Nvidia’s Reported $250 Billion OpenAI Backstop: Inside the Ohio AI Data Center Financing Plan

Nvidia is reportedly negotiating an extraordinary financial guarantee of roughly $250 billion that could help OpenAI lease and finance a proposed 10-gigawatt artificial-intelligence data center campus in southern Ohio. The reported arrangement is not a completed loan, a signed guarantee, or a direct $250 billion cash payment. It is a potential credit backstop designed to make lenders more comfortable funding the construction and long-term lease obligations of a project being developed by SB Energy, a SoftBank Group company.

The distinction matters. A guarantee can create enormous contingent exposure without requiring the guarantor to transfer the full amount at closing. Nvidia would generally be called upon only if the relevant borrower, tenant, or project vehicle failed to meet covered obligations, subject to whatever limits, collateral, performance conditions, and loss-sharing provisions appear in the final contracts. Yet the scale under discussion is so large that investors treated the report as more than an ordinary commercial partnership. Nvidia shares fell about 5% on July 27, 2026, after The Wall Street Journal reported the talks.

The proposal also reportedly sits beside separate discussions under which Nvidia could help finance as much as $350 billion of chips for the site. Those two figures should not be added together and described as a $600 billion check from Nvidia. They refer to different potential arrangements: one involving lease and project financing, and another involving the acquisition or financing of computing hardware. Neither has been finalized, and the eventual commitments could be smaller, phased over many years, shared with other parties, or abandoned altogether.

Even with those qualifications, the talks expose a defining tension in the AI investment cycle. Nvidia has become extraordinarily profitable by selling the processors, networking equipment, systems, and software used to build AI infrastructure. OpenAI is among the most important sources of demand for that infrastructure, directly and through cloud providers. Helping OpenAI secure more computing capacity could preserve years of demand for Nvidia products. It could also transfer part of the customer’s financing risk back to the supplier, creating the appearance of a circular system in which capital supplied or guaranteed by Nvidia helps fund purchases that ultimately support Nvidia’s own revenue.

Last updated: July 28, 2026, 10:00 UTC. The reported negotiations were still ongoing at the research cutoff. Nvidia, OpenAI, and the U.S. Commerce Department had not publicly confirmed a final agreement.

Key Takeaways

• Main development: Nvidia is reportedly discussing a roughly $250 billion guarantee connected to OpenAI’s proposed lease of a 10-gigawatt data center campus being developed by SB Energy in southern Ohio.
• Not a direct payment: The reported backstop would support lease and project debt. It would not necessarily require Nvidia to fund $250 billion upfront, and it would become an actual loss only if covered obligations defaulted and contractual protections proved insufficient.
• Separate chip discussions: Nvidia is also reportedly considering ways to finance up to $350 billion of chips for the campus. That hardware financing is distinct from the proposed lease and debt guarantee.
• Project scale: The campus could cost more than $500 billion including computing hardware and would be built on federal and private land around the Department of Energy’s Portsmouth Site in Pike County, Ohio.
• Energy plan: The federal government says SB Energy plans 10 gigawatts of new generation, including at least 9.2 gigawatts of natural-gas generation, alongside major transmission construction.
• Market response: Nvidia shares fell about 5% on July 27 as investors reconsidered how much credit, customer, and project risk the chipmaker may be willing to assume.
• Status: The negotiations are advanced according to the original reporting, but no signed guarantee, final tenant agreement, complete lender package, or definitive hardware-financing contract had been announced by the research cutoff.

What Nvidia and OpenAI Are Reportedly Negotiating

The reported transaction begins with a practical problem: a project requiring hundreds of billions of dollars cannot be financed on the strength of an ambitious technology roadmap alone. Lenders need contractual revenue, credible tenants, collateral, completion protections, and counterparties capable of absorbing losses if construction is delayed or demand falls short. OpenAI is one of the world’s most valuable private companies, but valuation is not the same as credit quality. It does not publish the kind of audited public financial statements available for listed companies, it remains unprofitable, and it does not possess the long public-debt history or investment-grade rating that would ordinarily support a lease of this magnitude.

SB Energy, the proposed developer, can assemble land, power, construction, and operating arrangements. OpenAI can supply the demand for computing capacity. Nvidia can supply much of the hardware. The financing challenge is to persuade banks and capital-market investors that the lease payments and project debt will remain dependable across a period long enough to recover construction costs.

A Nvidia guarantee could change that risk calculation. By promising to cover specified obligations if OpenAI or a related project entity cannot perform, Nvidia could lend its own financial strength to the financing structure. The guarantee might reduce borrowing costs, increase the amount lenders are willing to provide, extend maturities, or allow the project to proceed with less equity capital. In project-finance language, the Nvidia support would function as a form of credit enhancement. In market commentary, such arrangements are sometimes described as a credit wrapper because the stronger company’s credit sits around the weaker or unrated exposure.

The exact legal structure is unknown. It could involve guarantees of rent, debt service, construction completion, residual equipment value, or a combination of obligations. It could be capped by phase, decline as debt amortizes, include collateral posted by OpenAI or SoftBank, and require other guarantors to share losses. It could also contain performance milestones that prevent later phases from being financed until earlier capacity is operating. Without the contracts, it is impossible to know whether the headline amount represents maximum gross exposure, expected exposure, a series of separate commitments, or a theoretical ceiling across the full life of the campus.

That uncertainty should restrain both alarm and enthusiasm. Describing the proposal as Nvidia “spending $250 billion” overstates the immediate cash requirement. Describing it as costless because it is “only a guarantee” understates the risk. A guarantee is valuable precisely because the guarantor is expected to pay if specified failures occur. The central analytical question is not whether Nvidia transfers $250 billion on day one. It is how much risk Nvidia would accept, how likely that risk is to crystallize, how long it would remain outstanding, and what compensation or strategic benefit Nvidia would receive in return.

