Introduction

In August 2026, OpenAI signed a 20-year lease for a data center that could eventually consume 8 gigawatts of power—enough to run roughly eight nuclear power plants, or more than the entire city of San Francisco used in 2024 . The facility will be built on the site of a former Cold War-era uranium enrichment plant in Piketon, Ohio, and will run “exclusively” on Nvidia chips .

The deal comes with a $105 billion financing commitment from Nvidia, the company’s largest such guarantee to date . It’s part of a broader pattern in which AI developers rely on financial backing from their own suppliers to build the infrastructure underpinning the technology .

Here’s how the project works, why it’s controversial, and what it signals about the scale of America’s AI buildout.

The Project: 8 Gigawatts in Piketon, Ohio

The data center will be developed by SB Energy, a subsidiary of SoftBank, on the site of the former Portsmouth Gaseous Diffusion Plant in Piketon . The facility will initially have 4.25 GW of capacity, with the option to add another 3.75 GW .

For context: a nuclear power plant typically provides around 1 GW of energy. San Francisco consumed just over 5 GW in 2024 . This single data center, at full buildout, would consume more power than the entire city.

OpenAI expects the project to create 35,000 temporary construction jobs over six years and 2,500 long-term operating jobs . The company has committed $40 million to a community grant fund supporting workforce training, utilities, healthcare, and small business development .

The Financing Structure: Nvidia’s $105 Billion Guarantee

The most unusual element of the deal is Nvidia’s role. The chipmaker will provide $105 billion in credit support for the data center lease . Under the agreement, if OpenAI becomes insolvent and defaults on the lease or fails to make payments on time, Nvidia will be responsible for covering the payments .

This is a financing guarantee, not an equity investment. Nvidia previously considered a backstop of up to $250 billion, according to reports from the Wall Street Journal and Bloomberg in July . The final figure is less than half that.

Nvidia will also invest $1.5 billion** into SB Energy . And SoftBank, through SB Energy, will invest more than **$4 billion into local energy infrastructure and commit to building at least 10 GW of new energy generation to power the data center .

The company preemptively pushed back on critiques about “circular financing”—the concern that Nvidia is effectively funding its own customers, inflating demand and creating a feedback loop . “OpenAI will pay the lease,” Nvidia wrote in a blog post, adding that the deal was made with “the same discipline we apply to supply-chain management” .

Why This Deal Matters

The Scale of AI Infrastructure

This project is part of a much larger wave of investment. Alphabet, Meta, Microsoft, and Amazon have collectively committed nearly $2.4 trillion** in future AI infrastructure spending, according to Bloomberg . Alphabet alone reported **$902 billion in purchase commitments, contractual obligations, and uncommenced leases—more than nine times the figure from a year earlier .

Microsoft disclosed an $80 billion backlog** of Azure orders it cannot fulfill due to power constraints . Amazon’s CEO Andy Jassy said that even with **$220 billion in capital expenditure planned for 2026, it still isn’t enough to meet cloud infrastructure demand .

The Power Problem

The Ohio data center’s 8 GW capacity illustrates a central tension in the AI boom: the infrastructure is being built faster than the power grid can support it. That’s why SB Energy is committing to build 10 GW of new generation—the data center can’t rely on existing grid capacity .

Microsoft, Google, Meta, and Amazon have all signed long-term nuclear power agreements to secure baseload electricity for their AI facilities. Google recently announced a €13 billion investment in Finland, including a 22-year nuclear power purchase agreement .

The Political Backlash

More than 70% of Americans oppose building data centers near their homes, according to a Gallup poll from March 2026 . Rising electricity bills, attributed in part to data center demand, are fueling debate in the midterm elections .

OpenAI and SB Energy are trying to get ahead of this opposition. Their commitments to cover energy costs without passing them to local residents, create jobs, and fund community programs are a direct response to the growing hostility .

The Broader AI Infrastructure Wave

The Ohio project is one of several massive data center deals announced in recent months:

Anthropic’s $45 billion lease in West Virginia, a six-year contract with cloud provider Nscale that will deliver 460 MW of computing capacity using Nvidia’s next-generation Vera Rubin platform .

AWS’s $6 billion investment** in a third data center campus in Shreveport, Louisiana, bringing its total Louisiana investment to **$18 billion .

Meta’s plans for a 5 GW data center in Louisiana and a 1 GW facility in Ohio .

Stargate, the joint venture between OpenAI, SoftBank, Oracle, and MGX, targeting $500 billion in AI infrastructure investment by 2029 with 10 GW of capacity across multiple sites .

Globally, the AI infrastructure buildout is accelerating. In India, Tata Consultancy Services announced a $7.4 billion** AI data center campus in Hyderabad with up to **1 GW** capacity . AM Intelligence ordered **9,000 Nvidia Vera Rubin systems** as part of an **$8 billion plan to build 1 GW of computing capacity .

What Comes Next

The Ohio data center represents the most ambitious single AI infrastructure project announced to date. At 8 GW, it dwarfs anything currently operating. The $105 billion financing guarantee from Nvidia is unprecedented in scale.

But the deal also illustrates the structural risks in the AI buildout. Nvidia is guaranteeing the lease payments of a company that will use Nvidia chips to run AI models. If AI demand slows—or if OpenAI’s business model doesn’t generate the revenue to cover the lease—Nvidia could be on the hook for billions.

The company insists the deal is disciplined. “OpenAI will pay the lease,” Nvidia wrote . Whether that confidence holds will depend on whether the AI boom continues at its current pace.

For now, the message is clear: the scale of AI infrastructure investment has reached a level with no historical precedent. The Ohio data center, at 8 GW, is the clearest symbol of that ambition—and its risks.