OpenAI stands on the verge of securing one of the largest single-site computing commitments in history. The company is in advanced negotiations to lease a proposed 10-gigawatt data center campus in southern Ohio. This facility, if realized, would dwarf existing AI infrastructure projects. Its scale demands attention from executives, investors and policymakers alike.
The campus would rise on a mix of federal land managed by the Department of Energy and adjacent private parcels in Pike County. Developer SB Energy, a unit of SoftBank, would lead construction. OpenAI would take control of the computing equipment under a 20-year lease. Rent payments kick in only after operations commence. The first 800-megawatt phase targets 2028. Full build-out could run at least $500 billion at today’s prices for chips, power infrastructure, labor and materials. Reuters reported these details, citing people with direct knowledge.
Nvidia plays a central role. It would supply the hardware. It would also extend financial guarantees for both OpenAI’s lease obligations and SB Energy’s project financing. Such backing reduces risk for all parties. It also tightens the strategic knot between the two AI leaders. Their September 2025 partnership already outlined plans for at least 10 gigawatts of Nvidia systems across OpenAI’s footprint. This Ohio site would form a cornerstone.
Power sits at the heart of the challenge. Ten gigawatts equals the output of roughly ten large nuclear reactors. Current U.S. data centers consumed about 183 terawatt-hours in 2024, or over 4 percent of national electricity. Projections show that figure more than doubling by 2030. AI-specific servers drive much of the surge. Their consumption grows at 30 percent annually in baseline forecasts. The International Energy Agency laid out these numbers.
But. Reliability matters more than raw volume. AI training and inference run nonstop. Intermittent renewables fall short without massive storage. Natural gas offers speed but carries emissions and price volatility. Nuclear delivers steady baseload power. Tech companies have signed deals to restart retired plants and fund small modular reactors. Goldman Sachs analysts estimate 85 to 90 gigawatts of new nuclear capacity may be needed globally to meet data-center growth alone. Their January 2025 research spelled out the gap.
OpenAI’s move reflects urgency. The company paused a major U.K. data-center plan earlier this year over regulatory delays and energy costs. Domestic sites face their own hurdles. Grid interconnection queues stretch for years. Transmission upgrades cost billions and spark local opposition. Ohio’s location on former uranium-enrichment grounds offers historical ties to federal energy infrastructure. That pedigree may ease some permitting. Still, building at this magnitude will test every assumption about supply chains, skilled labor and capital formation.
Costs look staggering on paper. Five hundred billion dollars exceeds the annual GDP of many countries. Chip prices fluctuate with demand. Power-purchase agreements must lock in rates for decades. Construction inflation has already hit data-center projects hard. Yet the economics rest on explosive AI adoption. Inference workloads, once a small slice, now surge at 122 percent compound annual growth rates through 2028. Each new model generation demands more compute. Revenue from enterprise subscriptions, API calls and future agentic systems must cover the tab.
Sam Altman has spoken openly about the capital required. In earlier interviews he described the path to superintelligence as needing unprecedented infrastructure. The Ohio project aligns with that vision. It also fits SoftBank’s long bet on OpenAI. Masayoshi Son’s firm sees exponential returns if artificial intelligence transforms industries from drug discovery to software engineering.
Investors watch closely. Nvidia shares reacted modestly to the initial reports, reflecting broad expectations that the company would anchor such deals. Microsoft, OpenAI’s largest backer, maintains its own massive data-center commitments. The Ohio lease structure lets OpenAI avoid full balance-sheet ownership while securing capacity. Payments begin only when revenue-generating models run on the hardware. That alignment helps. It does not eliminate execution risk.
Regulatory questions linger. Antitrust authorities examine the Nvidia-OpenAI relationship. Energy regulators must approve the massive draw on regional grids. Environmental groups will scrutinize water usage for cooling and any nuclear components. The Department of Energy’s involvement could streamline approvals but also invites congressional oversight. Federal land brings additional layers of review.
Competitors pursue parallel paths. Anthropic secured $35 billion in financing from Apollo and Blackstone for its own expansion, using Broadcom silicon. Google, Amazon and Meta each announced nuclear restarts or new reactor deals. The race for power has become as intense as the race for talent or algorithms. One analyst called it the new arms race in bits and electrons.
Success in Ohio would mark a milestone. It would signal that private markets can mobilize half a trillion dollars for a single computing site. It would demonstrate that hyperscale AI infrastructure can be financed through leases, guarantees and long-term power contracts rather than pure equity. Failure or delay would underscore the physical constraints that software optimism sometimes overlooks. Grids do not scale with prompts. Reactors cannot be spun up overnight.
And the timeline is tight. First power in 2028 leaves little margin for setbacks. Chip deliveries, transformer manufacturing and skilled trades must all align. Global competition for the same equipment intensifies. Taiwan and South Korea remain critical nodes in the semiconductor supply chain. Geopolitical tensions could disrupt flows at any moment.
OpenAI’s discussions reveal a maturing industry. Early experimentation with cloud rentals has given way to bespoke, gigawatt-scale campuses purpose-built for training frontier models. The Ohio project, if completed, would become the company’s largest facility by a wide margin. It would consume electricity equivalent to several major metropolitan areas. Its output, measured in intelligence rather than lumens or horsepower, could reshape entire economic sectors.
Analysts at the RAND Corporation modeled exponential growth scenarios. Their work shows AI data-center power demand could require 10 gigawatts of new capacity in the United States in 2025 alone. That matches the scale of this single Ohio proposal. The numbers are no longer abstract. They translate into concrete bids for land, fuel and financing.
Whether the deal closes remains uncertain. Both OpenAI and Nvidia declined immediate comment to reporters. SB Energy has yet to issue a formal statement. Yet the outlines have already shifted industry expectations. A 10-gigawatt anchor tenant gives developers confidence to invest in transmission, substations and cooling plants. It gives utilities visibility into multi-decade demand. It gives policymakers a tangible example of what artificial intelligence actually requires beyond clever code.
The story is still unfolding. New details may emerge on power sources, exact financing terms or additional partners. For now the proposal stands as a stark illustration. Compute abundance does not come cheap. It demands capital on a scale once reserved for national infrastructure programs. And it forces a reckoning with the physical world that underpins the digital one. Short sentences capture the stakes. Longer ones reveal the interlocking dependencies. The next few quarters will test whether ambition outruns infrastructure reality. Or whether the pieces fall into place, powering a new era of machine intelligence from the fields of southern Ohio.
WebProNews is an iEntry Publication