The reported $350 billion chip financing is a separate issue

The Wall Street Journal and Reuters’ account of the report said the proposed $250 billion guarantee would cover the data center lease and debt financing rather than the Nvidia processors installed inside the facility. Separate discussions could involve financing up to $350 billion of Nvidia chips for OpenAI.

Hardware financing could take several forms. Nvidia might provide extended payment terms, arrange loans through financial partners, support a special-purpose vehicle that owns the equipment, guarantee a portion of equipment debt, lease systems rather than sell them outright, or invest in entities that purchase the chips. Each structure has different consequences for cash flow, revenue recognition, balance-sheet exposure, and loss risk.

The distinction is important because a long-dated chip-financing arrangement may weaken the normal separation between a sale and the financing of that sale. If Nvidia recognizes revenue before collecting all cash, investors must assess the quality of receivables and the strength of the customer. If the equipment is leased, revenue may be recognized over time rather than at delivery. If Nvidia guarantees third-party financing, the company may retain contingent exposure even after another lender supplies the cash. If the chips serve as collateral, their future resale value will depend on technology cycles that can make expensive accelerators obsolete faster than conventional real estate or power assets.

No public document available at the cutoff establishes which approach is being negotiated. The prudent interpretation is therefore narrow: substantial chip financing is reportedly under discussion, but its final amount, structure, timing, and accounting treatment remain unknown.

What Is Confirmed, What Is Reported, and What Remains Uncertain

The project combines public announcements, company disclosures, and anonymously sourced reporting. Treating all three as equally certain would produce a misleading article.

Confirmed through public sources

• The Department of Energy has announced a partnership involving SB Energy, SoftBank Group, and AEP Ohio for a 10-gigawatt data center and power complex at the Portsmouth Site in Pike County.
• The federal plan includes 10 gigawatts of new generation, including at least 9.2 gigawatts of natural-gas generation, and $4.2 billion of new transmission infrastructure with AEP Ohio.
• DOE is leasing federal land at the former Portsmouth Gaseous Diffusion Plant to an SB Energy-affiliated entity, and the initial development area includes 189 acres.
• Nvidia and OpenAI have an extensive strategic relationship. In September 2025, they announced a letter of intent covering at least 10 gigawatts of Nvidia systems and a potential investment by Nvidia of up to $100 billion as capacity was deployed.
• Nvidia has already entered smaller facility-lease guarantee arrangements. Its April 2026 quarterly filing disclosed maximum gross exposure of $3.5 billion across partner lease guarantees, partly mitigated by $712 million held in escrow.
• Nvidia’s fiscal first-quarter 2027 results showed $81.6 billion of quarterly revenue, including $75.2 billion from Data Center, and $50.3 billion in cash, cash equivalents, and marketable debt securities at April 26, 2026.

Reported but not independently confirmed by the companies

• Nvidia is negotiating a roughly $250 billion guarantee tied to OpenAI’s lease and project financing for the Ohio campus.
• Nvidia is separately discussing financing up to $350 billion of chips for OpenAI at the site.
• OpenAI is among the parties showing the strongest interest in becoming the tenant, while Anthropic, Microsoft, and Google have also expressed interest.
• The complete project could cost more than $500 billion when computing hardware is included.
• The first phase could provide approximately 800 megawatts in 2028.

Still unknown

• Whether Nvidia and OpenAI will sign any guarantee.
• The identity of the legal borrower, tenant, and guaranteed parties.
• Whether the guarantee would be unconditional or limited by performance tests, collateral, phases, maturities, or loss-sharing arrangements.
• Whether the reported $250 billion is a maximum across all phases or a practical expected exposure.
• How much equity SoftBank, SB Energy, OpenAI, Nvidia, government-linked entities, infrastructure funds, and other investors would contribute.
• Which lenders would provide debt and what credit rating, interest rate, covenants, and maturity the financing would receive.
• Whether the chip financing would be a loan, lease, guarantee, deferred-payment plan, equity-backed vehicle, or combination.
• How Nvidia would account for the commitments and what disclosures regulators and auditors would require.
• Whether OpenAI would occupy all 10 gigawatts, share the campus, or commit only to the initial phase.
• How the final project would address construction, fuel, pipeline, water, emissions, permitting, and transmission risks.

The most consequential fact is therefore the status: this is an exceptionally large negotiation, not a completed financing.

Why Nvidia Would Consider Taking This Risk

Chief Executive Jensen Huang has positioned Nvidia as an infrastructure-platform company rather than a stand-alone chip vendor. The simplest explanation for considering the guarantee is demand protection. Nvidia’s success depends on customers continuing to build systems that use its accelerators, networking products, software, and rack-scale designs. The Ohio campus would not merely represent a large sale. At full scale, it could create a multi-year stream of hardware deployments, upgrades, replacement cycles, service relationships, and software usage. A guarantee that helps unlock the project could therefore secure revenue that might otherwise be delayed, reduced, or diverted to competing processors.

That incentive is unusually powerful because the AI hardware market is moving from a period of scarcity into a period in which financing, power, and customer economics may become as important as chip availability. During the first phase of the generative-AI boom, cloud companies and model developers competed to obtain every advanced accelerator they could. Nvidia’s constraint was primarily supply. As the buildout expands into projects measured in gigawatts, the constraint shifts. The industry must finance land, generation, transmission, substations, cooling, buildings, servers, networking, and long-term operating expenses before it can sell enough AI services to recover those costs.

