Nvidia in Talks to Backstop $250 Billion OpenAI Data Center Deal

What happened

Nvidia is in discussions to provide a financial backstop of approximately $250 billion for OpenAI as the ChatGPT creator seeks to lease a massive data center campus in southern Ohio. The 10-gigawatt facility is being developed by SB Energy, a subsidiary of SoftBank Group, which raised $500 million from OpenAI in 2024. The $250 billion backstop would cover lease costs and certain construction expenses, but not the graphics processing units that will be installed. Including the chips, the total project cost could exceed $500 billion. Nvidia is also reportedly in talks to finance OpenAI's chip purchases separately, potentially adding another $350 billion in equipment financing. The campus is expected to provide computing capacity equal to the power consumption of about eight million homes, making it one of the largest AI infrastructure projects globally. SB Energy plans to take the unit public later this year at a valuation exceeding $50 billion.

Why it matters

This financing arrangement highlights the extraordinary capital requirements of AI infrastructure buildout and Nvidia's evolving role beyond chip supplier. By backstopping OpenAI's financing, Nvidia would leverage its strong investment-grade credit to help OpenAI secure cheaper debt than the unprofitable startup could obtain independently. The deal deepens the financial ties between two of the AI industry's most important companies, with Nvidia already having made a $30 billion investment in OpenAI earlier this year. For Oracle and other cloud infrastructure providers, this signals that competition for AI compute capacity will intensify dramatically. The sheer scale of the Ohio campus suggests that hyperscalers and AI developers view massive data center expansion as essential to maintaining competitive advantage, potentially driving up costs and timelines for securing compute resources across the industry.

Bigger picture

The potential Nvidia-OpenAI financing deal reflects a broader pattern in the AI ecosystem where chipmakers are increasingly providing capital backing to customers. Nvidia previously entered a $6.3 billion capacity backstop deal with CoreWeave and has invested in multiple AI developers including Anthropic. This circular financing model raises questions about the sustainability of AI infrastructure spending and whether returns will justify the massive capital commitments. The deal also comes as publicly traded hyperscalers like Meta, Alphabet, Amazon, and Microsoft face investor pressure to demonstrate returns on their own escalating capital expenditures. The 10-gigawatt campus would be built over several years and likely house multiple data centers, similar to Meta's Hyperion campus in Louisiana which provides half the capacity and hosts up to 11 buildings. For database and cloud infrastructure providers, the race for AI compute capacity is creating unprecedented demand for data center equipment, power infrastructure, and networking components.

What to watch

Monitor whether Nvidia formalizes the backstop agreement and discloses the terms in its quarterly filings, as transparency around financial commitments of this scale will indicate how the board views the associated risks. Watch for updates on the Ohio campus construction timeline and chip selection, as the facility would likely use Nvidia's upcoming Rubin Ultra or Feynman chip series launching in 2027-2028 rather than current-generation silicon. Track SB Energy's planned public offering later this year to gauge investor appetite for AI infrastructure projects at scale. Observe how other hyperscalers respond to this development in their capital expenditure plans and whether they accelerate their own data center buildouts. Finally, watch for earnings commentary from cloud infrastructure providers on how megaprojects of this scale impact pricing power, equipment availability, and competitive dynamics in the AI compute market.