Nvidia has spent the past three years as the most valuable company on the planet by selling the chips that train the world's largest AI systems. Now it is reportedly weighing something far more unusual than another hardware sale: a financial guarantee large enough to reshape how a single data center gets built. According to reports that surfaced in late July, the chipmaker is in talks to backstop roughly $250 billion in financing for OpenAI's planned mega-campus in southern Ohio, a facility that could become one of the largest computing sites ever attempted.
What Happened
The reported arrangement, first detailed by The Wall Street Journal, would see Nvidia provide a guarantee of about $250 billion to help OpenAI secure the lease and construction financing for a 10-gigawatt data center in Piketon, Ohio. A guarantee of this kind does not mean Nvidia writes a check up front. Instead, the company would lend its balance sheet and credit standing to the project, promising lenders and landlords that obligations will be met even if OpenAI stumbles. That distinction matters, because OpenAI remains deeply unprofitable despite its enormous private valuation, which makes conventional financing for a project of this scale extraordinarily difficult to arrange.
The $250 billion figure covers real estate and lease commitments, but it is not the only number in play. Reports indicate that Nvidia is separately discussing a financing package worth roughly $350 billion tied to OpenAI's purchases of Nvidia chips over the life of the campus. Taken together, the two arrangements would represent one of the most ambitious vendor-financing structures in corporate history, with a single supplier effectively underwriting both the building that houses its products and the products themselves. The talks are still ongoing, and people familiar with the discussions have cautioned that terms are not final and the deal may not close in its current form.
For context, the total cost of the Ohio campus has been estimated at more than $500 billion once chips, labor, and power are included. That is a staggering sum for a single site, and it explains why the financing has to be assembled so creatively. No traditional bank syndicate is likely to shoulder that exposure to an unprofitable customer without a highly rated partner standing behind it. Nvidia, with its fortress balance sheet, is one of the few entities on earth that can plausibly play that role.
Why It Matters
The significance of this deal goes well beyond one data center. It signals a shift in how the AI build-out is being financed, and it deepens the interdependence between the company that makes AI chips and the company that has become their largest buyer. When a supplier guarantees its customer's debt, the line between vendor and investor blurs. Nvidia's motivation is straightforward enough: guaranteeing OpenAI's expansion helps guarantee demand for Nvidia's own accelerators for years to come. If the Ohio campus rises, it will be filled with Nvidia silicon, locking in revenue that justifies the risk.
_007.png){kind=link}
The chosen location adds a layer of history that is hard to ignore. The campus would sit on the site of the former Portsmouth Gaseous Diffusion Plant, a roughly 3,700-acre federal complex in Pike County that once enriched uranium during the Cold War before civilian nuclear fuel work wound down and operations ceased in 2001. A site built to power one era's arms race would be repurposed to power another kind of race entirely, this one measured in floating-point operations rather than kilotons. The symbolism is striking, but the practical appeal is more mundane: the land is vast, federally controlled, and already wired for heavy industrial use.
Critics have long warned that AI infrastructure spending has taken on the characteristics of a self-reinforcing loop, where chipmakers, model developers, and cloud providers invest in one another in ways that inflate reported demand. A guarantee structure like this one will intensify that debate. Supporters counter that the compute is genuinely needed, that frontier models keep improving with scale, and that someone has to finance the physical plant if the technology is going to advance. Both readings can be partly true at once, which is exactly why the arrangement is drawing so much scrutiny.
Reaction
Wall Street's response has been a mix of enthusiasm and unease. On one hand, an Nvidia-backed OpenAI campus reassures investors that demand for high-end accelerators is not a passing spike but a durable, multi-year commitment. On the other, analysts have flagged the circularity of a chipmaker underwriting its biggest customer's ability to buy chips, a structure that concentrates risk in ways that are difficult to model. If AI monetization disappoints, the same interlocking commitments that look brilliant in a boom could amplify the pain in a downturn.
Commentators across the technology and finance press have drawn comparisons to earlier infrastructure booms, from railroads to telecom fiber, where vendor financing accelerated construction but also seeded fragility. The optimists note that unlike speculative fiber that sat dark for years, AI compute is being consumed almost as fast as it can be built, with utilization rates that make the railroad analogy imperfect. Skeptics respond that demand forecasts have a way of looking unbreakable right up until they break. What nearly everyone agrees on is that the sheer scale of the numbers changes the stakes for the entire sector.