Nvidia can respond in two ways. It can remain a conventional supplier and accept whatever level of demand customers can independently finance. Or it can use the wealth created by the chip boom to enlarge the market itself. Equity investments, lease guarantees, supplier credit, infrastructure funds, and strategic partnerships can help customers deploy more systems sooner. The reward is a larger installed base and greater dependence on Nvidia’s platform. The cost is that Nvidia begins to absorb risks previously held by customers, banks, landlords, and infrastructure investors.

A supplier can strengthen its ecosystem without giving money away

A guarantee can be economically rational when the supplier receives enough compensation and control. Nvidia could be paid a guarantee fee, receive warrants or equity, obtain long-term purchase commitments, secure priority rights over hardware deployment, require collateral, or gain influence over the technical design of the campus. Its existing facility-lease guarantees were provided in exchange for warrants, according to the company’s quarterly filing. That detail demonstrates that Nvidia does not necessarily view guarantees as charitable support. They can be strategic investments with potential upside.

The value of that upside depends on the contract. Warrants in a rapidly growing infrastructure company can offset expected credit losses. Minimum purchase obligations can protect Nvidia’s manufacturing plan. Restrictions on competing hardware can defend market share. Escrow, parent guarantees, and staged commitments can limit downside. A large gross figure may also overstate economic risk if exposure declines automatically as lease payments are made and if later phases require fresh approval.

There is nevertheless a difference between supporting a few billion dollars of partner leases and wrapping a project associated with a $250 billion headline. The larger the commitment, the more difficult it becomes to describe the activity as a peripheral extension of ordinary sales. It begins to resemble a capital-allocation strategy in its own right.

OpenAI is strategically important to Nvidia

OpenAI’s influence extends beyond its direct purchasing power. Its models shape expectations across the cloud industry, enterprise software, consumer applications, and developer tools. When OpenAI requires more compute, cloud providers build more capacity. When it deploys a new generation of models, competitors often respond with additional training and inference spending. Supporting OpenAI can therefore stimulate demand throughout the ecosystem.

The relationship has also become financially intertwined. Nvidia participated in OpenAI’s 2026 fundraising, and the companies had previously announced a plan under which Nvidia intended to invest progressively as OpenAI deployed at least 10 gigawatts of Nvidia systems. The final form of that earlier plan changed as the parties negotiated a broader funding round, illustrating that headline commitments in this market can be revised before money is deployed.

The Ohio discussions should be viewed in that context. They are not an isolated favor offered to an unfamiliar customer. They are a possible next stage in a strategic partnership linking model development, chip demand, infrastructure deployment, and capital formation.

Can Nvidia Financially Support a Guarantee This Large?

Nvidia is one of the few companies with the profitability, market capitalization, and access to capital needed even to contemplate an arrangement of this size. That does not mean the exposure would be immaterial.

For the fiscal year ended January 25, 2026, Nvidia reported $215.9 billion of revenue and $102.7 billion of operating cash flow. At year-end it held $62.6 billion in cash, cash equivalents, and marketable securities. In the quarter ended April 26, 2026, revenue reached $81.6 billion, operating cash flow was $50.3 billion, and the company held $50.3 billion in cash, cash equivalents, and marketable debt securities plus $30.2 billion of marketable equity securities.

Those figures show immense financial capacity, but they also illustrate why the form of the guarantee matters. A $250 billion maximum exposure would exceed Nvidia’s cash and marketable debt securities by several times. Nvidia could not treat the arrangement as though it were a fully cash-collateralized obligation without radically changing its balance sheet. The practical structure would almost certainly depend on long maturities, phased construction, declining exposure, collateral, co-guarantors, and the low expected probability that every covered obligation fails simultaneously.

Sources: Nvidia’s fiscal 2026 Form 10-K, Nvidia’s fiscal first-quarter 2027 Form 10-Q, and Nvidia’s first-quarter earnings release. Amounts are reported under U.S. GAAP except where the company separately identifies non-GAAP measures.

Nvidia has already warned investors about this category of risk

Nvidia’s annual report is unusually relevant because it anticipated the issue before the Ohio report. The company said it had been asked to offer financing arrangements supporting customers’ and partners’ data center buildouts. It warned that commercial arrangements can expose Nvidia to counterparties that fail to meet commitments, secure financing, complete infrastructure, or avoid financial distress. It also said financing arrangements could reduce upfront cash flow and increase credit risk.

By April, the theoretical risk had begun to become a reported balance-sheet item. Nvidia disclosed agreements guaranteeing partners’ facility lease obligations in the event of default, with $3.5 billion of maximum gross exposure. The partners had placed $712 million in escrow, and the exposure was scheduled to decline as lease payments were made over five to seven years.

That existing structure provides a useful template for thinking about Ohio. It shows how Nvidia can use guarantees to accelerate the ecosystem while receiving warrants and requiring protection. It also shows why investors will demand precise disclosure. A guarantee’s headline amount is only the starting point. Escrow, collateral, payment schedules, counterparty quality, probability of default, and recovery rights determine its economic value.

The Ohio proposal would be more than seventy times the disclosed maximum gross exposure of Nvidia’s existing facility guarantees. Even if the final structure is heavily phased, the comparison explains the market’s reaction. Investors were not learning that Nvidia had invented a completely new tool. They were learning that the company might deploy a familiar tool at a radically different scale.

The Circular Financing Concern

“Circular financing” is not a formal accounting diagnosis. It is a description of an economic pattern. The concern arises when a supplier invests in, lends to, guarantees, or otherwise supports a customer that uses the capital to buy the supplier’s products. The transaction can be commercially legitimate, but it complicates the interpretation of demand. Investors must ask how much purchasing would have occurred without the supplier’s financial support.

In the reported Ohio structure, Nvidia could help OpenAI obtain a lease, OpenAI could use the campus to deploy large quantities of Nvidia systems, and Nvidia could earn revenue from those systems. If Nvidia also finances the chips, the loop becomes tighter: Nvidia helps provide or guarantee the capital used to acquire Nvidia products.

