OpenAI’s Ohio AI Campus Turns a $500 Billion Vision Into a Financing Test

OpenAI is reportedly negotiating to lease a 10-gigawatt Ohio data center campus whose fully developed cost could surpass $500 billion. The proposal would place Nvidia hardware inside infrastructure developed by SoftBank-owned SB Energy. Nvidia could also provide financial support for the arrangement.

That combination makes the story larger than another Google News headline about record AI spending. OpenAI would control the computing equipment through a long-term lease without directly developing the entire campus. Nvidia could help secure both OpenAI’s lease obligations and SB Energy’s construction financing.

The result is an unusually concentrated test of the AI economy. One company would supply the chips, support the financing, and benefit when the tenant consumes more computing capacity. OpenAI would gain infrastructure without funding every component upfront.

The arrangement remains under negotiation, and its final structure has not been publicly confirmed. The reported $500 billion estimate assumes a complete buildout at current prices for chips, labor, energy, and construction materials.

A separate public project already gives the talks a physical foundation. Federal officials, SB Energy, and AEP Ohio broke ground in March 2026 on a 10-gigawatt technology campus near Piketon, Ohio.

The central conflict is therefore clear. OpenAI wants infrastructure on a scale that existing cloud contracts cannot easily provide. Yet reaching that scale requires financial guarantees, dedicated generation, new transmission, and confidence that demand will remain high for years.

The Ohio Project Is Real, but OpenAI’s Role Is Not Final

The physical campus has entered development, while OpenAI’s tenancy and Nvidia’s financial role remain reported negotiations.

The proposed campus sits at the Department of Energy’s Portsmouth Site in Pike County. The federal property formerly housed a gaseous diffusion plant used for uranium enrichment.

Federal cleanup continues at the location, but portions of the land are being released for industrial reuse. A project company affiliated with SB Energy has leased an initial 189 acres for AI and cloud infrastructure.

The Department of Energy documented that lease in an April 2026 site assessment. The assessment covers associated buildings, equipment, and systems planned for a phased AI campus.

SB Energy, the Department of Energy, and AEP Ohio held a groundbreaking event on March 20. Federal officials described the development as a planned 10-gigawatt data center paired with equivalent new generating capacity.

A gigawatt measures one billion watts of electrical power. A 10-gigawatt campus would therefore represent an industrial energy system, not a conventional collection of server buildings.

The first development phase covers 189 acres, while the broader campus would expand over time. The Department of Energy says construction should begin during 2026.

OpenAI was not identified as the campus tenant in the initial federal announcement. Later reporting connected the company to advanced lease negotiations with SB Energy.

According to data center reporting, OpenAI would control Nvidia equipment installed at the site through a long-term lease. Nvidia reportedly could guarantee OpenAI’s obligations and support financing for future construction.

None of those reported terms should be treated as a completed contract. OpenAI, Nvidia, and SB Energy have not publicly released a definitive lease covering the entire 10-gigawatt campus.

The distinction matters because several commitments are being compressed into one Google News narrative. The campus exists as a federal land and infrastructure project. OpenAI’s occupancy is a separate negotiation, while Nvidia’s potential guarantee adds another contractual layer.

The reported $500 billion figure also describes a possible full buildout. It is not a disclosed construction budget already committed to the Ohio site.

That estimate reflects current prices for accelerators, power systems, labor, and supporting infrastructure. Those inputs can change considerably before later phases begin.

Readers should consequently separate three facts. Ohio has a planned 10-gigawatt campus, OpenAI is reportedly considering a lease, and Nvidia is reportedly discussing financial support. Only the first has detailed public documentation.

Why OpenAI Wants Another 10 Gigawatts

OpenAI’s infrastructure strategy now treats computing capacity as a product constraint, not simply an operating expense.

The company has repeatedly argued that limited compute restricts the speed, availability, and capability of its services. More infrastructure would support model training, everyday inference, enterprise workloads, and products that demand sustained processing.

Inference is the computation used when a trained model answers a request. Its demand grows with users and usage, even when OpenAI is not training a new flagship model.

That creates a different capacity problem from a traditional software service. Better models can require larger clusters during development, while successful products generate continuing demand after release.

OpenAI’s answer has been to diversify beyond its historic dependence on Microsoft. Its infrastructure relationships now span Oracle, SoftBank, CoreWeave, Nvidia, and other hardware or energy partners.

