OpenAI and NVIDIA have partnered to build a massive AI factory at the PORTS-Pike Technology Campus in Portsmouth, Ohio, backed by an initial 4.25-gigawatt deployment. Scheduled to scale between 2028 and 2030, the multi-gigawatt infrastructure initiative anchors a broader commitment representing up to 16 gigawatts of NVIDIA compute and roughly $600 billion through the end of the decade.
The Bottleneck Shifts From Silicon to Land, Power, and Shell
Modern frontier AI labs face a structural capital mismatch. Their demand for massive large language model (LLM) training and low-latency inference compute scales exponentially, but their balance sheets and credit profiles cannot always match the multi-decade capital expenditure required to build physical data centers from scratch.
Compute is revenue in the modern tech economy.
Most mega-scale cloud service providers and investment-grade enterprises possess the balance sheets to secure this real estate independently. Frontier labs, however, often outpace their own long-term financing capabilities. To bypass this friction, NVIDIA is applying the same disciplined supply-chain management it uses for advanced node packaging and memory allocation directly to real estate and energy infrastructure.
Anatomy of the PORTS-Pike Deployment
The newly announced collaboration targets the exceptional PORTS-Pike Technology Campus in Portsmouth, Ohio. OpenAI will act as the primary tenant, designing, building, and operating an industrial-scale AI factory powered by NVIDIA’s full-stack DSX platform. This includes tightly integrated racks of GPUs, CPUs, high-speed networking switches, and proprietary infrastructure management software.
The numbers behind the site are staggering. The initial deployment aims for 4.25 gigawatts of continuous power capacity. According to infrastructure projections, each distinct hardware generation deployed at PORTS-Pike can house approximately 1.5 million NVIDIA GPUs, representing an estimated $150 billion to $200 billion in hardware revenue per cycle. Because the facility is engineered to support multiple upgrade paths over a 20-year span, the site acts as a permanent generational anchor.
Furthermore, NVIDIA retains the option to expand the arrangement at PORTS-Pike beyond the initial 4.25 gigawatts, tapping an additional 3.75 gigawatts of remaining site capacity. Combined with existing and planned projects, OpenAI’s total commitments to NVIDIA infrastructure through 2030 span roughly 12 gigawatts, with a ceiling of 16 gigawatts if the Ohio site fully expands.
Structuring Risk: Guarantees and Residual Value
NVIDIA has clarified the specific legal and financial boundaries of its exposure at PORTS-Pike. The company is supporting the site’s LPS infrastructure for approximately 4 gigawatts over a 20-year term, restricting its backing to defined segments of lease and power payments alongside a targeted residual-value commitment. NVIDIA is not footing the total bill for the site or covering every operational obligation of the tenant.
These guarantees will phase in sequentially as individual data center pods come online between 2028 and 2030. As OpenAI executes its scheduled lease payments and operational capacity scales up, NVIDIA’s financial exposure systematically steps down.
What happens if market dynamics shift or a tenant pulls back? NVIDIA points to the inherent fungibility of its hardware ecosystem. Because the entire hardware stack is unified under the CUDA programming model—allowing developers to optimize code across generations—the underlying compute infrastructure remains remarkably versatile. If necessary, high-density AI clusters can be reallocated across NVIDIA’s global ecosystem of cloud providers, sovereign AI builders, enterprises, and startups. Exceptional real estate paired with standardized accelerated computing architecture creates a liquid, financeable asset class.
The Infrastructure of Intelligence
NVIDIA’s evolution from an independent graphics card manufacturer into a full-stack data center architect is now complete. By stepping directly into land, power, and shell logistics for strategic partners, the company is actively engineering the supply lines of the intelligence economy.