The phrase becomes especially sensitive when several companies participate in overlapping deals. OpenAI receives equity from chipmakers and cloud providers, signs long-term compute contracts with those same firms, and uses the resulting capacity to sell AI services. Nvidia invests in model developers and infrastructure providers that buy Nvidia hardware. Cloud companies invest in AI laboratories that commit to their clouds. Chipmakers offer equity, warrants, or financing to encourage customers to adopt their architectures. Each transaction may have strategic logic, but the aggregate network can make it difficult to distinguish independent end demand from demand supported by the ecosystem’s own capital.

Why circularity does not automatically mean the revenue is fake

Vendor financing is common in capital-intensive industries. Aircraft manufacturers support airline purchases. Industrial-equipment companies finance machinery. Telecommunications suppliers have offered credit to network operators. Automakers maintain finance divisions. A supplier may understand the asset better than a general-purpose lender and may rationally accept credit exposure to win a long-term customer.

The existence of financing does not mean the equipment lacks value or that the customer lacks revenue. OpenAI has a large user base, meaningful subscription income, enterprise demand, and access to exceptional amounts of equity capital. Nvidia’s products are scarce, productive assets used across many workloads. The Ohio project also includes power and transmission infrastructure that could retain value even if one tenant’s strategy changes.

A guarantee can therefore solve a genuine market failure. Lenders may be reluctant to underwrite a private AI company’s long-dated lease because its business model is new and its financial disclosures are limited. Nvidia may possess superior information about hardware demand, model scaling, customer usage, and system economics. If that information justifies confidence, Nvidia can profit by accepting risk that outside lenders price too conservatively.

Why the concern remains legitimate

The risk is not that every supported sale is fictitious. The risk is that financing can pull future demand into the present, conceal weak customer economics, or encourage infrastructure that would not clear an independent return threshold. A supplier whose revenue depends on continued expansion may be more optimistic than a lender whose return depends on being repaid.

Several warning signs deserve attention:

• Customer concentration: Nvidia already reports that a limited number of direct and indirect customers account for a meaningful share of revenue. Supporting a major customer financially can increase that concentration even when the legal counterparties differ.
• Long asset lives versus short technology cycles: Data center buildings and power plants may operate for decades, while AI chips can be surpassed within a few years. Financing must remain viable even as hardware generations change.
• Revenue quality: Sales funded by extended terms or guaranteed debt may convert to cash more slowly and expose Nvidia to losses after revenue is recorded.
• Correlated downside: If AI demand disappoints, OpenAI’s ability to pay could weaken at the same time that Nvidia’s chip revenue, equity investments, and collateral values decline.
• Capital-allocation drift: Nvidia could move from a high-margin, relatively asset-light supplier model toward a structure requiring more guarantees, investments, and infrastructure exposure.
• Reduced price signals: When suppliers and strategic investors absorb financing risk, the apparent demand for capacity may not reflect the price an independent customer would pay without support.

The best test is transparency. Investors need the amount and duration of exposure, the identity and credit quality of counterparties, expected fees, collateral, loss-sharing, termination rights, and the relationship between supported transactions and Nvidia’s recognized revenue. Without that information, the market will apply its own discount.

The comparison with the dot-com and telecom eras

Critics have compared AI vendor financing with telecommunications-equipment deals around the turn of the century, when suppliers extended credit to network operators that then purchased the suppliers’ equipment. Some operators failed, leaving vendors with bad debts, returned equipment, and revenue that had looked stronger than the underlying economics.

The analogy is useful but incomplete. Today’s largest AI participants include highly profitable cloud companies, and Nvidia has far stronger cash generation than many telecom-equipment vendors had. AI systems already produce substantial commercial revenue, and computing capacity can be redirected among workloads more easily than some specialized telecom networks could be repurposed. OpenAI has also raised unprecedented amounts of equity capital.

The differences do not eliminate the core lesson. Financing cannot create durable end demand by itself. If customers do not generate enough cash from AI services, the ecosystem eventually has to reduce spending, restructure obligations, raise more equity, or transfer losses to guarantors and lenders. The amount of capital involved makes that lesson more important, not less.

OpenAI’s Compute Ambition and the Financing Gap

OpenAI Chief Executive Sam Altman has repeatedly made access to far more computing capacity central to the company’s strategy. OpenAI’s problem is not a lack of strategic ambition. It is the mismatch between the speed at which the company wants to deploy computing capacity and the pace at which an unprofitable private enterprise can generate cash internally.

OpenAI has raised capital on a scale rarely seen in private markets. Reuters reported that a 2026 funding round included commitments from Amazon, Nvidia, and SoftBank and was followed by a post-money valuation of approximately $852 billion. That valuation reflects expectations about OpenAI’s future earnings, market position, and potential public listing. It does not place $852 billion of cash on the company’s balance sheet, guarantee lenders repayment, or prove that every proposed data center will earn an adequate return.

The company’s infrastructure commitments have expanded across multiple partners. OpenAI has announced Stargate projects with Oracle and SoftBank, a large Nvidia systems partnership, use of Amazon’s Trainium capacity, and a collaboration with Broadcom on custom accelerators. This diversification reduces dependence on any one cloud or chip supplier, but it also creates a complicated web of minimum commitments, technical roadmaps, capital needs, and execution dependencies.

The Ohio campus represents a further step: moving from purchasing cloud services toward greater control over the physical infrastructure. Control can improve scheduling, hardware optimization, capacity certainty, and bargaining power. It can also transfer construction, financing, utilization, and technology-refresh risk closer to OpenAI.

Why controlling infrastructure can be strategically valuable

Cloud rental offers flexibility. A customer can buy capacity as needed, avoid owning rapidly depreciating equipment, and rely on the provider to operate buildings, power systems, networks, and cooling. That model helped OpenAI scale quickly through Microsoft Azure and later through additional providers.