The company introduced Stargate in January 2025 as a plan to invest $500 billion in United States AI infrastructure over four years. The initial announcement targeted 10 gigawatts of capacity across multiple locations.

By September 2025, OpenAI, Oracle, and SoftBank said their announced sites had brought planned Stargate capacity close to seven gigawatts. The locations included projects in Texas, New Mexico, Ohio, and another Midwestern site.

The earlier Ohio announcement concerned Lordstown, where SoftBank had broken ground on a facility designed to scale with other sites. That development is distinct from the much larger Portsmouth campus near Piketon.

OpenAI later said it was evaluating more locations beyond Stargate’s original goal. Its April 2026 infrastructure strategy framed expanded capacity as necessary for consumer, business, developer, and government demand.

The Portsmouth talks would advance that expansion through leasing. OpenAI could obtain operational control of equipment without acting as the landowner, utility, power developer, and construction manager.

Leasing also spreads obligations across time. However, it does not make the underlying expense disappear. Someone must finance the buildings, generators, transmission lines, cooling equipment, and accelerators before OpenAI can use them.

That is where Nvidia’s reported role becomes significant. A guarantee from the chip supplier could make lenders more comfortable financing a project whose tenant carries extraordinary long-term obligations.

The structure would also reinforce Nvidia’s demand. If OpenAI occupies the campus, the project would consume large quantities of Nvidia systems and supporting networking equipment.

OpenAI gains another route to compute, Nvidia gains a major equipment customer, and SB Energy gains a prospective anchor tenant. Each participant therefore helps reduce a different risk faced by the others.

This mechanism explains why the proposal has attracted more attention than routine Google News coverage of new data centers. It combines technology procurement with energy development and structured finance at an uncommon scale.

It also reveals OpenAI’s core challenge. The company wants to reserve infrastructure years before anyone can know precisely how efficient its future models will become.

Nvidia’s Backing Would Blur Supplier and Financier

The proposal tests whether AI demand can finance itself through tightly connected relationships among chipmakers, developers, and infrastructure owners.

Nvidia normally earns money by selling computing systems and related technology. A financial guarantee would place the company deeper inside its customer’s infrastructure obligations.

Under the reported arrangement, Nvidia could support OpenAI’s lease and SB Energy’s future project financing. That support might lower perceived risk for lenders and accelerate construction.

The benefit for Nvidia is direct. A completed OpenAI campus designed around Nvidia hardware could become one of the largest deployments of its systems.

The risk is equally clear. Nvidia would become more exposed to the financial capacity of a major customer whose purchases already contribute to demand for AI hardware.

This does not automatically make the arrangement unsound. Suppliers have long supported customers through credit, leases, purchase commitments, and project financing.

However, the scale changes the analysis. A reported guarantee tied to hundreds of billions in possible development would connect equipment demand to the supplier’s own financial strength.

Investors must then distinguish end-user demand from financing-assisted demand. The campus ultimately needs profitable workloads, not merely an ability to secure construction funding.

That question applies to OpenAI’s wider infrastructure network. The company has announced or pursued capacity with several partners, each using different ownership and financing structures.

Oracle supplies cloud infrastructure. Microsoft remains an important computing and commercial partner. CoreWeave specializes in rented accelerator capacity, while SoftBank combines investment ambitions with energy and infrastructure assets.

Nvidia sits across many of these relationships because its accelerators remain central to high-end AI computing. It benefits whether customers purchase systems, lease them, or rent access through cloud providers.

The Ohio proposal would give Nvidia another function. Its balance sheet could help transform OpenAI’s expected future payments into financeable infrastructure today.

That structure moves the primary competition away from a simple OpenAI-versus-Microsoft story. The more important conflict is OpenAI’s capacity promise versus the financing reality behind it.

OpenAI can sign several computing agreements, but every commitment depends on power, construction schedules, equipment delivery, and future revenue. Those dependencies do not vanish when different partners assume them.

The arrangement also creates concentration risk. If OpenAI reduces its required capacity, SB Energy could face an underused campus, while Nvidia could lose equipment demand and assume guarantee exposure.

If demand exceeds expectations, the structure looks prescient. OpenAI secures scarce capacity, Nvidia sells systems, and the developer expands later phases with an established tenant.

