OpenAI Ohio Data Center: 8 GW Compute Campus Backed by Nvidia

OpenAI has locked in a data center lease so large it redefines what "scale" means for AI infrastructure. The 8-gigawatt campus in Pike County, Ohio, backed by Nvidia financing and built by SoftBank's SB Energy, will come online in phases starting in 2028. For builders, this signals a shift toward industrial-scale compute campuses that could reshape model training economics and data center geography.

How the deal is structured

OpenAI signed a 10-year lease for the facility, which will be built and owned by SB Energy. Nvidia is providing up to $105 billion in financing to support credit and compute, securing its role as the exclusive AI compute provider at the PORTS-Pike Technology Campus. SB Energy will manage the site under a 20-year lease to OpenAI and plans to invest at least $4.2 billion in regional grid infrastructure.

The initial capacity target is 4.25 gigawatts, with an option to scale to 8 gigawatts total. Nvidia is also investing $1.5 billion directly in SB Energy, further intertwining the chipmaker with the project's long-term viability.

What 8 gigawatts of compute means

Eight IT-gigawatts is an order of magnitude beyond any single AI data center announced to date. For comparison, a large existing facility might run at 100-200 megawatts. The campus will be powered by 10 gigawatts of new energy generation, including up to 9.2 gigawatts of gas-fired power, with funding tied to a 2025 U.S.-Japan trade deal.

For builders, this scale directly affects model training timelines and inference costs. A campus this large can train multiple frontier models simultaneously or handle massive inference loads for products like ChatGPT and future agent systems. The compute is owned by OpenAI under a long-term lease, which gives the company predictable access to GPU clusters without the capital outlay of building its own chip supply.

Why this matters for AI builders

The OpenAI Ohio data center signals that the frontier of AI infrastructure is moving toward vertically integrated, hyperscale campuses owned by a small set of players. This has two practical implications:

First, compute concentration risk increases. If OpenAI's next-generation models are trained exclusively on this campus, any power or cooling disruption affects the entire pipeline. Second, access to Nvidia's latest hardware will be tiered. Nvidia's exclusive provider role means OpenAI gets first or exclusive access to new GPU generations, widening the gap between frontier labs and smaller shops.

For teams building on OpenAI's API, the long-term effect may be lower inference prices as training costs amortize over more capacity, but also reduced bargaining power as OpenAI controls its own hardware supply.

Caveats and open questions

Several details remain uncertain. The $105 billion figure comes from Nvidia's announcement, but early reports discussed a $250 billion backstop that was later reduced. The final terms of the Nvidia guarantee and exact capacity phasing are not public.

Regulatory approval for the gas-fired power buildout is not guaranteed, and the project sits on a former uranium-enrichment site, which may carry environmental remediation obligations. Nvidia CEO Jensen Huang has publicly rejected claims of "circular" AI financing, but the interlocking ownership between OpenAI, SB Energy, and SoftBank raises valid questions about risk concentration.

For builders, the key takeaway is that this project, if completed, will create the most concentrated pool of AI compute in the world by 2028. Whether that benefits the broader ecosystem or entrenches existing power structures depends on how OpenAI chooses to make that capacity available through its platform.