Analysis
Nvidia is in talks to provide a roughly $250 billion financial backstop that would let OpenAI lease computing capacity from a 10-gigawatt data center campus under construction in Pike County, Ohio, according to the Wall Street Journal, with CNBC and other outlets confirming the discussions Monday. The facility, sited on a decommissioned uranium-enrichment plant roughly 50 miles south of Columbus, is being built by SB Energy, the power development arm of SoftBank, and the backstop would apply specifically to the lease and construction financing rather than the chips themselves. In a parallel and separate negotiation, Nvidia is also discussing financing up to $350 billion of OpenAI's chip purchases, meaning the chipmaker could ultimately stand behind as much as $600 billion of the roughly $500 billion-plus total project cost once land, power, construction and silicon are all counted.
The Ohio campus has been described as the largest data center project ever announced, and its financing structure is unusual even by 2026's inflated AI-infrastructure standards. The power supplying the site is controlled by the federal government and is being funded in part by Japan, which committed $33 billion to a natural-gas power project on the same federal land as part of its broader trade agreement with Washington; Commerce Secretary Howard Lutnick is directly involved in deciding which companies get access to that power. That combination -- a foreign government funding the electricity, a chipmaker backstopping the lease, and a cabinet secretary allocating access -- has no clean precedent in prior data center buildouts.
The backstop talks land squarely inside Nvidia's now-familiar playbook of using its balance sheet to guarantee demand for its own hardware roadmap, a strategy it has already run with xAI, CoreWeave, and this month's more-than-$500-billion SK Group initiative in South Korea. OpenAI, for its part, has been assembling an unprecedented patchwork of compute commitments this year -- its own roughly $30 billion-plus, 3.2-gigawatt Georgia campus, the SoftBank-backed Stargate program, and now a second SoftBank-adjacent Ohio site -- while Microsoft has simultaneously been building in-house MAI models specifically to reduce its own reliance on, and cost exposure to, OpenAI.
Set against recent comparables, the scale here is still extraordinary: SK Hynix and SK Telecom's Nvidia-backed Korea initiative topped $500 billion in combined commitments just days earlier, and Nvidia's $1.5 billion Amkor prepayment for US advanced packaging capacity looks almost modest by comparison. A $250 billion backstop -- before even counting the separate $350 billion chip-financing conversation -- would be roughly a third of Nvidia's entire market capitalization committed to guaranteeing a single customer's real estate and power lease, a concentration of counterparty risk that has no analog in the chip industry's history.
For founders and GPs, the read-through is that the AI infrastructure arms race has moved decisively past venture-scale capital into sovereign-and-hyperscaler-scale financial engineering, where the actual constraint is no longer talent or even chips but structured balance-sheet guarantees between a handful of companies and governments. Any startup pitching itself as "infrastructure for the AI buildout" is now implicitly competing for attention against deals measured in the hundreds of billions, and LPs underwriting AI-adjacent infrastructure funds need a real view on Nvidia's aggregate counterparty exposure, not just chip demand.
The bear case is exactly the one Bloomberg detailed the same day: this is the definition of circular financing, where Nvidia extends credit to a customer whose primary use of that credit is buying more Nvidia chips, and the market's reaction was immediate -- the cost of insuring Nvidia's debt against default jumped by the most on record Monday as investors absorbed reports that the company is now discussing more than $750 billion in AI-related financial commitments in total. If OpenAI's revenue growth doesn't keep pace with the lease obligations Nvidia is guaranteeing, the backstop transforms from a demand-generation tool into genuine balance-sheet risk for the one company the entire AI trade is priced around.
What to watch: whether the $250 billion backstop and the separate $350 billion chip-financing talks are formalized as reported or scaled back once terms are finalized, how credit-default-swap pricing on Nvidia's debt behaves over the coming weeks, and whether Commerce Secretary Lutnick's role in allocating Ohio's federally controlled power draws Congressional scrutiny given the AI kill-switch and open-weight policy fights already underway in Washington this month.
