Nvidia commits up to US$100 billion in OpenAI deal

By Zara Kincaid · 2026-07-20

What was announced, and what it is not yet

Nvidia and OpenAI announced a letter of intent on 22 September 2025 to deploy at least 10 gigawatts of Nvidia systems for OpenAI's next-generation AI infrastructureverifiedVerified Source: openai.com.^[1] A letter of intent is not a binding contract. Both companies retain room to renegotiate terms as deployment milestones are reached.

OpenAI named Nvidia its preferred strategic compute and networking partner for its AI factory growth plans.^[1] That designation carries real commercial weight: it effectively locks in Nvidia's architecture as the default for OpenAI's infrastructure expansion at a moment when rivals including AMD and custom silicon from Google and Amazon are competing hard for the same contracts.

The scale: 10 gigawatts, US$100 billion and what those numbers mean

Nvidia intends to invest up to US$100 billion in OpenAI progressively as each gigawatt of systems is deployed.verifiedVerified Source: openai.com^[1] The phased structure ties Nvidia's financial commitment directly to OpenAI's build-out pace, so the full figure depends on OpenAI executing across all 10 gigawatts.

A 10-gigawatt deployment represents one of the largest concentrated compute capacities ever announced, comparable to the output of multiple large-scale power plants.^[1] Training frontier AI models can require hundreds of megawatts of continuous power, and a commitment at this scale will drive substantial new data centre construction and energy infrastructure. The first gigawatt is scheduled for the second half of 2026 on Nvidia's Vera Rubin platform.^[1]

Jensen Huang, Nvidia's founder and CEO, framed the deal as the continuation of a decade-long technology race. "NVIDIA and OpenAI have pushed each other for a decade, from the first DGX supercomputer to the breakthrough of ChatGPT. This investment and infrastructure partnership mark the next leap forward, deploying 10 gigawatts to power the next era of intelligence," Huang said.^[1]

Why a chipmaker is funding its biggest customer

Nvidia built its dominance on GPUs and accelerated computing platforms that underpin virtually every major AI training run, from early DGX supercomputers through to the upcoming Vera Rubin architecture.^[1] OpenAI is the company that has done the most to demonstrate that demand publicly, so Nvidia funding OpenAI is, structurally, Nvidia securing its own order book.

Sam Altman, OpenAI's co-founder and CEO, put the logic plainly. "Everything starts with compute. Compute infrastructure will be the basis for the economy of the future, and we will utilise what we're building with NVIDIA to both create new AI breakthroughs and empower people and businesses with them at scale," Altman said.^[1]

OpenAI now serves over 700 million weekly active users across global enterprises, small businesses and developersverifiedVerified Source: openai.com.^[1] At that scale, the compute requirement is not speculative; it is already present and growing. For Nvidia, the investment thesis is that every dollar placed into OpenAI returns multiples in hardware orders.

Risks and what could go wrong

A chipmaker promising to invest US$100 billion in its biggest customer creates a circular dynamic that raises legitimate questions about financial risk and market concentration.^[1] If OpenAI's revenues stall or its valuation corrects, Nvidia's investment exposure could prove costly. The phased, gigawatt-linked structure limits Nvidia's immediate outlay but does not eliminate the risk.

Market concentration is a separate concern. Naming Nvidia the preferred strategic partner entrenches a supplier relationship that is already dominant across the industry, and regulators in the United States, European Union and elsewhere have begun scrutinising the concentration of AI infrastructure in the hands of a small number of companies. A deal at this scale will draw attention.

Energy sustainability sits alongside the financial questions. Frontier AI's energy demands are growing faster than grids and green power supply can absorb in most regions, and a 10-gigawatt commitment, if fully deployed, would represent a draw on power infrastructure comparable to multiple large national grids running simultaneously. The long-term viability of that trajectory remains an open question for regulators, utilities and climate advocates alike.

What it means for the AI arms race and for businesses

Nvidia's move signals that the compute arms race has entered a phase where the hardware supplier itself is prepared to bankroll the consumption side of the market.^[1] OpenAI has now secured preferential access at a scale that smaller rivals cannot match, changing the competitive dynamics for every other AI company seeking Nvidia capacity.

For businesses building on OpenAI's APIs and platforms, the announcement suggests raw compute will not be the binding constraint in the near term, at least for OpenAI's product lines. Altman's framing of compute infrastructure as the basis for the economy of the future signals that OpenAI intends to use the additional capacity to push capability frontiers further and faster, which typically flows through to new model releases and expanded enterprise offerings.^[1]

The Vera Rubin platform, on which the first gigawatt is to be deployed in the second half of 2026, is Nvidia's next-generation architecture and represents a significant capability step from current hardware.^[1] OpenAI's early-deployment status on that platform gives it a meaningful head start over competitors who will access the same hardware later and at smaller scale.

Zara Kincaid

Zara Kincaid covers breaking news and sport for Bushletter. Fast and verb-led, he writes with a news-wire cadence and no patience for PR spin.