Nvidia is in talks to guarantee roughly $250 billion of financing so OpenAI can lease a 10-gigawatt data center rising on a decommissioned uranium enrichment site in southern Ohio, according to a Wall Street Journal report published Sunday, July 26 — an arrangement that, if completed, would rank among the largest corporate credit backstops ever extended and would tie the world's most valuable chipmaker directly to the debt of its single biggest customer.

The guarantee would let SoftBank's energy subsidiary, SB Energy, raise construction and lease debt on the strength of Nvidia's balance sheet rather than OpenAI's. OpenAI, which carries no investment-grade credit rating, would become the anchor tenant of the Piketon, Ohio, campus — its first facility occupied as a lessee rather than as a customer of Microsoft, Amazon, or Oracle. The full campus could cost more than $500 billion once chips are installed, and Nvidia is separately discussing financing as much as $350 billion of the accelerators that would fill it. Terms have not been settled, and the people cited in the reporting cautioned the deal could still collapse. Reuters said it could not immediately verify the figures, and Nvidia and OpenAI did not respond to requests for comment.

A former weapons site becomes an AI megacampus

The site itself is extraordinary. SB Energy broke ground on March 20 at the former Portsmouth Gaseous Diffusion Plant, which enriched uranium for the U.S. weapons program from 1954 until 2001 and is still being decontaminated. The Department of Energy listed it among 16 federal sites opened to data center construction, and SB Energy leases the land rather than owning it. Powering the campus will require 9.2 gigawatts of new natural gas generation plus $4.2 billion of transmission work with AEP Ohio — capacity underwritten in part by the $33.3 billion Japan committed under its trade agreement with the United States. The first phase, roughly 800 megawatts, is expected in 2028. Commerce Secretary Howard Lutnick controls allocation of the site's power, and executives from Anthropic, Microsoft, and Google have all lobbied him for access in recent weeks.

The Ohio project dwarfs OpenAI's separately announced 3.2-gigawatt Project Camellia in Georgia and reflects a strategic pivot: OpenAI abandoned building its own data centers last year, opting instead to lease compute at scale. The company has since raised its projected compute spending to roughly $750 billion through 2030, up from about $600 billion earlier this year. Nvidia has already invested $30 billion into OpenAI this year, following a $100 billion commitment disclosed in 2025.

Why the numbers strain belief

The scale of the proposed guarantee is what has drawn the sharpest scrutiny. Nvidia's most recent quarterly filing caps its maximum gross exposure across all disclosed partner-facility lease guarantees at $3.5 billion, with just $712 million held in escrow against it. A $250 billion backstop would be roughly 71 times that entire disclosed guarantee book. It exceeds Nvidia's full-year fiscal 2026 revenue of $215.9 billion and is nearly four times the $62.6 billion in cash and marketable securities the company held when its fiscal year closed on January 25.

Those figures explain why the report reignited the "circular financing" debate that has shadowed the AI trade all year — the concern that a chip supplier is effectively funding its own customer's demand for its chips. Investor Michael Burry, who disclosed and expanded a short position against Nvidia in a July 24 post, reacted bluntly on X: "Around and around we go. Nvidia to guarantee $200 billion of ChatGPT's spending on $NVDA chips." Burry has cited the Bank for International Settlements' 2026 annual report, which flagged AI infrastructure financing arrangements as carrying systemic risk through circular structures.

Aleksandar Tomic, an associate dean at Boston College, drew a direct line to the dot-com era. "The demand is not as big as it appears to be because the companies are buying from each other, using their own money to some degree," he said. "They're essentially using Nvidia's money to buy Nvidia chips." He questioned why the structure is needed at all: "If everything is so hunky-dory and everything is full steam ahead in the entire AI ecosystem, why is this kind of deal even needed?"

Not everyone reads it as a warning sign. Michael Monaghan, founder of Founder's ETF, argued the criticism overstates the case. "Like a lot of things in life, there is a downside or upside spin," he said. "That's not necessarily circular financing. That's just an economy. I understand the pushback, but I think there is less substance there." Bulls contend the backstop reduces "demand fade" risk and signals Nvidia is willing to underwrite the next wave of AI capacity, supporting long-duration GPU orders. Nvidia shares nonetheless fell about 4.9 percent on Monday.

What to watch next

The deal remains unconfirmed and unsigned, so the first thing to watch is whether Nvidia formally acknowledges the guarantee — and how it would account for a liability of this magnitude, which its filings currently describe as a credit derivative with "immaterial" fair value. Also critical: whether Lutnick allocates the site's power to OpenAI over rival bidders, whether SB Energy secures lenders willing to accept Nvidia's credit at this scale, and whether the mounting political backlash against data centers — New York last week became the first state to impose a construction moratorium — reaches a site already backed by federal land and a foreign trade commitment. If the arrangement holds, it will set a template for how the AI buildout gets financed. If it unravels, it may become the clearest test yet of whether the sector's demand is as real as its price tags suggest.

"They're essentially using Nvidia's money to buy Nvidia chips."
— Aleksandar Tomic, Associate dean, Boston College
$250B
Financing backstop
10 GW
Data center capacity
$500B+
Full campus cost
$350B
Separate chip financing