Skip to content
Rush Commerce
AI & Automation3 min read

Nvidia's $250B OpenAI backstop: keep your AI contracts short

Nvidia is reportedly guaranteeing ~$250B of financing for OpenAI's 10GW Ohio campus. When your vendor needs a co-signer, don't lock in multi-year AI pricing.

The Wall Street Journal reported this weekend that Nvidia is in talks to guarantee roughly $250 billion in financing so OpenAI can lease a data center campus. Read the shape of that: the chip vendor is co-signing the loan its customer needs in order to buy the chips. It isn't a scandal — vendor financing is old and legal — but it is a fact about your supply chain, and it should change how long you're willing to lock in any AI vendor's pricing.

What actually happened

Per WSJ reporting relayed by Benzinga and DigiTimes, Nvidia is negotiating a roughly $250 billion backstop on the lease and construction debt behind a 10-gigawatt campus in Piketon, southern Ohio — a decommissioned uranium-enrichment site about 50 miles from Columbus, being developed by SoftBank's SB Energy subsidiary. The first phase is roughly 800 megawatts, arriving around 2028. The full project could exceed $500 billion including silicon, with a separate discussion covering some $350 billion in chip-purchase financing.

The guarantee exists for a specific reason: OpenAI does not carry an investment-grade credit rating, so it cannot borrow at those terms on its own name. Microsoft, Google, and Anthropic have reportedly looked at the same site. Investor Michael Burry's summary of the structure was four words: "Around and around we go."

Why AI vendor financing matters for your business

The number to sit with isn't $250 billion. It's 2028. The capacity is three years out, the debt is now, and the revenue servicing that debt is monthly subscriptions and per-token API spend — a line item that includes yours. That's a maturity mismatch, and mismatches get resolved by repricing the near-term product.

So price accordingly. Keep the term short. An annual commit for a discount is fine. A three-year enterprise agreement priced against a capacity roadmap that hasn't been built is a bet on someone else's balance sheet, and you don't get a seat at that table. Prepaid credits are the same bet with worse liquidity.

Then measure the thing that actually matters: cost per completed task, not cost per million tokens. A model that's twice the list price and finishes the job in one pass is cheaper. Re-run that math every quarter, because it moves — and if the number ever makes a switch worth it, you want the switch to be a config change, not a project.

Key takeaways

  • Nvidia is reportedly in talks to backstop ~$250B of financing for a 10GW OpenAI campus in Piketon, Ohio, developed by SoftBank's SB Energy
  • The total project could top $500B, with another ~$350B in chip-purchase financing discussed separately; first phase is ~800MW around 2028
  • The guarantee is needed because OpenAI lacks an investment-grade credit rating and can't borrow at those terms alone
  • Keep AI commitments annual or shorter, skip prepaid credits, and track cost per completed task so a vendor switch stays a config change

Don't know what a task actually costs you to automate? Then you can't tell whether a vendor's price change matters. Start with the numbers: run our ROI calculator, then tell us what you're trying to automate and we'll tell you what it takes to run it on more than one provider.

Sources: Benzinga, DigiTimes.

  • #ai-infrastructure
  • #vendor-risk
  • #openai
  • #nvidia
  • #contracts
TR

Tommy Rush — Founder, Rush Commerce

Operator turned builder. 15+ years running operations — now shipping the systems businesses run on. More

Get The Rush Report weekly — one email, zero fluff.