At very large scale, rental has limitations. The cloud provider earns a margin, controls deployment schedules, allocates scarce accelerators among customers, and may design infrastructure around a broader customer base rather than one model developer’s exact requirements. A company training and serving frontier models may believe that custom rack layouts, networking, memory systems, power distribution, and software can materially improve cost per unit of useful computation.

Direct control also creates an option on future demand. If OpenAI believes model capability, enterprise adoption, AI agents, coding systems, video generation, and scientific applications will grow rapidly, securing a decade of power and land may appear cheaper than repeatedly bidding for short-term cloud capacity.

The downside is utilization. A cloud provider can spread capacity across thousands of customers. A primarily single-tenant campus depends more heavily on the tenant’s workload. If model efficiency improves faster than demand, if competitors win users, if custom chips reduce Nvidia requirements, or if regulators constrain deployment, the project could carry more capacity than OpenAI needs. The lease remains payable even when servers are underused.

Valuation is not a substitute for cash flow

Private-market valuations often create confusion in infrastructure reporting. When investors value OpenAI at hundreds of billions of dollars, they are purchasing equity based on expected future outcomes. Lenders financing a data center focus on a different question: what contractual cash flow will pay interest and principal on time?

An equity investor can tolerate years of losses if the company’s eventual value rises. A project lender normally expects scheduled payment regardless of whether the tenant’s next model succeeds. That is why a company can be extremely valuable and still need a credit wrapper. Nvidia’s guarantee would not necessarily imply that OpenAI is close to failure. It would reflect the fact that a private, loss-making company is being asked to support obligations comparable to the largest infrastructure financings in history.

OpenAI’s ability to raise equity is a real protection. So are its subscription revenue, enterprise contracts, strategic partners, and potential access to public markets. But those protections need to be compared with the volume and duration of its compute commitments. A company can raise record capital and still overcommit if infrastructure spending grows faster than revenue and new financing.

OpenAI’s diversification weakens and strengthens Nvidia at the same time

OpenAI’s relationships with AMD, Broadcom, Amazon, Microsoft, Oracle, and other suppliers reduce Nvidia’s exclusivity. Custom accelerators and competing GPUs can improve OpenAI’s negotiating position and reduce the risk of dependence on one vendor. For Nvidia, that makes a large Ohio partnership strategically attractive: financing support may help anchor future deployments on Nvidia’s platform.

Yet the same diversification creates risk for Nvidia. A guarantee could remain outstanding while OpenAI shifts an increasing share of workloads to other processors. If the contract does not secure sufficient Nvidia demand, the company could assume credit exposure without receiving the expected hardware revenue. Any final agreement will therefore be judged partly on the commercial commitments attached to the guarantee.

SoftBank and SB Energy’s Role in the Ohio Project

The Ohio campus is not simply an OpenAI–Nvidia transaction. SB Energy is the developer assembling the physical project, and SoftBank’s strategic relationship with OpenAI is central to the proposed structure.

SB Energy began as a renewable-energy developer and has expanded into powered data center sites. Its public materials list the PORTS Technology Campus as a 10-gigawatt site in early construction. The company’s role includes coordinating land, generation, transmission, construction, and the commercial arrangements needed to make the campus usable by a technology tenant.

SoftBank, led by Masayoshi Son, has made OpenAI a core part of its AI strategy. It has invested in OpenAI, participated in Stargate, and positioned itself as a capital provider and infrastructure partner. That creates aligned incentives. SoftBank benefits if OpenAI grows and consumes more compute. OpenAI benefits from SoftBank’s willingness to organize projects too large for conventional venture funding. SB Energy benefits from securing a tenant with enormous demand.

The alignment is commercially useful, but it also adds another layer of circularity. A major OpenAI investor is linked to the developer seeking an OpenAI lease. Nvidia, another OpenAI investor and supplier, may guarantee the lease and finance the chips. The project’s economics therefore depend on related strategic interests as well as independent lender appetite.

What the developer must deliver

At 10 gigawatts, the developer’s challenge extends far beyond constructing server halls. It must coordinate:

• Federal and private land rights;
• environmental review and remediation at a former uranium-enrichment site;
• natural-gas generation and fuel supply;
• pipeline capacity;
• high-voltage transmission and substations;
• water and cooling systems;
• roads, fiber, security, and workforce access;
• procurement of transformers, turbines, switchgear, and other long-lead equipment;
• construction financing before lease revenue begins;
• tenant specifications that may change as chip designs evolve;
• operating systems capable of maintaining extremely high reliability.

Every component has its own schedule. A building can be ready before transmission. Turbines can arrive before a pipeline. Servers can be delivered before cooling is commissioned. A delayed permit can postpone revenue while interest continues to accrue. This is why project lenders seek guarantees and why guarantees can become expensive even when the final asset is valuable.

SoftBank’s negotiating leverage

SoftBank can offer more than money. Through SB Energy, it can present OpenAI with a large, coordinated site rather than requiring the model developer to assemble dozens of smaller contracts. Through its equity relationship, it can accept a longer time horizon than a conventional developer. Through its government relationships in Japan and the United States, it can participate in a project tied to trade and industrial policy.

That combination gives SoftBank leverage over tenants and suppliers. Nvidia wants the chips selected. OpenAI wants the capacity. Governments want investment, jobs, and AI infrastructure. Lenders want strong counterparties. SB Energy sits at the point where those interests meet.

The developer’s leverage is not absolute. A site without a creditworthy tenant cannot support its full financing. OpenAI has alternative cloud and data center relationships. Nvidia can support other model developers. Government backing does not eliminate construction economics. The final contract must therefore allocate risk among parties that need one another but do not share identical interests.

