Story
August 17, 2026

Nvidia’s $105 Billion Ohio Gamble Puts AI’s Power Bill in the Spotlight

Nvidia’s backing for OpenAI’s planned Ohio campus promises jobs and vast computing capacity, but its gas-heavy power plan and complex financing are sharpening fears about costs and an AI bubble.

Nvidia is betting that the next phase of artificial intelligence will be built not just on chips, but on an industrial-scale promise to secure land, power and financing. In Ohio, that promise carries a price tag large enough to revive concerns over energy bills and AI’s increasingly intertwined balance sheets.

The project centers on OpenAI’s planned Ports-Pike data center near Cincinnati, to be built and owned by SoftBank-linked SB Energy. Nvidia said it would invest $1.5 billion in the developer and become the campus’s exclusive compute-infrastructure supplier, while offering up to $105 billion in credit support for construction and related obligations. The site could grow from 4.25 gigawatts of capacity to 8 gigawatts.

OpenAI then announced a 10-year lease, framing the campus as a jobs and infrastructure play: 35,000 construction jobs and 2,500 permanent operating roles, with initial capacity expected in 2028. Supporters see a template for building the vast computing backbone AI companies say they need.

But the scale of the power plan is the political and financial pressure point. SB Energy intends to build at least 10 gigawatts of new generation, including a proposed 9.2-gigawatt natural-gas plant on a former federal uranium-enrichment site. That plant alone is expected to cost $33 billion; gas-plant construction costs have risen 66% in two years, and future competition with export markets could push gas prices sharply higher in some regions.

The financing has also drawn scrutiny because Nvidia supplies the chips, invests in the developer and guarantees major lease and power-payment obligations. Nvidia chief executive Jensen Huang rejected the circular-financing critique: “Is this circular financing? No. OpenAI will pay the lease.” The dispute is over more than semantics: proponents call it long-term infrastructure underwriting, while critics see a high-stakes chain of commitments that could magnify any stumble in AI demand.