ByThe Nvidia OpenAI deal is the financing and infrastructure partnership the two companies announced in August 2026, under which Nvidia is backing up to $105 billion for a new AI data center campus in Pike County, Ohio. OpenAI will lease the site for 20 years and run it exclusively on Nvidia GPUs, with the first phase of capacity expected online by 2028. It is the latest, and largest, step in a partnership that already includes a $100 billion pledge from 2025 and a $30 billion commitment from early 2026.
If you follow AI news, you have probably seen three different dollar figures attached to Nvidia and OpenAI this year, and it is easy to assume they are the same deal reported three times. They are not. This guide breaks down exactly what was announced in August 2026, how the financing actually works, where the money is going, how it stacks up against rival infrastructure spending, and what it means for anyone building on top of AI models like ChatGPT, Claude, or Gemini.
What Is the Nvidia OpenAI Deal?
The Nvidia OpenAI deal announced on August 17, 2026 is a financing agreement, not a straightforward cash investment. Nvidia is providing up to $105 billion in financing support to secure land, power, and infrastructure for a new data center complex known as the PORTS-Pike Technology Campus in Pike County, Ohio, according to Nvidia and OpenAI’s joint announcement. SB Energy will build and own the physical site, and OpenAI will lease the campus for 20 years to run its AI workloads.
The arrangement sits inside a broader strategic partnership between the two companies to deploy 10 gigawatts of Nvidia systems for OpenAI, first outlined by OpenAI’s official partnership announcement. Under that plan, Nvidia intends to invest up to $100 billion progressively as each gigawatt of capacity is deployed, while this Ohio-specific credit line covers the lease and power payments that get the site built in the first place. In practice, that means the August 2026 news is best read as the first concrete, named project inside a much larger multi-year commitment, rather than a brand-new pledge on top of it.
How Much Money Is Nvidia Actually Committing?
The headline number, $105 billion, is a financing ceiling, not a guaranteed lump-sum payment. Nvidia is backing “defined portions of lease and power payments” for the Ohio campus, which functions more like a loan guarantee than a direct equity check, according to reporting from Bloomberg (August 17, 2026). That distinction matters because it means Nvidia’s actual cash exposure depends on how much of the campus gets built and how OpenAI’s lease payments are structured over the 20-year term.
- Nvidia backs land, power, and shell-building costs for the Ohio campus, not the GPUs themselves.
- SB Energy owns and builds the physical site; OpenAI is the tenant on a 20-year lease.
- The complex will run exclusively on Nvidia GPUs, tying the financing directly to future chip orders.
- Fortune reported the effective deal size came in roughly $145 billion lower than early rumors suggested, a sign that some of the earlier headline figures were inflated.
Put simply, Nvidia is not writing OpenAI a $105 billion check. It is agreeing to stand behind the debt and lease obligations that make a bank, or SB Energy’s own financing partners, comfortable building an $100-billion-plus campus for a single tenant. If OpenAI’s business falters and cannot cover its lease, Nvidia’s financing backstop is what gets called on.
Where Is the New Data Center, and When Will It Open?
The site is in Pike County, Ohio, roughly in the southern part of the state, and is being developed by SB Energy as the PORTS-Pike Technology Campus. The initial phase targets 4.25 gigawatts of computing capacity, with an option to add another 3.75 gigawatts later, for a possible total of 8 gigawatts, according to the joint Nvidia-OpenAI announcement. The first 800 megawatts of that capacity is expected to come online by 2028, which is fast by data-center construction standards but still years away from full build-out.
Once complete, the campus will run exclusively on Nvidia GPUs, which CNBC estimated could amount to roughly 1.5 million chips deployed at the site through 2030, worth an estimated $150 billion to $200 billion in potential revenue for Nvidia per hardware generation. That figure explains why Nvidia is willing to underwrite the infrastructure costs: every gigawatt of capacity built for OpenAI becomes a guaranteed multi-year order book for Nvidia’s chip business, well beyond what a single financing round could ever capture on its own.
How Big Is 8 Gigawatts of AI Data Center Capacity?
Gigawatt figures get thrown around in AI coverage without much context, so it helps to translate the number. One gigawatt is roughly enough continuous electricity to power around 750,000 average U.S. homes, based on typical residential consumption figures the U.S. Energy Information Administration publishes each year. An 8-gigawatt campus, the full build-out target for the Ohio site, would draw an amount of power comparable to a mid-size American city running at full demand around the clock, every day of the year.
