Nvidia Cuts Its OpenAI Data Center Guarantee Nearly in Half

The Nvidia OpenAI data center deal just got smaller, at least on paper. Nvidia has revised the financial guarantee it plans to offer for OpenAI’s massive Ohio computing campus, trimming the figure from roughly 250 billion dollars down to under 120 billion dollars. The Wall Street Journal first reported the change on Friday, citing people close to the negotiations, and a Reuters report confirmed the broad strokes shortly after.

The scaled back Nvidia OpenAI data center deal still ranks among the largest corporate financing arrangements ever discussed in the AI industry. It is simply not as enormous as the number that circulated in late July, when the two companies were reportedly weighing a quarter trillion dollar backstop for a single site.

A binding agreement on the smaller Nvidia OpenAI data center deal could be signed within days, according to people familiar with the talks, though neither company has confirmed the exact terms publicly.

Why the Nvidia OpenAI Data Center Deal Shrank

Investors are the reason the Nvidia OpenAI data center deal changed shape. Once word spread that Nvidia might guarantee up to 250 billion dollars in lease payments and construction debt for a project it does not own, shareholders started asking pointed questions about how much risk the chipmaker should absorb on behalf of a customer, even one as important as OpenAI.

Nvidia sits in an unusual position. It sells the GPUs that power OpenAI’s models, invests directly in OpenAI, and now finds itself effectively co-signing loans for the data centers those chips will run inside. That layered exposure worried analysts who track how concentrated Nvidia’s balance sheet has become around a single customer relationship.

The revised structure reportedly covers only the first phase of the Ohio project rather than the entire multi year buildout, which lowers Nvidia’s near term exposure while still letting construction begin on schedule.

Inside the 10 Gigawatt Ohio Campus

The site sits in Pike County, Ohio, and is being developed by SB Energy, a subsidiary of SoftBank. When finished, the campus is expected to draw 10 gigawatts of power, enough to rival the output of several nuclear reactors, and would rank as the largest data center project announced anywhere to date.

OpenAI is still negotiating a binding lease for the full campus even as the smaller Nvidia OpenAI data center deal covers only the opening phase. Nvidia’s guarantee is separate from the cost of the chips themselves, which reporting suggests could require its own financing arrangement worth up to 350 billion dollars.

That distinction matters for anyone trying to size up the real numbers. The Ohio data center guarantee is about construction and lease obligations, essentially making lenders comfortable enough to break ground. The chip financing is a different pool of money entirely, and neither piece has been finalized.

What the OpenAI Nvidia Backstop Means for the Wider AI Buildout

The Nvidia OpenAI data center deal is one thread in a much larger tapestry of AI infrastructure financing. Alphabet, Microsoft, Amazon, Nvidia, Oracle and Meta have collectively committed close to 1.5 trillion dollars toward chips, data centers and energy capacity over the next several years, according to recent industry tallies. Meta’s own infrastructure ambitions, detailed in Welp Magazine coverage of the Muse Glimmer superintelligence push, show how every major lab is racing to lock down compute at a similar scale.

Nvidia is not slowing down elsewhere either. Earlier in the week it partnered with six major financial institutions to launch compute financing platforms aimed at pulling in more than 500 billion dollars in third party capital for AI infrastructure financing. The scaled back Ohio guarantee looks less like retreat and more like Nvidia spreading its commitments across more partners instead of shouldering each one alone, a pattern also visible in how AMD recently invested up to 5 billion dollars in Anthropic to diversify AI infrastructure financing across chipmakers.

For OpenAI, the stakes are just as high. The company remains unprofitable despite a valuation near 852 billion dollars, and its ability to fund infrastructure at this scale is under constant scrutiny from analysts and, increasingly, from its own investors and business partners.

How This Fits OpenAI’s Broader Chip and Compute Strategy

OpenAI has been working multiple angles on compute at once. The company has explored building its own silicon, a move Welp Magazine covered in depth when it explained why OpenAI built the Jalapeno chip instead of buying more GPUs, while also leaning on outside capital through arrangements like the 10 billion dollar Stripe and OpenRouter infrastructure deal. The Nvidia OpenAI data center deal fits that same pattern of spreading bets across chips, cash and partnerships rather than relying on one source.

Demand for the physical building blocks behind all of this keeps climbing too. Chipmakers up and down the supply chain, including TSMC after its record revenue quarter driven by AI chip demand, are feeling the pull of an industry that cannot seem to build data centers fast enough.

A Pattern Investors Should Watch

This is not the first time a headline grabbing AI financing number has shrunk once a deal reached the negotiating table. As compute demand keeps climbing, expect more of these enormous proposed figures to get revised downward as lenders, boards and shareholders weigh in before ink hits paper.

The Nvidia OpenAI data center deal also signals something about how the market is maturing. A year ago, headline numbers rarely faced this kind of public pushback. Now every billion dollar guarantee gets scrutinized the moment it leaks, and companies are adjusting their public posture accordingly, favoring phased commitments over one massive upfront promise.

For more coverage of how the biggest AI labs are financing the infrastructure race, keep following Welp Magazine tech desk as this story and others like the AMD Anthropic investment and OpenAI chip strategy continue to develop.