This uncertainty is why the phrase “could exceed $500 billion” needs context. It describes a possible endpoint under current cost assumptions, not a single check or completed investment decision.

The meaningful question is whether each phase can secure a tenant, power, financing, and hardware on acceptable terms. A record headline cannot answer that question.

Power Is the Project’s Hardest Constraint

The Ohio campus cannot reach its planned scale unless its energy system arrives alongside the servers.

SB Energy plans to develop 10 gigawatts of new generation for the campus. At least 9.2 gigawatts would come from natural gas facilities under the public framework.

The Department of Energy says Japanese funding would provide $33.3 billion for that gas generation. The power assets would connect with the local grid rather than operating as an isolated electrical island.

AEP Ohio is also planning new high-voltage transmission infrastructure. The utility says SB Energy has committed to cover $4.2 billion of those investments.

The proposal uses 765-kilovolt transmission lines, which move large amounts of electricity efficiently across long distances. AEP expects power to begin flowing to the site in 2029.

Those timelines show why a 10-gigawatt announcement does not equal immediate computing capacity. Generating plants, substations, transmission corridors, data halls, and chip deliveries must proceed in coordinated phases.

The Department of Energy’s project outline says the development could create 10,000 construction jobs and more than 2,000 permanent positions. Those are government projections, not independently verified employment totals.

Federal officials also say excess generation and transmission capacity would be available to other grid users. SB Energy has committed to paying for supporting infrastructure and accelerated site cleanup.

Those promises address a growing political concern. Communities increasingly question whether data centers will raise electricity bills or absorb grid capacity needed by existing residents and businesses.

The Ohio plan attempts to answer that criticism by pairing the new load with dedicated generation and privately funded transmission. Whether that protection works will depend on future regulatory approvals and operating arrangements.

Natural gas introduces another tradeoff. It can provide dispatchable power, meaning operators can increase output when the campus needs electricity.

Yet a 9.2-gigawatt gas fleet would also create long-term fuel dependence and emissions. The final environmental impact will depend on plant efficiency, utilization, methane leakage, and future generation changes.

The location presents additional scrutiny because the campus occupies land undergoing federal environmental cleanup. The Department of Energy continues monitoring the former uranium-enrichment complex.

Federal ownership can simplify land coordination, but it does not eliminate environmental responsibilities. Construction phases still need assessments, permits, remediation work, and safe integration with the existing site.

National forecasts make the power issue broader than Ohio. The Department of Energy’s resource hub cites estimates that data centers could consume 11.8 percent of United States electricity by 2030.

That estimate sits within a range, and actual demand will depend on deployment speed and hardware efficiency. Nevertheless, it illustrates the pressure facing utilities across major data center markets.

OpenAI’s proposed tenancy would transform the Portsmouth project from speculative capacity into a concentrated industrial load. It would also make energy delivery part of the company’s product roadmap.

A delayed transmission line could postpone access to computing systems even if the data halls were complete. A delayed turbine could have the same effect.

The hardest part of scaling AI is therefore becoming physical coordination. Software releases can move quickly, but power plants and transmission networks follow multi-year schedules.

The $500 Billion Estimate Needs a Stress Test

The project’s headline value depends on full deployment, current input prices, and demand that remains strong throughout a long construction cycle.

The estimate reportedly includes chips, labor, power, and other materials needed for the entire campus. It should not be confused with Stargate’s separate $500 billion national commitment.

The two figures resemble each other, but they describe different concepts. Stargate set an investment ambition across United States infrastructure, while the Ohio estimate concerns one possible campus buildout.

That distinction is easily lost when Google News displays several stories with similar numbers. It is essential for evaluating what OpenAI and its partners have actually committed.

The first uncertainty is utilization. A campus creates economic value only when customers use its computing capacity consistently enough to cover equipment, energy, maintenance, and financing costs.

OpenAI has strong demand for ChatGPT and its developer services. Public interest alone, however, does not disclose the margins generated by workloads running on increasingly expensive infrastructure.

The second uncertainty is model efficiency. Software improvements can reduce the computing required for a given task, while more capable models can create entirely new demand.

Both effects can occur together. Lower costs encourage more usage, but they can also alter which chips and cluster designs remain economically attractive.