That scale is exactly why these deals take years to build rather than months. Utilities have to plan new transmission lines and, in many cases, new generation capacity; local grids have to be reinforced; and the chips themselves have to be manufactured, shipped, and installed in a coordinated sequence. Nvidia financing the lease and power costs upfront is, in effect, a way of pulling that multi-year timeline forward instead of waiting for OpenAI to fund it out of operating revenue.
Why Does OpenAI Need This Much Computing Power?
ChatGPT and OpenAI’s API now serve hundreds of millions of users and a growing number of enterprise customers, and both training the next generation of models and serving inference requests at that scale require enormous, continuous GPU capacity. Every new model generation, from the kind of reasoning upgrades seen in GPT-5.6 to agentic coding tools, needs more compute than the last, not less, because larger context windows, longer reasoning chains, and multi-step agent workflows all multiply the number of tokens a model processes per user request.
Competing labs are making the same bet. Google has scaled Gemini’s infrastructure alongside releases like Gemini 3.7 Flash, and Anthropic has done the same for models such as Claude Opus 5. That is the core logic behind gigawatt-scale deals like this one: model providers are no longer just buying GPUs off a price list, they are locking in years of future capacity through financing arrangements, long-term leases, and dedicated power contracts, because both the chip supply and the grid capacity to run frontier AI models are effectively rationed right now.
How Does the Nvidia OpenAI Deal Compare to Earlier Investment Rounds?
Nvidia and OpenAI have now announced three distinct financial arrangements within about a year, and each one covers something different. The table below lays out how the August 2026 Ohio financing fits alongside the two earlier commitments.
| Announcement | Timing | Amount | What It Covers |
|---|---|---|---|
| Initial Nvidia-OpenAI pact | 2025 | Up to $100 billion | Progressive investment tied to each gigawatt of Nvidia systems deployed |
| Early 2026 commitment | Early 2026 | $30 billion | Additional pledge toward OpenAI’s compute build-out |
| Ohio data center financing | August 17, 2026 | Up to $105 billion | Land, power, and lease-payment backing for the Pike County, Ohio campus |
Read together, these three rounds show a shift from Nvidia making a general-purpose pledge toward Nvidia financing a specific, named piece of physical infrastructure it can point to, the PORTS-Pike Technology Campus, with a fixed 20-year tenant already under lease. Each round is also structured differently, which is part of why headline totals reported in different articles this year have not always matched.
How Does This Fit Into the Broader AI Infrastructure Race?
Nvidia and OpenAI are not the only companies making gigawatt-scale bets in 2026. Amazon, Google, Meta, and Microsoft have all announced their own expansions of AI data center capacity this year, and industry analysts have noted that combined AI infrastructure commitments across Big Tech now run into the trillions of dollars when leases, chip orders, and power contracts are added together. Tesla and SpaceX have separately floated plans for a new semiconductor factory to help supply the same AI buildout.
What makes the Nvidia OpenAI deal distinct in that race is the direct chip-supplier-to-customer financing structure. Instead of OpenAI negotiating a lease independently and then placing a separate GPU order, Nvidia is financing the building that will house its own hardware. That model could become more common if it works, since it lets a chip maker guarantee demand for its next several product generations while helping a key customer avoid raising as much outside capital.
What Are the Key Milestones to Watch Between Now and 2028?
Because this is a multi-year build-out rather than a single event, the most useful way to track the Nvidia OpenAI deal is by milestone rather than by headline. These are the checkpoints that will show whether the campus is on schedule.
- Site preparation and permitting for the PORTS-Pike Technology Campus in Pike County, Ohio, which typically includes local zoning approval and utility interconnection agreements.
- Construction of the first phase of the shell buildings and power infrastructure, which is the portion Nvidia’s financing directly underwrites.
- Delivery and installation of Nvidia GPU hardware, expected to track Nvidia’s ongoing chip generation releases through 2030.
- The first 800 megawatts going live, targeted for 2028, which will be the first real test of whether the project is on schedule.
- Expansion toward the full 4.25-to-8-gigawatt target, contingent on OpenAI exercising its option for the additional 3.75 gigawatts of capacity.
Each of these milestones is also a point where the deal could slow down or change shape. Permitting delays, grid capacity disputes with Ohio utilities, or a shift in OpenAI’s own compute strategy could all push the 2028 target back, in the same way large data center projects from other hyperscalers have slipped in past years.
What Happens If OpenAI Cannot Meet Its Lease Payments?
