OpenAI is in advanced talks to lease a proposed 10-gigawatt data center campus in southern Ohio, with NVIDIA expected to guarantee the lease obligations and the developer’s financing, according to The Information, which cited people familiar with the discussions.
Ten gigawatts is the figure that matters. That is roughly the output of ten large nuclear reactors, committed to a single AI compute campus under a 20-year lease. The campus could cost at least $500 billion to build at current prices for chips, power, and construction. OpenAI would control the computing equipment and begin payments once the site starts operating, with the first phase expected in 2028.
The structure is the latest iteration of a pattern that has become routine across frontier AI: the chip vendor helps finance the buildout that will be filled with its own chips. Google agreed to backstop Anthropic’s $35 billion chip commitment. Google separately arranged to rent 110,000 GPUs from SpaceX. NVIDIA is now the reported financial backstop for the largest single-site compute commitment disclosed to date. In each case, the supplier’s balance sheet sits behind the buyer’s infrastructure obligation.
The OpenAI-NVIDIA relationship has been deepening for longer than this deal alone. In September 2025, the two companies announced a partnership to deploy at least 10 gigawatts of NVIDIA systems, with NVIDIA stating it intended to invest up to $100 billion in OpenAI as each gigawatt came online. The Ohio campus would formalize that commitment by adding a lease guarantee on top of the hardware supply agreement. The first phase is scheduled to use NVIDIA’s Vera Rubin platform.
The site itself aligns with a project the US Department of Energy announced in March: the redevelopment of the former Portsmouth Gaseous Diffusion Plant near Piketon, Ohio. SB Energy, a SoftBank Group company, committed to building 10 gigawatts of new power generation capacity at the site, including at least 9.2 gigawatts fueled by natural gas. The department did not name a tenant when it announced the project.
The circular financing structure carries specific risks worth naming. When NVIDIA guarantees OpenAI’s lease and backs its financing, it is no longer acting purely as a vendor. Sanchit Vir Gogia, chief analyst at Greyhound Research, described the shift to Network World: “For enterprises, standardizing on OpenAI is therefore no longer a model decision. It is exposure to a single economic gravity field spanning silicon, power, capital, and regulatory attention.”
The 2028 operating date is not incidental to the funding story. OpenAI has publicly described a target of building automated AI researchers by roughly that period, and the company filed its S-1 earlier this year. A 20-year, 10-gigawatt infrastructure commitment is exactly the kind of long-dated liability that public-market investors will price against OpenAI’s revenue trajectory. The compute commitment and the capability roadmap are the same bet, sized at roughly half a trillion dollars.
Neil Shah, vice president for research at Counterpoint Research, told Network World that projects at this scale typically take a decade to reach full capacity and carry significant execution risk. “A 10-gigawatt site won’t just appear overnight,” he said. Gogia added that scale alone does not solve compute scarcity: “More compute does not cure scarcity. It reschedules it.” The financing, he noted, surfaces downstream as minimum commitments and usage thresholds even as token prices fall.
The lease remains under negotiation, and questions around financing terms, permitting, and deployment timelines are unresolved, according to The Information.
Enterprise teams currently building long-term contracts with OpenAI should treat this deal as a structural disclosure, not a capability announcement: a 20-year NVIDIA-backed obligation shapes pricing floors, vendor concentration risk, and competitive dynamics in ways that will take years to fully resolve.
Reported by Network World (networkworld.com), 2026-06-10.