The third uncertainty concerns hardware cycles. AI accelerators improve rapidly, while data center buildings, power plants, and transmission assets remain in service much longer.

OpenAI and SB Energy must therefore design infrastructure that can accept future systems. Otherwise, equipment selected during early planning could constrain later phases.

The fourth uncertainty is financing. A Nvidia guarantee could improve access to capital, but the final exposure, conditions, and duration remain unknown.

Guarantees also do not remove operating risk. They redistribute losses if the tenant cannot meet its obligations or the project cannot refinance on expected terms.

The fifth uncertainty is execution. The Ohio campus combines federal land reuse, environmental cleanup, gas generation, high-voltage transmission, data center construction, and advanced computing equipment.

Each component has a different regulator, supplier base, and schedule. A delay in one part can prevent the others from producing revenue.

AEP’s target of beginning power delivery in 2029 provides a useful checkpoint. It suggests that meaningful grid-connected capacity remains several years away, even if early site work advances sooner.

The government’s public descriptions also contain assertions that require future proof. Officials say dedicated generation and developer-funded transmission will protect consumer electricity costs.

That outcome will depend on rate treatment, construction overruns, fuel prices, and actual use of shared grid assets. It cannot be confirmed at groundbreaking.

The employment estimates require similar caution. Large projects create substantial temporary construction work, but permanent staffing at highly automated data centers can be much smaller.

The project could still deliver major regional benefits through tax revenue, supplier activity, cleanup funding, and transmission investment. Those effects should be measured rather than assumed.

OpenAI’s exact commitment remains the largest verification gap. A signed lease, disclosed initial capacity, and firm equipment orders would make the proposal more concrete.

Until then, the cautious conclusion is straightforward. The Portsmouth campus is an active infrastructure development, while OpenAI’s reported deal remains a potentially transformative tenancy.

Three Signals Will Show Whether the Campus Can Scale

A signed OpenAI lease, funded power milestones, and an initial operating phase will matter more than another headline estimate.

The first signal is a definitive tenancy agreement. OpenAI, SB Energy, or Nvidia would need to disclose the initial capacity, lease duration, guarantee structure, and deployment schedule.

Those details would clarify whether OpenAI is reserving the full campus or beginning with a smaller phase. They would also show how financial risk is divided.

A confirmed Nvidia guarantee would strengthen the case that the project can attract financing. Narrower support, conditional backing, or no final agreement would weaken expectations for rapid expansion.

The second signal is progress on generation and transmission. AEP’s planned 2029 power date creates a measurable infrastructure milestone.

Regulatory approvals, turbine contracts, transmission routing, and construction starts will indicate whether that schedule remains credible. Delays would directly limit when servers can begin operating.

The transmission plan also promises that SB Energy will pay for the new infrastructure. Regulatory filings should reveal how those costs are assigned and protected from overruns.

Evidence that Ohio customers remain insulated would support the project’s political case. Disputes over rates or grid access would weaken it.

The third signal is the first operating phase. Investors and customers need evidence of delivered computing capacity, not only permitted land and announced power.

An initial phase would reveal which Nvidia systems OpenAI deploys, how much power becomes available, and whether construction costs match early assumptions.

It would also show whether OpenAI uses the campus for model training, inference, enterprise services, or a mixture. That workload mix affects utilization and revenue potential.

Success would not require the entire 10 gigawatts to appear at once. A phased campus can establish its economics before later construction begins.

Failure to activate an initial phase would make the $500 billion estimate less meaningful. It would suggest that financing and infrastructure announcements are advancing faster than deployable demand.

For developers and enterprise buyers, the outcome affects more than OpenAI’s balance sheet. Additional capacity can influence API availability, service reliability, model release timing, and usage limits.

For utilities and communities, Portsmouth will test whether dedicated generation can protect existing customers while serving an unprecedented new load. That claim deserves transparent measurement.

For Nvidia, the campus could show whether supplier-backed financing can accelerate AI infrastructure without creating excessive customer concentration.

The Ohio project is therefore not merely a larger Stargate site. It is a test of whether the AI industry can coordinate demand, credit, electricity, and hardware at one location.

Watch for contracts before capacity claims, energized infrastructure before job projections, and operating servers before accepting the full valuation. Those signals will determine whether the campus becomes productive infrastructure or an expensive option on future AI demand.