Nvidia Weighs $250 Billion Guarantee for OpenAI Ohio Data Center
SoftBank-developed 10-gigawatt campus could cost over $500 billion including silicon
Nvidia is reportedly in advanced discussions to guarantee roughly $250 billion in construction and lease financing to support OpenAI's development of a massive 10-gigawatt data center in Piketon, Ohio. Developed by SoftBank's energy subsidiary, the unprecedented project represents a monumental shift in how Big Tech finances the extreme energy and physical infrastructure required to sustain the AI expansion.
Key details
The proposed 10-gigawatt data center campus would be built on the site of a former uranium enrichment plant in Piketon, Ohio, a location chosen because it already features heavy-duty power interconnections and necessary industrial permits. SoftBank's energy subsidiary, SB Energy, is developing the site, which is expected to cost over $500 billion in total once Nvidia's silicon hardware is fully integrated.
To finance this colossal buildout, Nvidia is discussing a $250 billion backstop for the lease and construction debt to ease lenders' concerns regarding OpenAI’s lack of an investment-grade credit rating. Separately, discussions are underway for an additional $350 billion in financing specifically for OpenAI to purchase Nvidia's specialized chips. The project is planned to be completed in phases, with the first phase delivering approximately 800 megawatts of capacity by 2028. Additionally, the U.S. government is heavily involved in power allocation for the site under a recent trade agreement, which includes a pledge from Japan to invest $33 billion in a supporting natural gas plant.
Why this matters
A 10-gigawatt data center campus consumes a small city's worth of electricity—equivalent to the output of roughly ten large nuclear reactors. Supporting this level of compute-intensive AI workloads requires immense electricity supply and physical grid integration, highlighting that energy availability remains the primary bottleneck for scaling generative AI models. By stepping in to guarantee $250 billion of its customer's debt, Nvidia is directly underwriting the physical infrastructure required to deploy its own chips, showcasing how resource constraints are forcing silicon providers into the real estate and energy financing markets.
Context
This proposed project is more than three times the size of the 3-gigawatt data center campus OpenAI announced recently in Georgia. It follows a growing industry trend of repurposing retired nuclear, industrial, or fossil-fuel facilities due to their pre-existing transmission infrastructure and cooling access. As AI infrastructure spending exceeds $700 billion annually, off-balance-sheet commitments and complex financing arrangements have become the norm, with some estimates placing Big Tech's off-balance-sheet AI liabilities at nearly $1.65 trillion.
Risks and open questions
The scale of Nvidia's proposed financial intervention has sparked intense debates among economists and market experts regarding "circular financing" in the AI sector. Critics, including veteran investor Michael Burry, argue that a chip vendor guaranteeing its largest customer's lease obligations to fund purchases of its own chips concentrates severe financial risk on a single balance sheet. If the massive capital investments in AI do not yield corresponding revenue, or if grid delays prevent the Piketon site from coming online in a timely manner, both the developer and the guarantor face staggering exposure.
What happens next
The talks about the Piketon, Ohio site and its financing guarantees remain in progress and subject to change. Observers are watching for official confirmation of the Nvidia backstop, which would require significant disclosures under SEC reporting rules. Meanwhile, other tech giants including Anthropic, Microsoft, and Google have reportedly approached U.S. Commerce Secretary Howard Lutnick to lobby for power allocation at the same federal site, indicating a fierce multi-company scramble for Ohio's remaining energy capacity.
Source: The Next Web Published on AI Usage Global, author: AUG Bot
