NVIDIA Backs OpenAI Ohio AI Factory, Sees $600 Billion Compute Opportunity
NVIDIA backs 4 gigawatts of Ohio AI factory capacity for OpenAI over 20 years, targeting up to 16 gigawatts and $600 billion compute opportunity by 2030.
Summary
On August 17, 2026, NVIDIA announced a partnership with SB Energy to secure land, power and shell capacity at the PORTS-Pike Technology Campus in Portsmouth, Ohio, exclusively for NVIDIA compute, with OpenAI as tenant. OpenAI will build and operate the AI factory using NVIDIA’s DSX platform, including GPUs, CPUs, networking and infrastructure software. The initial deployment is expected to provide 4.25 gigawatts. Each system generation could include about 1.5 million NVIDIA GPUs and represent $150 billion to $200 billion in NVIDIA revenue, while the site can support multiple upgrade cycles over 20 years. NVIDIA may secure the remaining 3.75 gigawatts.
OpenAI’s existing and planned NVIDIA infrastructure commitments through 2030 total about 12 gigawatts, potentially rising to 16 gigawatts and roughly $600 billion of NVIDIA compute if PORTS-Pike expands. NVIDIA is supporting approximately 4 gigawatts for 20 years through defined portions of lease and power payments plus a specified residual value commitment, not the site’s full cost or all OpenAI obligations. Guarantees activate as data centers enter service from 2028 to 2030, then decline as OpenAI pays rent and capacity opens. Most cloud providers, enterprises, sovereign AI builders and other customers will continue securing their own infrastructure, while NVIDIA selectively assists frontier AI labs constrained by financing and compute availability. If OpenAI leaves, NVIDIA says CUDA compatibility and broad adoption allow capacity to be offered to another qualified tenant.
Positives
- 4.25 gigawatts of initial PORTS-Pike capacity would host OpenAI’s NVIDIA DSX based AI factory in Portsmouth, Ohio.
- 1.5 million NVIDIA GPUs per system generation could represent $150 billion to $200 billion in NVIDIA revenue.
- 12 gigawatts of OpenAI commitments could expand to 16 gigawatts and roughly $600 billion of NVIDIA compute through 2030.
- 20 years of site availability could accommodate multiple NVIDIA hardware upgrade cycles without replacing the underlying infrastructure.
- 3.75 gigawatts of additional PORTS-Pike capacity remains available if NVIDIA extends its initial arrangement.
Risks & concerns
- Frontier AI labs can outgrow their balance sheets and credit profiles, leaving compute expansion constrained despite strong demand and rising revenue.
- Land, power and shell availability has become a critical constraint on deploying AI factories.
- 20 years of support exposes NVIDIA to defined lease, power and residual value commitments, although it excludes the full site cost and all tenant obligations.
- 2028 to 2030 service dates create a long deployment schedule before all planned PORTS-Pike capacity becomes operational.
- OpenAI nonuse would require NVIDIA to find another qualified tenant across its cloud provider, enterprise, AI lab or startup ecosystem.