The Portsmouth Site: From Uranium Enrichment to AI Infrastructure

The planned campus is associated with the Department of Energy’s Portsmouth Site near Piketon in Pike County, southern Ohio. The location gives the project a powerful industrial-policy narrative: federal land once used for uranium enrichment would be redeveloped for artificial-intelligence infrastructure.

The Portsmouth Gaseous Diffusion Plant operated as a uranium-enrichment facility from 1954 until 2001. The broader DOE site covers more than 3,700 acres and remains subject to an extensive environmental cleanup program that began in 1989. Deactivation, demolition, waste management, and remediation continue under federal and state oversight.

DOE has said the first data center phase will use 189 acres of leased federal land. SB Energy has committed to support accelerated cleanup and remediation, while the government presents redevelopment as a way to turn a Cold War industrial property into a new technology campus.

That history creates both opportunity and obligation. The site offers large tracts of controlled land, existing industrial infrastructure, workforce experience, and federal involvement. It also carries a legacy of contamination and complex regulatory responsibilities. Construction schedules must remain compatible with cleanup activities, worker safety, waste management, and land-release requirements.

Descriptions of the project should not imply that the entire former plant has already been cleaned and transferred for unrestricted use. DOE is making selected land available while the larger remediation mission continues. The interaction between cleanup and development is part of the project, not a completed precondition.

Why southern Ohio is attractive

Data center development has concentrated in places with available land, power, fiber, supportive governments, and access to skilled labor. Southern Ohio offers proximity to major Midwestern and eastern markets without the same land costs and congestion found in Northern Virginia or other mature hubs. The Portsmouth site also allows public authorities to coordinate land and power planning on a scale difficult to achieve in fragmented private markets.

The region has a manufacturing and energy workforce, but a project of this size would still compete for electricians, pipefitters, welders, equipment operators, engineers, and technicians. DOE has projected 10,000 construction jobs and more than 2,000 permanent positions. Those are official project estimates rather than guaranteed employment outcomes. Actual job creation will depend on the final scale, construction sequence, automation, local hiring, and the number of permanent operating functions located on site.

The economic impact could extend beyond direct employment. Transmission, generation, pipelines, suppliers, housing, logistics, and local services may attract investment. The risks include pressure on roads, housing, public services, water systems, and local labor markets. A $40 million community benefits agreement announced by DOE is intended to support surrounding communities, but the project’s full local costs and benefits cannot be known before construction and operating plans are final.

A 10-Gigawatt Data Center Is Really an Energy Project

The phrase “10-gigawatt data center” can sound like a measurement of computing alone. In practical terms, it describes an energy system with a data center attached.

Ten gigawatts is 10,000 megawatts of power demand at full load. If a completed campus drew that amount continuously for a full year, it would consume approximately 87.6 terawatt-hours of electricity. That calculated figure is about 61% of Ohio’s 2024 in-state net electricity generation of 142.7 terawatt-hours. The comparison is illustrative rather than a forecast: the campus would be built in phases, may not operate at maximum load every hour, and is paired with plans for new generation.

The scale explains why ordinary grid connection is insufficient. DOE says SB Energy plans 10 gigawatts of new power generation, including at least 9.2 gigawatts of natural-gas generation. The project also involves $4.2 billion of new transmission infrastructure with AEP Ohio, including high-voltage lines and substations. Japanese funding of $33.3 billion for the gas-generation component is linked to a broader U.S.–Japan trade and investment agreement.

Why natural gas dominates the initial plan

AI data centers require continuous power, rapid construction, and the ability to respond to changes in load. Natural-gas plants can provide dispatchable electricity and can often be developed faster than new large nuclear facilities. Existing gas supply networks and established turbine technology make the fuel attractive for a project targeting an initial operating phase in 2028.

The tradeoff is exposure to fuel prices, pipeline constraints, carbon emissions, local air permitting, and long-lived infrastructure that may conflict with corporate climate targets. Even efficient combined-cycle generation emits carbon dioxide. A 9.2-gigawatt fleet would be a major addition to regional fossil generation, making emissions policy and fuel availability material project risks.

Renewables, storage, nuclear power, and grid purchases could supplement the system over time. The International Energy Agency expects renewables and natural gas to lead near-term growth in power supply for data centers, with nuclear and other technologies contributing more later. The Ohio plan’s announced first phase, however, is built around a very large gas-generation commitment.

The project’s scale must be understood in phases

A 10-gigawatt announcement does not mean 10 gigawatts of servers will switch on in 2028. The reported first phase is approximately 800 megawatts. Later phases would require additional buildings, generation, transmission, equipment, financing, and tenant commitments. Each phase creates a decision point.

Phasing reduces risk because the parties can observe demand and performance before committing to the entire campus. It also introduces uncertainty because the headline project may never reach its theoretical maximum. Many infrastructure announcements describe ultimate site capacity rather than contracted near-term load.

For investors, the initial 800 megawatts may be more important than the 10-gigawatt vision. The first phase can establish construction cost, operating reliability, OpenAI utilization, lender appetite, and the economics of Nvidia systems at scale. Successful execution could unlock later phases. Delays or disappointing returns could lead the parties to resize the project without formally abandoning the site.

Ohio’s Grid Safeguards Are Part of the Financing Story

A project of this size cannot be evaluated only through the balance sheets of Nvidia, OpenAI, and SoftBank. The local utility framework matters because the cost of serving a large data center can otherwise migrate to households and ordinary businesses. Ohio regulators and AEP Ohio have already tried to address that concern through a special tariff for very large data-center customers.

