OpenAI is in advanced talks to lease a computing campus on federal land in Pike County, Ohio, a site that used to enrich uranium for nuclear weapons during the Cold War. The proposed deal involves a 20-year lease, with Nvidia acting as financial guarantor for OpenAI's payments and for the project developer's financing. If finalized, the campus could cost at least $500 billion to fully build out.
To put that number in context: the entire Northern Virginia data center market, which is the largest in the world, held roughly 5 gigawatts of capacity in 2025. This single Ohio site is planned for 10 gigawatts, roughly double that. Even the first phase, targeted for 2028, covers only 800 megawatts.
SoftBank's energy subsidiary, SB Energy, is building the power and physical infrastructure at the site. In March, construction formally broke ground under a partnership with the US Department of Energy. The power plan leans heavily on natural gas: at least 9.2 gigawatts out of the planned 10 will be gas-fired. SB Energy has committed $4.2 billion to upgrade electrical transmission lines in southern Ohio and has pledged not to raise electricity bills for local residents.
Nvidia's role here is the part worth watching. This is not a standard vendor relationship. Nvidia has committed to investing up to $100 billion in OpenAI across a broader partnership announced in September 2025, and the Ohio deal would add another layer, making Nvidia a financial backstop for the lease itself. A chip company guaranteeing a customer's 20-year rent and a developer's financing is not a sales arrangement. It is a structural tie that binds Nvidia's financial health to OpenAI's ability to pay its bills.
This is happening across the industry, not just at OpenAI. Apollo and Blackstone announced a $35 billion financing package for Anthropic, OpenAI's main competitor, just this week. That deal uses Broadcom's custom chips rather than Nvidia's, and targets over 20 gigawatts of computing capacity through 2028. The pattern is the same: chip companies and financial firms are locking in AI labs with long-term, tightly structured financing deals.
For businesses that use OpenAI's tools today, this matters in ways that go beyond headlines. When your software vendor is locked into 20-year commitments backed by a single chip supplier, the flexibility of the whole supply chain narrows. Minimum usage thresholds, reservation tiers, and pricing structures tend to become more rigid when infrastructure commitments are this large and this long. The cost of building this much computing power does not disappear; it gets embedded in contracts that eventually flow downstream.
The deal still has a long path to completion. Permitting, regulatory review, and power grid approvals are all unresolved. Full buildout of the campus is expected to take at least a decade. OpenAI also filed confidentially for an initial public offering on June 8, which means its financial disclosures, including the details of infrastructure commitments like this one, will eventually become public in a way they never have been before.
That transparency is coming regardless of when the IPO actually happens. When OpenAI's public prospectus lands, businesses that depend on its tools will, for the first time, be able to read audited information about the financial health and contractual obligations of a vendor many of them have quietly made central to their operations.