Analysis
Nvidia is in talks to guarantee roughly $250 billion in lending that would let OpenAI lease a 10-gigawatt data center campus being built by SoftBank subsidiary SB Energy in Piketon, Ohio -- on the site of a former federal uranium-enrichment facility -- according to reporting Monday. Nvidia is separately said to be discussing financing of up to $350 billion to help cover the chips that would fill the campus, putting the combined project financing tied to Nvidia potentially above $500 billion.
The arrangement extends a pattern that has defined much of Nvidia's 2026: rather than simply selling GPUs to OpenAI at arm's length, Nvidia increasingly finances the infrastructure that houses them, effectively vendor-financing demand for its own product. Critics have flagged this as a form of circular financing -- Nvidia's revenue depends partly on capital it itself is providing -- while supporters argue it's a rational way for the best-capitalized company in the AI stack to accelerate a buildout that would otherwise be gated by OpenAI's and SoftBank's own balance sheets.
The deal's structure -- an outside party guaranteeing debt so a lab can lease compute it doesn't own outright -- closely echoes Meta's newly announced $14 billion data center joint venture with BlackRock, where BlackRock-managed funds take an 80% ownership stake financed partly with debt while Meta leases back the capacity. Across both deals, the common thread is hyperscalers and labs increasingly keeping gigawatt-scale AI infrastructure off their own core balance sheets, shifting the underlying leverage into financing vehicles that are harder for public markets to price.
The scale here dwarfs even that: a single 10-gigawatt campus represents a meaningful fraction of all announced AI data center capacity globally, and the choice of a former uranium-enrichment site underscores how much AI buildout is now competing for the same grid capacity and industrial real estate that energy-intensive legacy industries once occupied.
For GPs and LPs, the Nvidia-OpenAI Ohio talks are the clearest evidence yet that AI infrastructure financing has outgrown traditional corporate balance sheets entirely -- this is project finance at a scale usually reserved for national infrastructure, not corporate capex. The bear case is that Nvidia is underwriting demand it cannot be certain will materialize at this scale, concentrating an enormous amount of counterparty risk in a single customer relationship.
What to watch: whether the financing terms are finalized before OpenAI's next major funding milestone, how much of the $250 billion guarantee ends up on Nvidia's own balance sheet versus structured off it, and whether other hyperscalers respond with comparable vendor-financing arrangements of their own.