AEP Ohio said in February 2026 that it had 5,642 megawatts of large-load commitments under binding agreements and another 12,219 megawatts of pre-tariff commitments, for a combined 17,861 megawatts. That total exceeded the utility’s existing system peak demand, which it placed at roughly 8,000 to 10,500 megawatts. The figures illustrate why data-center requests cannot simply be added to the grid on the same terms as a warehouse or office building.

The tariff requires qualifying data centers to pay for at least 85% of their subscribed energy demand, even if they use less. It also requires evidence of financial viability, imposes exit charges, and generally commits the customer for 12 years, including a four-year ramp period. Those provisions are designed to prevent other ratepayers from being left with the cost of transmission, substations, and generation built for a customer that delays, downsizes, or disappears.

That structure intersects directly with the reported Nvidia guarantee. A utility or project lender may insist on durable contractual payments before financing infrastructure. OpenAI’s private-company credit profile may not be sufficient on its own for every counterparty. A guarantee from Nvidia could improve confidence that lease and capacity payments will continue even if OpenAI’s finances deteriorate.

The guarantee therefore would not exist in isolation. It would sit beside utility commitments, minimum-payment provisions, construction contracts, power-purchase arrangements, equipment-financing documents, and government land agreements. Each contract transfers a different risk. The central question is not simply whether Nvidia “backs” OpenAI, but which payment obligations Nvidia would assume, under what circumstances, and for how long.

Why a minimum-payment tariff changes tenant behavior

A take-or-pay style obligation discourages speculative capacity reservations. Once a customer commits to paying for most of its subscribed load, reserving an extra gigawatt becomes expensive. That should make contracted demand more credible, but it can also amplify losses if expected AI usage fails to materialize.

For OpenAI, minimum payments would turn future electricity capacity into a fixed or quasi-fixed cost. For lenders, that improves revenue visibility. For Nvidia, a guarantee could transform part of OpenAI’s long-term operating obligation into a contingent liability. For Ohio ratepayers, the tariff is intended to keep those private risks from becoming public utility costs.

The effectiveness of that protection depends on enforcement, security deposits, collateral, and the creditworthiness of the guarantors. A contract is valuable only if the obligated party can perform when stress arrives. Nvidia’s scale makes it a powerful guarantor, but a $250 billion ceiling would be far larger than the lease guarantees Nvidia had disclosed in its April 2026 quarterly filing.

The U.S. Government Is Not a Passive Landlord

The federal role goes beyond leasing unused acreage. DOE controls the Portsmouth site, is responsible for environmental cleanup, and is facilitating a project intended to combine industrial redevelopment with AI infrastructure. The Commerce Department has also been involved in negotiations, according to reporting on the proposed financing. The transcript supplied with the CNBC segment incorrectly identified former Commerce Secretary Wilbur Ross; the current official involved is Commerce Secretary Howard Lutnick.

DOE has framed the development as a way to accelerate cleanup, create jobs, strengthen domestic AI capacity, and protect local consumers from the cost of the project’s electricity needs. SB Energy is expected to finance accelerated remediation of portions of the site needed for development. That arrangement gives the government a potential source of private capital for work that otherwise could remain a long-term federal obligation.

The site also carries symbolic value. Portsmouth was built for uranium enrichment during the Cold War. Reusing part of it for AI infrastructure connects two eras of strategic industrial policy: nuclear-material production and advanced computing. Both involve large federal landholdings, national-security arguments, energy-intensive facilities, and public-private coordination.

Symbolism does not remove execution risk. Federal land transactions, environmental review, cleanup standards, water needs, air permits, transmission approvals, and local infrastructure all can affect the schedule. The history of large industrial projects suggests that political support at announcement is not the same as timely delivery.

The U.S.–Japan financing link

DOE says $33.3 billion of Japanese funding is associated with the planned natural-gas generation. The arrangement reflects a broader pattern in which trade policy increasingly includes direct investment commitments in strategic U.S. industries. Japan brings capital and industrial partners; the United States offers access to land, energy projects, and a rapidly growing AI market.

SoftBank sits naturally at that intersection. It is a Japanese technology investor, a major OpenAI backer, and the owner of the developer connected to the Ohio campus. Its involvement can align national policy with corporate strategy, but it also concentrates several roles within one economic network.

The same group of partners may supply equity, develop the land, arrange power, support the tenant, and benefit from the purchase of computing equipment. That alignment can speed decisions. It can also make arm’s-length pricing and risk allocation harder for outsiders to evaluate.

Industrial policy can lower risk without eliminating it

Government participation can improve access to land, permitting coordination, infrastructure, and capital. It may also signal that a project has strategic importance. None of those advantages guarantees commercial success.

Public support is most valuable when it solves a genuine coordination problem: utilities need credible load commitments before building; developers need power before signing tenants; tenants need facilities before ordering chips; chip suppliers need confidence in deployment before expanding production. A coordinated framework can break that deadlock.

The danger is that strategic urgency weakens commercial discipline. When every participant expects another participant or government partner to absorb downside, capital can be committed on assumptions that would not survive a conventional underwriting process. The Ohio project’s eventual contracts will determine whether public involvement reduces genuine bottlenecks or merely redistributes risk.

Why Nvidia Shares Fell After the Report

Nvidia shares declined roughly 5% on July 27, 2026, after the financing report became a focal point for investors. The move did not prove that the proposed guarantee is uneconomic, nor can one day’s trading be attributed to a single headline with certainty. It did show that the market treated the reported structure as a potential change in Nvidia’s risk profile.

Investors have rewarded Nvidia for selling scarce accelerators at high margins while customers fund their own data centers. A guarantee blurs that clean supplier relationship. Nvidia would still sell chips, but it might also help underwrite the customer’s ability to pay for the buildings and power required to use them.

That distinction matters for valuation. A conventional product sale generates revenue and cash, subject to collection risk. A guarantee can create a contingent obligation that may not require cash at inception but can become costly if the underlying customer defaults. It also ties the supplier more closely to the economics of the customer’s project.