What's Next
The Ohio project is not starting from a blank page. Earlier in 2026, the U.S. Department of Energy announced a partnership with SoftBank and its energy subsidiary, SB Energy, to redevelop the Piketon site, with SB Energy committing to build roughly 10 gigawatts of new power generation, the bulk of it from natural gas. That power commitment is the hidden linchpin of the whole plan, because a 10-gigawatt data center consumes electricity on the scale of a major metropolitan area, and no such campus can operate without dedicated generation built alongside it.
The site is also woven into the broader Stargate initiative, the $500 billion AI infrastructure venture that OpenAI, Oracle, and SoftBank unveiled at the start of 2025. If the financing comes together, construction is expected to move ahead through 2026, with an initial phase delivering several hundred megawatts of capacity within a couple of years and the full build stretching out across the rest of the decade. The near-term milestones to watch are whether the Nvidia guarantee is formally signed, how lenders price the risk, and whether the power infrastructure can be permitted and built on the aggressive timeline the project assumes.
Closing Thoughts
Strip away the eye-watering figures and the Ohio deal is really a story about conviction. Nvidia is betting that demand for AI compute is deep enough and durable enough to justify guaranteeing a customer's debt on a scale that would have seemed absurd a few years ago. OpenAI is betting that the models it trains on that compute will eventually generate enough value to make the whole edifice pay for itself. Both bets rest on the same assumption: that the current trajectory of AI capability continues, and that the appetite for it does not cool.
Whether that assumption holds is the defining question hanging over the entire industry. If it does, a repurposed Cold War site in rural Ohio could become one of the engines of the next decade of computing, and creative financing like this will look like foresight. If it does not, the same interlocking guarantees could become a cautionary tale about how quickly optimism can turn into overextension. For now, the deal is still a set of reported talks rather than a signed contract, which is a useful reminder that even the largest numbers in technology begin as negotiations that can still fall apart. It is worth watching closely, precisely because it captures the moment the AI build-out stopped being about products and started being about the balance sheets willing to finance them.
한글 요약
엔비디아가 오하이오주 파이크턴에 들어설 오픈AI의 초대형 데이터센터 건설을 위해 약 2,500억 달러 규모의 금융 보증을 논의 중인 것으로 7월 말 보도됐습니다. 이 보증은 엔비디아가 직접 현금을 내는 것이 아니라, 자사의 우량 신용을 빌려줘 아직 적자 상태인 오픈AI가 임대·건설 자금을 조달할 수 있도록 뒷받침하는 구조입니다. 여기에 더해 오픈AI의 칩 구매를 지원하는 별도의 약 3,500억 달러 규모 금융 패키지도 논의되고 있어, 전체 프로젝트 비용은 5,000억 달러를 넘어설 것으로 추산됩니다.
부지는 냉전 시기 우라는을 농축하던 포츠머스 가스확산공장 터로, 2001년 가동을 멈춘 약 3,700에이커의 연방 부지입니다. 앞서 미 에너지부는 소프트뱅크 및 자회사 SB에너지와 손잡고 이곳에 약 10기가와트 규모의 발전 설비를 짓기로 했으며, 대부분은 천연가스 발전으로 채워집니다. 이 부지는 오픈AI·오라클·소프트뱅크가 2025년 초 발표한 5,000억 달러 규모의 스타게이트 구상과도 연결돼 있습니다.
업계에서는 칩 공급사가 최대 고객의 부채를 보증하는 이 구조를 두고, 인공지능 인프라 투자가 서로 얽힌 순환 구조로 위험을 키운다는 우려와, 실제 연산 수요가 그만큼 크다는 반론이 엇갈립니다. 아직 최종 계약이 아닌 협상 단계인 만큼 조건이 바뀌거나 무산될 가능성도 있어, 보증의 정식 서명 여부와 자금 조달 조건, 발전 인프라 구축 일정이 앞으로의 관전 포인트입니다.
참고 / 출처: Reuters, TechCrunch, Data Center Dynamics, OpenAI