The Nvidia OpenAI data center financing talks now being reported would guarantee roughly 250 billion dollars so OpenAI can lease a 10-gigawatt campus being built in southern Ohio, first reported by the Wall Street Journal on 26 July and since confirmed in broad strokes by multiple outlets, including Al Jazeera. The scale of the number is difficult to hold in your head next to almost anything else in corporate finance: it is close to a full year of Nvidia’s own revenue, and roughly four times the company’s cash reserves, committed to underwriting a single tenant’s lease on a facility that has not yet been built.
The project sits on the site of the former Portsmouth Gaseous Diffusion Plant near Piketon, Ohio, a facility that enriched uranium for the US weapons programme from 1954 until 2001 and is still being decontaminated. SoftBank’s energy subsidiary, SB Energy, broke ground there in March alongside Energy Secretary Chris Wright, Commerce Secretary Howard Lutnick and SoftBank chairman Masayoshi Son. The full project, including the Nvidia chips that will eventually fill it, is expected to cost more than 500 billion dollars, with an initial phase of roughly 800 megawatts targeted for completion in 2028.
Why Nvidia OpenAI Data Center Financing Looks So Unusual
The mechanics of the deal exist because of a specific, structural problem: OpenAI does not carry an investment-grade credit rating, which makes it expensive and difficult for SB Energy to raise construction debt against OpenAI as the eventual tenant. Nvidia’s balance sheet effectively stands in for one, letting lenders price the debt against the chipmaker’s credit rather than the borrower’s. Nvidia has already put 30 billion dollars directly into OpenAI, and is separately discussing financing worth up to 350 billion dollars more for the chips that would go inside the Ohio facility once it is operational.
Analysts have already started asking the obvious question about circularity. LiveAIWire’s earlier reporting on how AI data centre demand is already straining real communities found that the industry’s build-out has consistently outpaced the regulatory and financial scrutiny meant to accompany projects of this size, and this deal is the clearest example yet. Aleksandar Tomic, associate dean at Boston College, told Al Jazeera that the structure amounts to Nvidia lending OpenAI the means to keep buying Nvidia chips, calling the broader pattern a real concern about whether AI infrastructure demand reflects genuine customer revenue or simply capital moving in a circle between a small number of companies.
The Numbers Nvidia Would Be Standing Behind
Nvidia’s own financial disclosures make the scale of this specific ask clearer. A detailed breakdown of Nvidia’s public filings shows the company’s most recent quarterly report caps its total exposure across every partner facility lease guarantee it has disclosed at 3.5 billion dollars, a figure that shrinks further as those partners repay their lessors over time. A 250 billion dollar guarantee for the Ohio project alone would be roughly seventy times larger than that entire existing guarantee book combined. Nvidia closed its last full fiscal year with 62.6 billion dollars in cash and marketable securities against 215.9 billion dollars in revenue, meaning the reported guarantee figure would exceed the company’s entire cash position several times over.
None of the terms have been finalised, and multiple outlets reporting the story note the arrangement could still collapse before any agreement is signed. Commerce Secretary Lutnick controls the allocation of power at the Ohio site, and reporting indicates Anthropic, Microsoft and Google have all separately approached him about access to the same site in recent weeks, underscoring how constrained the underlying electricity supply actually is relative to the industry’s collective ambitions.
Why OpenAI Wants This Specific Deal
For OpenAI, a completed lease would mark its first step toward controlling infrastructure directly as a tenant, rather than renting compute exclusively from Microsoft, Amazon and Oracle as it has done throughout its history. LiveAIWire’s coverage of OpenAI’s 852 billion dollar valuation and its delayed IPO plans found that the company’s projected compute spending has already climbed toward 750 billion dollars through 2030, even as it has never posted an annual profit and continues to miss some of its own internal revenue targets.
Locking in dedicated capacity years in advance, even through a financing structure this unusual, is one way OpenAI is trying to guarantee it does not lose the infrastructure race to a rival with deeper pockets or a more direct line to the chips it needs.
For Nvidia, the incentive runs in the opposite direction but points at the same outcome. Guaranteeing the lease effectively guarantees demand for Nvidia’s own chips inside the facility for years to come, at a moment when competition from AMD, Google’s own TPUs and a wave of Chinese open-weight models has made the future demand curve for Nvidia hardware considerably less certain than it looked two years ago. A structure that looks financially strange in isolation becomes more explicable once both companies’ underlying incentives are placed side by side.
The Political Backdrop the Deal Is Landing Into
The financing talks have emerged just as data centres have become a genuine electoral flashpoint across the United States. New York became the first state to impose a one-year moratorium on new data centre construction earlier in July, and at least a dozen other states have proposed similar measures.
In Texas, which trails only Virginia in total operational and planned data centres nationally, a sitting state senator lost his own primary after backing a local data centre project, while a Democratic congressional candidate has separately proposed ending state tax breaks for the industry. A recent Gallup poll cited in reporting on the trend found that 71 percent of Americans oppose data centre construction that does not give local communities a say in the decision.
That backdrop does not necessarily threaten the Nvidia OpenAI data center project directly, since federal land and a bespoke power arrangement with Japan already anchor it outside most local planning fights. LiveAIWire’s reporting on the industry’s search for computing capacity beyond the terrestrial grid entirely found that the same underlying pressure, a grid that cannot expand fast enough to match AI’s electricity demand, is what is now pushing serious money toward ideas as extreme as data centres in orbit.
The Ohio deal and the orbital experiments are two very different answers to the identical underlying problem: there is more demand for AI compute than there is grid capacity to power it, and every company racing to build that capacity is now improvising financing and siting solutions that would have looked implausible three years ago.
What Happens Next
Nvidia’s stock fell nearly five percent in the trading session immediately following the report, a reaction some analysts read as the market questioning why such an unusual guarantee is even necessary if underlying AI demand were as straightforwardly strong as the industry’s public messaging suggests. Whether the deal closes on anything resembling the terms currently being reported, and whether it survives the scrutiny that its scale is already attracting from both financial analysts and state legislators, will be one of the clearest tests yet of how much longer the industry’s current financing model can keep pace with its own infrastructure ambitions.
About the Author
Stuart Kerr is Technology Correspondent at LiveAIWire, covering artificial intelligence, emerging technology, and their impact on business, society, and everyday life. LiveAIWire publishes original AI journalism every weekday at liveaiwire.com.