OpenAI Partners with Nvidia and SoftBank for Massive Ohio AI Infrastructure Project

The facility will be developed and owned by SoftBank's SB Energy, with Nvidia underwriting the deal financially and supplying the underlying compute infrastructure. The configuration marks a departure from the conventional hyperscaler build pattern, where the model lab typically controls the capex stack end-to-end.

Deal structure and ownership topology

SB Energy, SoftBank's dedicated infrastructure subsidiary, will hold the physical asset; Nvidia positions as both financial backer and primary silicon supplier. OpenAI effectively becomes the anchor tenant rather than the capital allocator — a setup that decouples compute provisioning from balance-sheet exposure on OpenAI's side while concentrating infrastructure capex risk within SoftBank and the silicon dependency within Nvidia's roadmap.

The Ohio build lands against a backdrop of aggressive AI infrastructure spending across the sector, with Oracle reportedly planning fresh layoffs as it absorbs massive AI infrastructure costs, according to The Indian Express. The pattern indicates that the capex commitments reshaping capital priorities at incumbents are not isolated to OpenAI's orbit; balance sheets at hyperscalers, dedicated infrastructure vehicles like SB Energy, and chip vendors acting as both suppliers and financiers are all being repositioned in parallel.

Specifications still undisclosed

The critical unknowns for downstream observers are compute density, the Nvidia silicon generation targeted for deployment, the energization timeline, and whether the campus capacity will be reserved exclusively for OpenAI workloads or resold across the broader market. Subsequent disclosure on interconnect topology, memory bandwidth targets, or the site's power envelope would materially clarify the throughput ceiling available for large-scale training runs and high-throughput inference workloads.

Absent those figures, the deployment can only be benchmarked structurally. The Ohio project shifts financial risk upstream to SoftBank while binding Nvidia's silicon supply more tightly to a single hyperscale tenant. That dependency has direct operational consequences: any silicon refresh or supply disruption ripples downstream into OpenAI's training calendar and the inference budgets of its enterprise tier.

What to track next

For developers and operators, the practical signal is that OpenAI is locking in long-duration compute supply outside its own capex stack — a hedge against the GPU allocation volatility that has constrained training schedules across the industry. The capital profile of this class of deal — multi-billion-dollar infrastructure commitments underwritten across non-traditional partner stacks — has portfolio implications at every scale, from hyperscalers committing to long-tenor compute contracts down to questions of financial independence for individual investors positioning around the AI economy's compounding capex cycle.

Watch for the silicon generation specified at site energization, the announced power capacity in megawatts, and any reference to OpenAI's reservation share versus SB Energy's right to lease residual capacity to third parties. Those three data points together will determine whether the campus functions as a closed OpenAI enclave or a multi-tenant compute hub.