This is the scenario that makes the financing structure of the Nvidia OpenAI deal worth understanding in detail. Because Nvidia is backing defined portions of OpenAI’s lease and power payments rather than gifting cash outright, Nvidia’s downside is a guarantee obligation, not a sunk investment. If OpenAI’s revenue fell short and it could not cover its 20-year lease to SB Energy, Nvidia’s financing backstop would be what keeps the campus’s debt serviced, at least for the portions it agreed to cover.
That is precisely why analysts describe this as risk-sharing rather than risk elimination. Nvidia has strong incentives to make sure OpenAI succeeds, since Nvidia’s own GPU order book depends on it, but the arrangement does concentrate financial exposure between two companies instead of spreading it across a wider pool of investors, which is a structural difference from how data centers were typically financed before the current AI buildout.
What Are the Risks and Criticisms of the Deal?
The most common criticism is that this is a circular financing arrangement: Nvidia is effectively bankrolling the construction of a facility that will buy Nvidia’s own chips, which lets both companies book activity that looks like growth without new outside capital entering the system. Critics point out that if OpenAI’s revenue growth ever slows, Nvidia is left holding financing exposure tied to a single, enormous customer, concentrating risk that would normally be spread across many buyers.
There is also a simple scale problem. Building 4.25 to 8 gigawatts of new capacity in one region means securing an amount of electricity comparable to a mid-size city’s entire grid draw, and Ohio utilities and regulators will need to plan around that demand for years. Reporting from Fortune (August 18, 2026) noted that the finalized deal came in well below some of the eye-popping figures floated earlier in the summer, which it read as a signal that markets were pricing in more “artificial demand” for chips than the actual contract supports. That gap between rumored and confirmed numbers is itself a caution sign for anyone treating every AI infrastructure headline as final.
What Does the Nvidia OpenAI Deal Mean for AI Pricing and Access?
In the short term, the deal does not change what you pay for ChatGPT or the OpenAI API; the Ohio campus will not have meaningful capacity online until 2028 at the earliest. In the medium term, more guaranteed GPU supply for OpenAI should ease some of the capacity constraints that have throttled API rate limits and new-feature rollouts industry-wide, a dynamic that also affects how fast tools like AI coding assistants can ship new capabilities to developers.
The bigger picture is that compute has become the real bottleneck in AI, ahead of model architecture or training data. Whoever locks in the most reliable, largest-scale power and chip supply gets to ship faster and price more competitively, which is exactly what deals like this one are designed to secure years in advance. For everyday users, the practical takeaway is less about this week’s headline and more about watching whether AI subscription prices and rate limits loosen once campuses like this one actually start coming online in 2028 and beyond.
Frequently Asked Questions
How much is the Nvidia OpenAI deal actually worth?
Nvidia is backing up to $105 billion in financing for the Ohio data center’s land, power, and lease payments, according to the companies’ August 17, 2026 announcement. That figure is a financing ceiling tied to construction milestones, not a single upfront cash payment, so the amount Nvidia actually pays out will depend on how the campus is built over time.
Is Nvidia buying a stake in OpenAI with this deal?
No. This specific agreement is project financing for the Ohio campus, separate from any equity stake. It sits alongside, not instead of, Nvidia’s broader up-to-$100-billion systems partnership with OpenAI announced in 2025, which covers deployment of Nvidia hardware rather than ownership of OpenAI itself.
When will the Ohio data center come online?
The first 800 megawatts of capacity is expected online by 2028, with the full 4.25-to-8-gigawatt build-out continuing into the following years, per Nvidia and OpenAI’s official announcement. Large data center campuses typically ramp up capacity in phases rather than opening all at once.
Will this deal affect ChatGPT subscription prices?
Not immediately. The new capacity will not be operational until 2028 at the earliest, so it has no near-term effect on ChatGPT or API pricing. Over the long run, more secured compute supply could help stabilize costs and rate limits as demand keeps growing.
Why is this called a “circular” deal by critics?
Because Nvidia is financing infrastructure that will, in turn, be filled with Nvidia GPUs purchased by OpenAI, meaning money and hardware orders flow in a loop between the two companies rather than involving new third-party capital entering the system.
What other companies are involved in the Ohio project?
SB Energy is the developer that will build and own the PORTS-Pike Technology Campus. Nvidia provides the financing backing, and OpenAI is the 20-year tenant that will operate the site’s AI workloads once construction phases are complete.
The Nvidia OpenAI deal is ultimately a bet by both companies that AI compute demand keeps climbing for the rest of the decade. Whether that bet pays off will depend on how fast OpenAI’s revenue grows relative to the financing it now owes, and how quickly rivals building their own gigawatt-scale campuses, from Google to Meta to Amazon, can catch up.
Subscribe for Newsletter
