference workloads required to serve billions of global users.
“The cost per unit of intelligence will keep falling, and that’s great,” Altman said. “But the frontier of AI capability is rising fast, unlocking more uses every day. Without enough compute, society would have to choose between use cases — like developing a cancer cure or providing free education. No one wants to make that choice. The answer is more capacity.”
The first gigawatt of NVIDIA Vera Rubin systems is expected to begin producing outputs in the second half of 2026.
From First Server to Billion-Fold Scale
The collaboration between the two companies dates back nearly a decade. Huang personally delivered OpenAI’s first NVIDIA DGX system in 2016.
“This represents a billion times more computational power than that initial server,” Brockman said. “It’s what will allow us to make real breakthroughs, to create new models, and to empower every individual and business by scaling to the next level.”
The Beginning of Global AI Buildout
Huang underscored that this partnership is just the start of a worldwide AI infrastructure revolution.
“We’re literally going to connect intelligence to every application, every use case, every device,” Huang said. “And we’re just at the beginning. This is the first 10 gigawatts — I assure you of that.”
Watch the full CNBC interview below.
