Nvidia is negotiating to secure $250 billion in debt for OpenAI's data centers.

      Nvidia is reportedly negotiating to offer approximately $250 billion in financing guarantees for OpenAI, as stated by the Wall Street Journal on Sunday, citing sources knowledgeable about the discussions. This financial support would enable OpenAI to lease a 10-gigawatt campus being developed in southern Ohio by SoftBank’s energy subsidiary.

      The rationale for requiring such a guarantee is clear: OpenAI lacks an investment-grade credit rating. Nvidia's financial standing would effectively serve as a substitute, allowing lenders to assess the debt based on Nvidia's creditworthiness instead of that of the borrower.

      Financial coverage

      The $250 billion figure pertains to the leasing and construction debt for the data center, not the Nvidia chips that will be housed within it. Discussions regarding financing for chip purchases could potentially escalate to an additional $350 billion. The total expenditure for the entire initiative is projected to exceed $500 billion, including costs for silicon, with the first phase, approximately 800 megawatts, expected to be completed by 2028.

      Nvidia's support would aid in creating financing mechanisms that reassure lenders regarding the project's funding. Reuters stated it could not promptly confirm the report, and Nvidia, OpenAI, and the US Commerce Department did not provide comments when approached.

      The political aspect of the site

      The power supply for the Ohio campus is governed by the US government and separately financed by Japan through a recent trade agreement linked to Tokyo’s commitment to invest $33 billion in a natural gas plant. Commerce Secretary Howard Lutnick is involved in determining access allocations.

      OpenAI has been actively negotiating for the site for several weeks and is one of the top contenders. Anthropic, Microsoft, and Google have also recently made inquiries with Lutnick, according to the Journal.

      The circularity concern

      Investor Michael Burry, recognized for predicting the housing market crash of 2008, remarked on X: “Around and around we go. Nvidia to guarantee $200 billion of ChatGPT’s spending on $NVDA chips.” He had increased his short position in Nvidia on Friday at $210.28. Tech analyst Ed Zitron criticized the move, deeming it “an insane thing to do on so many levels and about as bearish as it gets, especially considering it is being built by SoftBank.” Both comments reflect the same structural issue: a chip supplier backing its customer's capacity to continue purchasing chips.

      Nvidia has previously allocated over $40 billion toward AI equity stakes, including around $30 billion in OpenAI. Its investment in Ilya Sutskever’s Safe Superintelligence exemplifies this pattern, where Nvidia finances firms that then spend that funding on Nvidia hardware.

      A precedent in structure

      Vendor guarantees for data center commitments are increasingly common, though the scale here is unprecedented. Google has backed approximately $44 billion for other companies’ data center leases, a figure that the proposed Nvidia guarantee would surpass by nearly six times.

      OpenAI's capital structure has grown more complex, with eight banks now supporting SoftBank’s record $40 billion loan to OpenAI, and SoftBank has independently sought margin loans secured by its OpenAI shares. A significant portion of these obligations remains outside conventional debt reporting. Research estimates that Big Tech’s off-balance-sheet AI commitments total around $1.65 trillion, with Meta alone responsible for about $420 billion, nearly triple its reported debt.

      Implications for OpenAI

      For OpenAI, this arrangement represents a preliminary step toward owning its computing resources rather than leasing them from Microsoft, Amazon, and Oracle. Control over infrastructure differentiates between merely negotiating for capacity and having the ability to dictate terms.

      For Nvidia, the incentive lies in ensuring demand certainty. By guaranteeing the construction, it secures years of chip orders from a customer that would otherwise face challenges in financing them, which is why some critics view the arrangement as problematic.

      The scale of concern

      Spending on AI infrastructure is anticipated to exceed $700 billion this year, with the industry expected to invest more than $3 trillion on data centers through 2028, largely funded by the chips inside. The central risk does not lie in the potential irrelevance of AI, but in the possibility that the expansion outpaces demand or that competitive pressures diminish returns faster than the debt can be repaid.

      No agreements have been finalized. The Journal emphasized that discussions surrounding the site and its financing are ongoing and subject to modification, which is significant regarding the proposed $250 billion guarantee.