Nvidia is exploring a staggering $250 billion financing guarantee to help OpenAI secure a lease for a 10-gigawatt data center campus being developed in Pike Township, Ohio. If completed, the arrangement would rank among the largest financing agreements ever underwritten in artificial intelligence.

The data center project is being developed by SoftBank’s energy subsidiary. The overall infrastructure build could eventually exceed $500 billion—including the Nvidia semiconductors expected to power the facilities. Nvidia’s guarantee would cover the lease and construction financing, leaving the cost of chips itself outside its scope. The company is simultaneously negotiating a separate commitment to fund OpenAI’s acquisition of Nvidia processors, a figure that could reach $350 billion.

Why Nvidia Is Betting This Much

For Nvidia, the math is simple: guaranteed demand for its chips for years. OpenAI currently rents compute capacity from Microsoft, Amazon, and Oracle. A deal of this scale would flip the leverage. Nvidia wouldn’t just sell chips—it would finance the infrastructure that runs them. Every dollar of chip sales is locked in. Every iteration of H100s, Blackwell, or future accelerators flows directly into OpenAI’s hands-owned data centers.

It’s a hedge against a post-monopoly future. Right now, Nvidia owns AI training hardware. But cloud providers are building custom silicon to reduce dependency. Microsoft is pushing in-house chips. Amazon has Trainium and Inferentia. Google has TPUs. If hyperscalers stop buying Nvidia’s most expensive parts, the company loses its golden goose. Locking in OpenAI’s workloads for the next decade is insurance against that day.

What This Means for OpenAI

Independence. OpenAI has borrowed massive compute capacity from cloud providers—a relationship that worked during the company’s growth phase but creates dependency. Owning infrastructure means controlling costs, latency, and redundancy. It means OpenAI can run AGI or whatever comes next without negotiating with cloud providers on their terms.

The deal also signals confidence. OpenAI is betting its future on AI requiring ever more compute, ever more power, ever more specialized hardware. A $250 billion guarantee from Nvidia is a vote of faith that the company’s models will be worth powering at that scale.

The Catch

Nvidia assumes risk. If OpenAI’s models hit capability ceilings or if the market shifts, Nvidia could be left underwriting massive unused capacity. That’s why the guarantee is structured carefully—it’s tied to the lease, not to chip purchases. OpenAI has to occupy and operate the data center, not just buy capacity from Nvidia that sits idle.

The deal also makes Nvidia a stakeholder in OpenAI’s success. If OpenAI fails or pivots away from Nvidia chips, the entire structure collapses. This is a tighter coupling than either party has publicly acknowledged.