NVIDIA is backing OpenAI’s Ohio AI campus located at PORTS-Pike, aligning hardware supply with hyperscaler site development.
NVIDIA has partnered with SB Energy to secure land, power, and shell capacity at the PORTS-Pike Technology Campus in Ohio for AI infrastructure, with OpenAI designated as the primary tenant. The agreement covers an initial 4.25 gigawatts at the Portsmouth campus, where OpenAI plans to build and operate an AI facility utilizing NVIDIA GPUs, CPUs, networking hardware, and infrastructure software. NVIDIA retains the option to extend the arrangement to encompass the site’s remaining 3.75 gigawatts.
This development underscores how access to land, electricity, and physical data centre space has emerged as a critical constraint, running parallel to traditional bottlenecks in chips, packaging, memory, and networking as demand for large-scale AI computing accelerates. In response, NVIDIA is applying the same supply-chain management strategies it has honed for semiconductors to these foundational infrastructure inputs.
Under the disclosed structure, NVIDIA will underwrite portions of the lease and power commitments for approximately 4 gigawatts over a 20-year term, with the site exclusively reserved for NVIDIA compute workloads. The guarantee does not cover the full cost of the campus or all of the tenant’s financial obligations. Instead, the support will phase in as individual data centres come online between 2028 and 2030. While OpenAI remains responsible for lease payments, NVIDIA is leveraging its market scale and visibility into customer demand to secure a location capable of hosting multiple generations of its systems. As capacity enters service and lease payments are fulfilled, NVIDIA’s financial exposure will correspondingly decrease.
NVIDIA characterizes the Ohio facility as a long-life asset designed for repeated upgrades as newer generations of compute hardware are deployed. Each successive generation installed at PORTS-Pike could require approximately 1.5 million NVIDIA GPUs and generate between USD $150 billion and USD $200 billion in revenue for the company. Over a two-decade horizon, the campus is expected to accommodate several upgrade cycles. The strategic emphasis remains on securing a permanent physical footprint with guaranteed long-term access to power and real estate, while continuously refreshing the computing equipment housed within.
On a broader scale, NVIDIA outlined that OpenAI’s existing and planned commitments represent roughly 12 gigawatts of NVIDIA compute through 2030. Should the Ohio arrangement expand beyond the initial 4.25 gigawatts, total exposure could rise to approximately 16 gigawatts. At that threshold, the opportunity would translate to roughly USD $600 billion in compute spend through 2030. These figures highlight the unprecedented capital intensity now concentrated among a select group of large AI customers as the industry races to secure data centre capacity and electricity supply.
Despite this intervention, most NVIDIA customers—including cloud service providers, enterprises, and state-backed AI initiatives—will continue to independently secure their own land, power, and shell capacity. These entities typically possess the balance sheets and financing structures required to execute long-term infrastructure contracts directly. NVIDIA’s involvement is specifically targeted at frontier AI laboratories, whose surging demand for training and inference computing is outpacing their ability to secure long-dated infrastructure on their own. According to NVIDIA, these organizations may demonstrate strong revenue growth and customer traction but often lack the financial profile necessary to underwrite large-scale site commitments.
NVIDIA maintains that this model does not constitute circular financing, emphasizing that OpenAI remains fully liable for lease payments. Rather, the structure is designed to lock in scarce infrastructure where NVIDIA has high confidence in sustained long-term demand for its computing systems. A critical component of this thesis is that the site will retain intrinsic value even if a single tenant departs. Capacity at PORTS-Pike could be reassigned to other clients across NVIDIA’s extensive customer base—which spans cloud providers, enterprises, AI laboratories, and start-ups—because the underlying hardware is widely adopted and readily redeployable.
NVIDIA ties this flexibility to the breadth of its software and developer ecosystem, noting that platform standardization significantly enhances the reuse of installed compute assets across different customers over time. This dynamic reinforces the company’s position that long-term site commitments should be viewed as productive, multi-tenant infrastructure rather than a single-customer wager.
The Ohio project signals a further evolution in NVIDIA’s role, transitioning from a pure chip supplier to an active participant in financing and securing the physical foundations required for AI deployment. Moving forward, the company will remain highly selective, concentrating only on what it describes as exceptional sites where sustained demand can reliably support multiple generations of NVIDIA compute.