NVIDIA to invest up to $100 billion in OpenAI for 10 gigawatts of AI infrastructure

NVIDIA intends to invest up to $100 billion in OpenAI progressively as the two companies deploy at least 10 gigawatts of NVIDIA systems — representing millions of GPUs — for OpenAI's next-generation AI infrastructure. The companies announced the letter of intent on September 22, 2025, from San Francisco and Santa Clara, calling it a landmark strategic partnership to train and run OpenAI's next generation of models on what NVIDIA and OpenAI describe as the path to deploying superintelligence.

The scale is difficult to overstate. Ten gigawatts of dedicated compute is the kind of power draw associated with entire cities of industrial infrastructure, and NVIDIA's commitment of up to $100 billion, disbursed as each gigawatt comes online, would rank among the largest corporate investments tied to a single AI company. The first phase is targeted for the second half of 2026 and will run on NVIDIA's Vera Rubin platform, the successor generation of NVIDIA's datacenter systems.

The stakes for the AI market are straightforward: whoever controls the compute pipeline controls the pace of frontier model development. This deal ties OpenAI's infrastructure growth directly to NVIDIA silicon at a moment when compute capacity, power availability, and datacenter construction have become the binding constraints on the industry.

What did the executives say?

NVIDIA founder and CEO Jensen Huang framed the deal as the continuation of a decade-long working relationship between the two companies.

"NVIDIA and OpenAI have pushed each other for a decade, from the first DGX supercomputer to the breakthrough of ChatGPT," Huang said. "This investment and infrastructure partnership mark the next leap forward—deploying 10 gigawatts to power the next era of intelligence."

OpenAI co-founder and CEO Sam Altman emphasized compute as the foundational resource for future economic activity.

"Everything starts with compute," Altman said. "Compute infrastructure will be the basis for the economy of the future, and we will utilize what we're building with NVIDIA to both create new AI breakthroughs and empower people and businesses with them at scale."

OpenAI co-founder and President Greg Brockman pointed to the companies' shared history and the reach of the systems already built on NVIDIA's platform.

"We've been working closely with NVIDIA since the early days of OpenAI," Brockman said. "We've utilized their platform to create AI systems that hundreds of millions of people use every day. We're excited to deploy 10 gigawatts of compute with NVIDIA to push back the frontier of intelligence and scale the benefits of this technology to everyone."

What does the partnership actually cover?

The letter of intent establishes several concrete commitments and working arrangements:

• Deployment scale: at least 10 gigawatts of NVIDIA systems, representing millions of GPUs, for OpenAI's next-generation AI infrastructure.

• Investment structure: NVIDIA intends to invest up to $100 billion in OpenAI progressively, tied to the deployment of each gigawatt of capacity.

• Timeline: the first gigawatt of NVIDIA systems is targeted to come online in the second half of 2026 on the Vera Rubin platform.

• Scope of buildout: the deployment includes datacenter and power capacity, not just chips.

• Partnership role: OpenAI will work with NVIDIA as a preferred strategic compute and networking partner for its AI factory growth plans.

• Co-optimization: the companies will jointly align their roadmaps — OpenAI's model and infrastructure software with NVIDIA's hardware and software.

NVIDIA and OpenAI stated they expect to finalize the details of this new phase of the strategic partnership in the coming weeks. The letter of intent, in other words, is a declaration of direction rather than a fully executed agreement.

How does this fit with OpenAI's other infrastructure deals?

The partnership does not replace OpenAI's existing infrastructure relationships. According to the announcement, it complements work OpenAI and NVIDIA are already doing with a broad network of collaborators — including Microsoft, Oracle, SoftBank, and Stargate partners — focused on building what the companies call the world's most advanced AI infrastructure.

That ecosystem matters for context. OpenAI's compute expansion has been financed and hosted through multiple channels, and the NVIDIA deal adds a direct equity-investment dimension alongside those arrangements. For NVIDIA, the up-to-$100 billion commitment deepens its financial exposure to its largest AI customers, aligning chip supply, capital, and model development within a single strategic relationship.

The co-optimization clause is the quiet technical core of the announcement. Aligning OpenAI's model and infrastructure software roadmaps with NVIDIA's hardware and software roadmarks means future model architectures and future GPU generations will be designed with each other in mind — a coupling that raises switching costs for both sides and concentrates innovation cycles between the two companies.

Why does the scale matter?

The numbers in the announcement define the ambition. Ten gigawatts is the floor, not the ceiling — the letter of intent specifies "at least" that capacity. NVIDIA's investment of up to $100 billion scales with actual deployment, disbursed gigawatt by gigawatt, which ties capital release to delivered infrastructure rather than promises.

The first-gigawatt date carries the most immediate weight. If the second-half 2026 target on Vera Rubin holds, OpenAI's next training runs would begin landing on NVIDIA's newest platform generation within roughly a year of the announcement. The remaining gigawatts would follow across a multi-year buildout of datacenter and power capacity.

OpenAI's user base gives the demand side of the equation. The company reports over 700 million weekly active users, plus what it describes as strong adoption across global enterprises, small businesses, and developers. Serving that audience — and training the models behind it — is what the 10 gigawatts are for.

What comes next?

The companies say they will finalize partnership details in the coming weeks, which means the specific financial terms, deployment locations, and gigawatt-by-gigawatt schedule remain open questions. What is already fixed is the direction: NVIDIA capital flowing into OpenAI as NVIDIA silicon comes online, beginning with a Vera Rubin-powered gigawatt in the second half of 2026 — and, in the companies' own framing, a shared roadmap aimed at superintelligence.