OpenAI and Nvidia
bet $105 billion on Ohio.
OpenAI has signed a deal with Nvidia covering a data-center site in Ohio. Nvidia is guaranteeing payments of up to $105 billion over as long as 15 years. It's the latest step in a relationship that deepened just three months after Nvidia's roughly $30 billion equity investment on May 10.
Why the alliance
moved this fast
From May's "equity" to August's "land." Another step forward in three and a half months.
The story starts on May 10. To help OpenAI secure priority access to its GPUs, Nvidia made a roughly $30 billion equity investment in the company. A chipmaker putting capital into its own biggest customer was an unusual move, and it was already read at the time as a sign of how tightly the two companies were binding themselves together.
Three and a half months later, on August 17, the two companies took their next step. They signed an agreement covering "Stargate Ohio," the data-center campus OpenAI is building in New Albany, Ohio (Licking County), under which Nvidia guarantees payments of up to $105 billion over 15 years. According to THE DECODER, it's the largest financial guarantee ever tied to a single data-center project.
The shift from equity to a lease guarantee is not a small distinction. Taking equity means holding shares; a lease guarantee is closer to Nvidia effectively backstopping OpenAI's ability to pay — which means Nvidia itself now shares the risk if demand for ChatGPT and other products doesn't grow as fast as planned.
This isn't renting land.
It's buying up tomorrow's compute, in advance.
How the deal is built
A 15-year guarantee. 1.4 gigawatts. More than 600,000 GPUs. Stargate Ohio, by the numbers.
Stargate Ohio is the roughly 1,100-acre campus OpenAI is building in New Albany. Under the new agreement, Nvidia guarantees payment for up to $105 billion of the GPU infrastructure procurement costs for that campus, spread over 15 years. Rather than a simple equipment sale, it's an unusual "vendor financing" arrangement in which Nvidia itself takes on the investment risk.
Elena Cho, who leads Nvidia's data-center business, is reported to have said at the announcement that "we're not just selling GPUs — we're betting on the demand behind them". An analysis by TechBuzz notes that this kind of vendor-financing structure lets Nvidia recognize revenue earlier, while leaving Nvidia itself exposed to impairment risk if OpenAI's demand comes in below projections.
At full build-out, the campus is expected to reach up to 1.4 gigawatts of IT load capacity, running more than 600,000 Nvidia Blackwell Ultra-generation GPUs. Phase 1 (400 megawatts) is slated to come online in 2028, followed by Phase 2 (900 megawatts) in 2029, with the site reaching full capacity in 2031.
The deal, in numbers
Renee Talbot, a regional economist at Ohio State University, estimates that once Stargate Ohio is fully operational it will support roughly 3,200 permanent jobs, calling it "a single data-center project large enough to move the region's employment statistics." That said, data-center jobs typically peak during construction and shrink sharply once a site moves into operation — the 8,500-job peak won't simply persist as a steady-state figure.
Who feels this, and how
Engineers, business leaders, indie developers — the impact lands differently for each.
Engineers & researchers
The direct impact is limited for now. But as the tightest point of GPU supply shifts from cloud majors toward hyperscale megaprojects, that shapes the medium-term outlook for API pricing and rate limits. For now, this is one to watch rather than act on.
Business & leadership
OpenAI's expanding supply could strengthen negotiating leverage for large enterprise deals — but the flip side of a contract this size is that a demand miss would ripple across the whole industry. Scrutinize SLA and repricing clauses closely before signing long-term contracts.
Indie developers & startups
Any direct benefit won't arrive before 2028 at the earliest. In the meantime, keep tracking API and GPU-cloud pricing trends, and it's worth avoiding over-reliance on a single provider in your stack.
Short term (weeks) — track pricing shifts
Watch official blogs and earnings materials from OpenAI and Nvidia's cloud partners for any price cuts or hikes tied to changes in GPU supply.
Medium term (months) — diversify vendors
The more GPU supply concentrates around a handful of megadeals, the higher the relative risk of depending on a single provider. Consider spreading API and GPU-cloud usage across more than one.
Long term (2028+) — judge the new capacity on results
Once Phase 1 comes online (expected 2028), confirm it's actually translating into better pricing or latency before revisiting your contracts. No need to move early.
Optimism only tells half the story
A number as large as $105 billion also feeds concerns that AI infrastructure spending is turning into a bubble. The guarantee still dwarfs OpenAI's current annual revenue, and the bet still rests on demand continuing to grow roughly as planned.
The power burden is real too. 1.4 gigawatts is comparable to the electricity draw of a mid-sized city, and details on grid-upgrade plans and renewable-energy procurement for the site have not yet been disclosed. The effect of concentrated data-center demand on electricity prices and grid capacity remains a live concern for residents around New Albany.
There's also a broader industry question: as GPU supply increasingly concentrates around megadeals between OpenAI and Nvidia specifically, it could make procurement harder for other AI companies and cloud providers. How far this "concentration of compute" goes — and what it does to diversity and competition across the industry — is worth watching closely.