Nvidia OpenAI Backstop for Ohio Drops Below $120B
The Nvidia OpenAI backstop for the proposed data center campus in Ohio has been cut from $250 billion to less than $120 billion, with the revised guarantee now covering only the first phase of the project. Nvidia and OpenAI could sign an agreement as early as this weekend, while OpenAI continues to negotiate a binding lease for the full 10-gigawatt site.
The scale-back of the Nvidia OpenAI backstop follows investor concern about the size of the risk Nvidia was carrying through large financing commitments. It also lands days after Nvidia moved the opposite way on structure: on Monday it partnered with six major financial institutions to launch compute-financing platforms targeting more than $500 billion in third-party capital for AI infrastructure. The two moves frame the central question of the AI buildout, namely who carries the credit risk when the largest customers cannot borrow on their own.
Nvidia OpenAI Backstop: From $250 Billion to First Phase
SB Energy, a unit of SoftBank, is developing the campus in southern Ohio. Completion would make it the largest data center project announced to date. The guarantee was designed to help OpenAI lease computing capacity at the site, and the deal would be Nvidia's largest customer-financing arrangement by a wide margin. OpenAI's push to own infrastructure rather than rent it from cloud providers also makes the project a direct challenge to the established cloud hierarchy.
Nvidia and OpenAI began talks in late July with a backstop of up to $250 billion under consideration. The terms being finalized now point to a figure closer to $100 billion, applied only to the first construction phase, with later stages left to be financed on their own merits. The revised figure emerged on Friday, three days after the platform launch, and OpenAI is still in talks on a binding lease that would cover the entire project, so the scope of the guarantee and the scope of the lease no longer match.
That mismatch matters. The Nvidia OpenAI backstop now covers only the first phase, but a lease for the full 10 gigawatts would bind OpenAI to capacity beyond that scope, leaving the economics of the later phases unsettled. SB Energy, the developer in the middle, relies on Nvidia's credit to borrow at acceptable rates, and with the support narrowed to phase one, the rest of the build has no equivalent backstop yet.
The financing layers around the Ohio project now look like this:
| Element | Originally discussed | Revised status |
|---|---|---|
| Ohio lease backstop | Up to $250 billion | Under $120 billion |
| Guarantee coverage | Full project | First phase only |
| Chip financing talks | Up to $350 billion | Still under negotiation |
| Compute-financing platforms | Not yet launched | Launched Monday with six banks, $500B+ target |
The Mechanics of a Circular Structure
The Nvidia OpenAI backstop exists because of a credit gap. OpenAI does not carry an investment-grade credit rating, so lenders have been unwilling to extend financing to the company directly on acceptable terms. Instead, SB Energy borrows against Nvidia's balance sheet, and the chipmaker absorbs contingent leasing debt in exchange for securing future demand for its processors.
A backstop of this kind functions as a credit wrap: if the borrower cannot meet lease payments, the guarantor steps in. In practice, lenders underwrite Nvidia's balance sheet rather than OpenAI's, which is why the size of the figure matters to investors far beyond the Ohio site itself.
That structure is circular in a specific sense: Nvidia guarantees the debt that pays for capacity, and that capacity buys Nvidia's silicon. The guarantee buys the chipmaker a long-term revenue floor, but only while the infrastructure stays utilized and cash-flowing. If end demand weakens, Nvidia is left standing behind the lease obligations alone.
What investors are pricing in is the failure case. If utilization falls short of projections, Nvidia holds lease obligations on capacity that would have to be resold at market rates while the debt service continues. A guarantee that acts as a revenue floor in a rising market becomes a balance-sheet weight in a downturn, and the Ohio terms suggest shareholders want that weight kept small.
The Ohio lease guarantee is not Nvidia's only exposure. A separate negotiation covers up to $350 billion in financing for the processors that would fill the campus. If both tracks close, Nvidia's contingent obligations would cover more than half of a buildout that could top $500 billion in total cost. That chip-financing track is separate, but lenders will price it against the same risk calculus that produced the smaller backstop.
What the Nvidia OpenAI Backstop Cut Signals
The trimmed Nvidia OpenAI backstop is the first visible constraint on the vendor-financing model that has powered the AI buildout. Chipmakers have increasingly acted as lenders of last resort for customers who want to own infrastructure without the balance sheets to fund it, and the Ohio revision shows that model now has a ceiling set by investor tolerance as much as by engineering capacity.
There is a direct tension with the platforms announced Monday. Those platforms are designed to route third-party capital into AI infrastructure, which would take weight off Nvidia's own books. The Ohio guarantee runs the other way, putting the company's balance sheet behind a single customer, and investors appear to have drawn a line at how much contingent debt one project can carry.
For Nvidia's shareholders, the reduction lowers the contingent liabilities attached to a single customer. The company's valuation rests on the durability of AI demand, and a guarantee of nearly a quarter-trillion dollars would have made that demand assumption explicit on the balance sheet. The narrower first-phase figure keeps the revenue-floor logic while capping the downside.
The Monday announcement was itself a signal that AI infrastructure is being treated as a new asset class, with banks lining up to channel institutional money into compute. The Ohio revision shows the limit of that enthusiasm when the credit behind the asset is a chipmaker's guarantee rather than the customer's own cash flow.
For OpenAI, the project is as much about independence as about capacity. Owning a proprietary campus would reduce its reliance on the cloud providers that currently supply its compute, and the financing structure is the price of that independence. The cut does not change the goal, but it does slow the pace at which the company can pursue it, and its ability to fund large-scale commitments remains under scrutiny.
For the wider market, the final Nvidia OpenAI backstop terms will set the template for how chipmakers, landlords and cloud customers divide infrastructure risk, including how much exposure a vendor takes, how much a developer such as SB Energy carries, and how much a customer must prove it can fund on its own. The two tracks now run in parallel: Nvidia is brokering third-party capital through the bank platforms while still guaranteeing the first phase of Ohio directly.
For decision-makers, the practical read is that Nvidia and its peers will keep financing AI capacity, but in smaller, phase-by-phase increments with more third-party capital in the stack. The trimmed Nvidia OpenAI backstop is a negotiated compromise, and a signing this weekend will show whether the new ceiling holds when the remaining phases of the Ohio project come up for financing.
Why This Matters
The shrinking of the Ohio guarantee is the point where investor discipline caught up with the vendor-financing engine behind the AI buildout. The final terms will split infrastructure risk between Nvidia, SoftBank's SB Energy and OpenAI, and that split sets the precedent for every large AI data center deal that follows.
Photo by Brecht Corbeel on Unsplash
Related Articles
- Nvidia OpenAI Circular Financing: $250 Billion Guarantee Reshapes AI Infrastructure Risk
- AI Business Roundup: Infrastructure Billions, Market Jitters, and a New Regulatory Era
- OpenAI IPO Delay: Inside the $1 Trillion Standoff Reshaping AI Finance
✔Human Verified
Researched and cross-referenced against primary sources by the Bytevyte editorial team. This article was generated with the assistance of artificial intelligence and reviewed by the Bytevyte editorial team.