OpenAI has signed onto one of the largest AI infrastructure projects now planned in the United States, securing about 8 gigawatts of IT capacity at the PORTS-Pike Technology Campus in Pike County, Ohio, with NVIDIA and SoftBank-backed SB Energy shaping the project around NVIDIA compute.
The August 17 announcement is not simply another large data center lease. NVIDIA is investing $1.5 billion in SB Energy and providing credit support tied to land, power, and shell buildout for the initial 4.25 gigawatts of IT capacity, with an option to extend to the remaining 3.75 gigawatts. SB Energy will build, own, and operate the site under a 20-year lease to OpenAI, while the campus is expected to host NVIDIA AI compute exclusively.
That structure matters because the bottleneck in frontier AI is moving beyond GPUs alone. The companies are treating land, electricity, transmission, building shells, cooling design, financing, and cluster reliability as one supply chain. OpenAI is buying future capacity, NVIDIA is helping make that capacity financeable, and SB Energy is turning a former federal uranium-enrichment site into a power-first AI campus.
What the companies announced
OpenAI said it has entered an agreement to secure approximately 8 gigawatts-IT at PORTS-Pike, working with SB Energy, NVIDIA, and the U.S. Department of Energy. The first 800 megawatts are expected to become available in 2028, largely using existing AEP infrastructure. The full project will be built in phases rather than arriving all at once.
OpenAI will begin paying only when completed capacity becomes available for lease. The company framed the commitment as a response to long-term frontier model training needs and growing demand for products such as ChatGPT and Codex. It also said NVIDIA and OpenAI are collaborating through the design, testing, and commissioning process, with a technical white paper planned on resilient infrastructure design, component qualification, and software-level workload management at cluster scale.
NVIDIA’s role is broader than supplying accelerators. The company is expected to be the exclusive AI compute infrastructure provider at the campus and to support the land, power, and shell capacity needed to bring the first 4.25 gigawatts online. In NVIDIA’s framing, the data center will use the company’s full-stack DSX AI factory platform, including GPUs, CPUs, networking, and related systems.
Why this is different from a normal cloud deal
Cloud contracts usually hide the physical machinery behind a service agreement. PORTS-Pike puts the physical layer in the foreground. OpenAI is not only reserving compute; it is entering a long-term relationship with a developer that must secure land, transmission, power generation, permits, environmental reviews, and financing before all of the capacity exists.
That is why NVIDIA’s participation is significant. A chip supplier that helps back the buildout around its own infrastructure can make it easier for the project developer to raise money, order long-lead electrical equipment, and commit to a campus sized around future AI demand. It also tightens the commercial loop: OpenAI needs compute, SB Energy needs capital and project certainty, and NVIDIA benefits if more AI factory capacity turns into more NVIDIA systems deployed at scale.
For enterprise readers, the useful lesson is that AI capacity planning is becoming infrastructure planning. Model performance, latency, and product availability increasingly depend on site selection, grid interconnection, power-market exposure, cooling constraints, and whether the provider can keep massive clusters reliable enough to run high-value workloads.
The power plan is part of the story
The Department of Energy had already described the Portsmouth site as part of a public-private effort to combine data center development with new energy infrastructure in southern Ohio. In a March fact sheet, DOE said SoftBank Group and SB Energy planned to build 10 gigawatts of new power generation, including at least 9.2 gigawatts of natural gas generation, to support 10 gigawatts of data center development. SB Energy and AEP Ohio also planned $4.2 billion in transmission infrastructure, including new high-voltage lines and substations.
OpenAI’s August update puts those earlier power claims into a more concrete customer context. The company said SB Energy will pay the full cost of grid upgrades and new transmission lines needed to serve the data center, and that those costs will not be shifted to Ohio or regional ratepayers. It also said the project will publicly report expected water use once the site design is finalized.
Those promises will need public follow-through. AI data centers are drawing more scrutiny because they can concentrate electricity demand in places where grid capacity, generation mix, water use, ratepayer exposure, and local land-use politics all collide. PORTS-Pike tries to answer that concern by pairing the data center with new generation, transmission investment, annual reporting, and local community commitments. The test will be whether those commitments survive permitting, financing, phased construction, and real operating conditions.
What OpenAI is offering Ohio
OpenAI is pairing the capacity agreement with a community-benefits package. The company expects 35,000 construction jobs over a six-year buildout through 2032 and 2,500 long-term operating jobs. It also pledged $40 million for a community grant fund, building on SB Energy’s separate $40 million commitment.
The company is also offering up to $84 million in Codex credits through ChatGPT for roughly 844,000 eligible Ohio college, community college, and technical school students aged 18 and older during the 2026-2027 academic year. Eligible students are supposed to receive $100 in credits through their ChatGPT accounts.
That student offer is small compared with the scale of the data center, but it is politically useful and strategically consistent. OpenAI is trying to connect the physical buildout of AI infrastructure with workforce development, local legitimacy, and expanded use of its own tools. For a project placed on land tied to the former Portsmouth Gaseous Diffusion Plant, that local story is not incidental; it is part of how the companies are asking Ohio to accept a new kind of industrial campus.
The open questions
The announcement leaves several important questions unresolved. The full campus depends on new infrastructure, permits, environmental reviews, and financing. The first 800 megawatts are targeted for 2028, but the larger buildout runs through 2032. That makes execution risk a central part of the story, especially as AI companies race to lock in capacity years before all of the hardware, power, and buildings are available.
The economics also deserve watching. NVIDIA’s investment and credit support could accelerate a project that ultimately buys enormous amounts of NVIDIA infrastructure. That does not make the deal improper, but it does show how the AI boom is creating more circular dependencies among model labs, chipmakers, infrastructure developers, energy providers, and investors.
For OpenAI customers, the near-term impact is indirect. ChatGPT will not suddenly change because a campus was announced in Ohio. The practical effect comes later if the project gives OpenAI more reliable access to large-scale training and inference capacity, reduces dependence on constrained third-party cloud supply, and helps support heavier products such as coding agents, enterprise deployments, and future frontier models.
PORTS-Pike is best understood as a signal about where AI competition is going. The next phase is not only about who has the strongest model or the newest GPU. It is about who can secure enough power, buildings, financing, networked compute, and community permission to keep those models running at industrial scale.
