NVIDIA has secured a dedicated allocation of low‑power semiconductor (LPS) capacity through a partnership with SB Energy at the PORTS‑Pike Technology Campus in Portsmouth, Ohio. The reservation includes not only the silicon die but also the associated land, power delivery, and shell infrastructure required to operate an AI factory. The site will host NVIDIA‑accelerated compute hardware, and OpenAI has been named the initial tenant to run its training and inference workloads on the reserved LPS resources.
This approach mirrors NVIDIA’s long‑standing practice of using its scale, multi‑year visibility, and supply‑chain partnerships to lock down critical semiconductor resources, a tactic already employed by major cloud providers and investment‑grade enterprises that can finance LPS independently through balance‑sheet strength and extended contracts. The new agreement is specifically intended for frontier AI labs that, despite exhibiting strong revenue trajectories, lack the credit history or balance‑sheet depth to negotiate multi‑year LPS commitments on their own, thereby providing them with pre‑secured compute infrastructure that includes the necessary land, power, and shell components.
- Location: PORTS‑Pike Technology Campus, Portsmouth, Ohio
- Partner: SB Energy (provides LPS capacity, land, power, shell)
- Purpose: Host NVIDIA AI‑factory compute (training & inference)
- Initial tenant: OpenAI
- Intended beneficiaries: Frontier AI labs lacking independent LPS procurement capability
Why this matters
Factually, NVIDIA has secured LPS capacity at the PORTS‑Pike site via SB Energy, with OpenAI as the first tenant. This indicates that the company is extending its infrastructure‑securing model to a site outside its traditional supply‑chain hubs. Inference: By making LPS, land, power, and shell available as a pre‑bundled package, the arrangement reduces the upfront infrastructure hurdle for frontier AI labs that cannot independently negotiate long‑term contracts, potentially shortening their time‑to‑train and deploy large models.