OpenAI is in advanced negotiations to lease a 10-gigawatt data center campus on federal land in Ohio, according to The Information as reported by Seeking Alpha. The full buildout could exceed $500 billion, making it one of the largest single infrastructure projects in the history of the technology industry. Nvidia is reportedly backing the project, though financing details have not been finalized.
Lease, Not Own
The structure is notable: OpenAI would control the equipment on a long-term lease rather than owning the campus outright, according to Gotrade. Payments would begin once the site is operational, with the first phase expected in 2028. That structure limits OpenAI’s balance sheet exposure while locking in compute capacity at a scale that no other AI lab has publicly pursued.
Ten gigawatts is roughly the power consumption of a mid-sized U.S. city. For context, the entire U.S. data center industry consumed an estimated 17 gigawatts in 2022, according to the Department of Energy. OpenAI’s Ohio campus alone would represent a significant fraction of current national data center power demand.
The Agent Compute Bottleneck
The scale of the project reflects the compute demands of running autonomous AI agents at enterprise scale. Multi-step reasoning, tool use, and persistent agent sessions require sustained inference capacity that dwarfs the one-shot query model of early chatbots. Training frontier models remains expensive, but the emerging constraint is inference: keeping millions of agent sessions running simultaneously, each executing multi-turn workflows that can stretch over hours or days.
OpenAI’s Ohio pursuit follows a pattern. Anthropic secured $35 billion in chip financing from Apollo and Blackstone in recent weeks. Google’s parent Alphabet announced an $80 billion equity raise earmarked for AI infrastructure. Microsoft disclosed plans for 10 new data centers in its latest earnings call. The common thread: every frontier lab and hyperscaler has concluded that compute capacity, not model quality, is the binding constraint for the next phase of AI deployment.
Capital Questions
The timing coincides with a reported stall in SoftBank’s $6 billion margin loan collateralized against its OpenAI equity stake. Lenders balked at the valuation of an unproven private company. OpenAI simultaneously pursuing a $500 billion infrastructure commitment while its largest investor struggles to borrow against its stake illustrates the tension between the industry’s long-horizon capital planning and near-term financial discipline.
OpenAI filed a confidential S-1 with the SEC last week. The Ohio data center commitment, if finalized, would appear on the company’s disclosure as a major long-term lease obligation, giving public market investors their first detailed look at the capital intensity required to operate frontier AI infrastructure at scale.
The Federal Land Factor
The use of federal land introduces a government dimension. Federal leases typically involve interagency review, environmental assessment, and political considerations that private land deals do not. The project would place OpenAI’s core compute infrastructure on government-controlled property, a detail that could draw scrutiny from lawmakers already debating the appropriate relationship between AI companies and the federal government.