OpenAI Seeks Credit Guarantee on Nvidia’s Balance Sheet
OpenAI is in discussions with Nvidia for a backstop of up to $250 billion that would help fund its plans to lease a massive new artificial intelligence data center, CNBC confirmed on Monday. The guarantee would let OpenAI raise debt on the strength of Nvidia’s credit, according to a person familiar with the talks who asked not to be named because the details are confidential. The Wall Street Journal first reported the negotiations.
The structure would cover lease and construction debt for a 10-gigawatt data center campus planned for Pike County, Ohio, on a site that once functioned as a uranium-enrichment plant. The chips that would eventually fill the facility are being negotiated separately, the person said. The total cost of the campus could exceed $500 billion.
What the Site Looks Like
Ten gigawatts is roughly equivalent to the annual power consumption of 8 million U.S. households, according to a CNBC analysis of Energy Information Administration data. The Ohio site is one of the largest single industrial power footprints ever considered for a private-sector buildout, and its prior nuclear history may help streamline permitting for on-site generation. AEP Ohio and other regional utilities have been briefed on potential substation and transmission upgrades that would be required to serve a campus of that scale.
The talks are still in progress and subject to change, according to a second person familiar with the plans. Nvidia declined to comment.
Why the Backstop Matters
OpenAI kick-started the AI boom with the launch of ChatGPT in 2022 and has been racing to secure computing capacity to meet future model and product demand. It faces heightened competition from Anthropic, Google, Amazon and Meta, all of which are collectively spending hundreds of billions of dollars on capital expenditure to support their own AI infrastructure. The pace of that buildout has put pressure on OpenAI to lock in power, land and credit capacity before rivals do.
By tying the data center’s lease and construction debt to Nvidia’s balance sheet, OpenAI can borrow against investment-grade credit without diluting its own equity further. The structure mirrors how hyperscale cloud operators financed early campus builds, but at a scale and concentration of risk that is unprecedented in the AI sector. Nvidia’s existing investment-grade rating gives lenders a familiar credit anchor and reduces the cost of debt versus what OpenAI could raise standalone.
From the September 2025 Commitment to Today’s Talks
In September, Nvidia said it would invest up to $100 billion in OpenAI as part of a strategic partnership under which the AI lab would deploy at least 10 gigawatts of Nvidia systems. That investment never fully materialized, though Nvidia did contribute $30 billion to the record-breaking funding round OpenAI closed in March at a valuation reported near $500 billion at the time.
Nvidia CEO Jensen Huang said the March round might be the last time the company invests in OpenAI before its IPO. OpenAI confidential filed for an initial public offering with the Securities and Exchange Commission in June, though the company has not disclosed a public timeline for the debut. The Ohio backstop, if it closes, would be structured to mature alongside that IPO path rather than depend on it.
Valuation and the Open-Weight Question
Private investors now value OpenAI at nearly $1 trillion on the bet that the company will maintain its lead in frontier AI and find a long-term workable business model. That valuation faces increasing uncertainty as a host of open-weight alternatives, largely out of China, undercut its pricing power across enterprise and consumer segments. Alibaba’s Qwen series, Moonshot’s Kimi models, and DeepSeek’s recent updates have all been cited by enterprise buyers as credible substitutes for parts of OpenAI’s API footprint.
Whether the Ohio campus is financed and built will hinge on three things: how much of the $250 billion Nvidia is willing to formally guarantee, how the chip-purchase line is sized separately, and how quickly Ohio regulators and utilities can clear a multi-gigawatt interconnection. Each of those is independently capable of slowing the timeline, and the project is being watched by utilities across the Midwest as a bellwether for hyperscale load growth.
If the deal closes, it would be the largest single private-sector credit commitment tied to AI infrastructure to date, exceeding the combined debt guarantees that supported the early cloud buildout across Northern Virginia and the Pacific Northwest. It would also mark the first time a chip vendor has acted as the primary credit anchor for an AI lab’s data center footprint at this scale.
The Broader AI Compute Race
The Ohio talks arrive against a backdrop of unprecedented AI infrastructure investment across the industry. Anthropic, Google and Meta have each committed tens of billions of dollars to data center capacity, and Microsoft has expanded its long-running partnership with OpenAI to include additional capacity commitments. The total committed AI infrastructure spend across the leading labs is now widely estimated to exceed $500 billion through the end of the decade, with power availability rather than chip supply emerging as the binding constraint.
That power constraint is precisely why the Ohio site is so attractive. A former uranium-enrichment plant carries with it legacy grid connections and large-scale industrial zoning that would be difficult to assemble at scale on a greenfield site. If OpenAI can close the backstop and lock in the lease terms, it would secure a multi-gigawatt footprint in a single transaction, which is materially faster than assembling the equivalent capacity across multiple smaller sites.
What to Watch Next
The next major signal will come from Nvidia’s commentary on its own capex guidance and any explicit reference to the Ohio site. OpenAI, which has not publicly commented on the Pike County project, is likely to address the financing structure only when the deal is finalized. Investors will also look for any update on the timing of OpenAI’s IPO and how the new debt and chip commitments are reflected in the company’s path to public-market profitability.
Regulators in Ohio and at the Federal Energy Regulatory Commission will have their own reviews to complete, and any grid interconnection delay could push the operational date into 2028 or later. Until then, the financing structure remains the single biggest open variable.