The market reaction likely reflected several overlapping concerns:

• Scale: A reported ceiling near $250 billion is enormous even for Nvidia.
• Concentration: OpenAI is already a strategically important customer and partner.
• Circularity: Nvidia may support financing that ultimately drives purchases of Nvidia systems.
• Disclosure: Investors do not yet know the final terms, collateral, duration, or triggers.
• Capital allocation: A guarantee could compete with other uses of Nvidia’s balance sheet and risk capacity.
• AI-spending skepticism: Markets are increasingly asking whether vast infrastructure commitments will earn adequate returns.

The share-price response should also be kept in proportion. Nvidia’s equity value can move by hundreds of billions of dollars on changes in growth expectations, interest rates, export policy, or broad technology sentiment. A 5% decline is material, but it does not reveal a settled judgment about the deal.

Why the headline number can overstate immediate exposure

A guarantee ceiling is not the same as an upfront payment. If the project is financed in stages, Nvidia’s exposure could begin far below $250 billion and grow only when construction milestones, tenant commitments, or equipment deployments are met. The agreement could include collateral, reimbursement rights, covenants, reserve accounts, or shared guarantees.

Those details could materially reduce expected loss. A $250 billion maximum covering many years and several phases is different from an unconditional guarantee of $250 billion on day one. The probability of default and the amount recoverable after default matter as much as the headline limit.

Investors nevertheless have reason to focus on the maximum. Tail risk is often invisible until conditions deteriorate. A guarantee can look inexpensive during rapid growth because no cash leaves the guarantor. Its cost appears when the customer, project, or financing market is under stress, precisely when the guarantor’s own business may also be weakening.

Why the market may be demanding a higher quality of AI revenue

The first phase of the AI infrastructure boom was dominated by scarcity. Cloud providers and AI labs competed for accelerators, and Nvidia’s growth was constrained more by supply than by demand. As spending commitments rise into the hundreds of billions, the debate shifts from access to return on invested capital.

Investors now have to ask who ultimately pays for AI compute. Enterprise subscriptions, consumer plans, advertising, API usage, government contracts, and productivity gains all can support demand. Yet the cash flows generated by AI services remain small relative to the infrastructure commitments being discussed across the industry.

Financing can bridge that timing gap. It cannot permanently replace profitable demand. The Ohio project will create durable economic value only if customers pay enough for the compute produced there to cover energy, chips, networking, maintenance, financing, and software development while still providing an acceptable return to capital providers.

The Strongest Case for the Guarantee

The favorable interpretation begins with Nvidia’s strategic position. The company is not merely selling a component; it is building an accelerated-computing platform that depends on large, coordinated deployments. If a credit guarantee enables OpenAI to construct infrastructure that would otherwise be delayed, Nvidia can expand the market for its systems while helping a leading AI developer maintain momentum.

A phased guarantee could be rational if four conditions hold. First, OpenAI demand must remain strong. Second, the facilities and power assets must retain value even if OpenAI falters. Third, Nvidia must receive compensation commensurate with the risk. Fourth, the contracts must limit exposure to projects that are actually completed and used.

Under those conditions, the guarantee resembles strategic market development. Industrial companies have long used financing, leasing, and customer support to accelerate adoption of expensive equipment. Aircraft manufacturers, automakers, telecom vendors, and energy-equipment suppliers all have helped customers finance purchases. The practice is not inherently improper or uneconomic.

Nvidia also has information advantages. It can observe accelerator orders, deployment schedules, utilization, model development, and customer demand across the AI market. It may be better positioned than a conventional lender to evaluate whether a large compute project is technologically viable and commercially necessary.

The company’s financial capacity is another argument. Nvidia generated more than $100 billion of operating cash flow in fiscal 2026 and continued to produce extraordinary cash flow in the first quarter of fiscal 2027. If its earnings remain strong, it can support commitments that would overwhelm a smaller supplier.

A guarantee could protect a broader ecosystem

OpenAI is important to Nvidia beyond direct purchases. Its models stimulate demand for cloud capacity, enterprise AI tools, inference services, developer software, and adjacent infrastructure. A slowdown in OpenAI deployment could affect cloud partners, data-center developers, networking vendors, and the perception of AI demand.

Supporting the Ohio campus could therefore protect network effects. More compute can support more capable models and wider product adoption; more adoption can create demand for inference; inference can justify additional infrastructure. Nvidia benefits at several points in that loop.

A successful campus could also become a reference architecture for sovereign and industrial AI projects. The ability to coordinate chips, networking, power, cooling, financing, and software at unprecedented scale would strengthen Nvidia’s position against rival accelerators and custom silicon.

The asset may be more reusable than the tenant

A data center built for OpenAI is not necessarily worthless without OpenAI. Buildings, substations, transmission connections, cooling systems, fiber routes, and generation assets can serve another hyperscaler or AI lab, although conversion may require time and expense. The report that Anthropic and Microsoft had also shown interest in the site suggests that demand may not depend on one tenant alone.

Residual value is critical to the guarantee analysis. If lenders can re-lease completed capacity to a replacement tenant, losses after default may be much smaller than the guaranteed amount. If the campus is highly customized, remote, incomplete, or technologically obsolete, recovery could be far worse.

The favorable case therefore depends less on the publicity value of “10 gigawatts” than on mundane underwriting details: land rights, power contracts, equipment ownership, step-in rights, lease transferability, completion guarantees, and collateral priority.

The Strongest Skeptical Case

The skeptical interpretation is that Nvidia is moving from selling the picks and shovels of the AI boom to financing the miners. That can preserve demand in the short run while weakening the quality of revenue and concentrating risk in customers whose own economics remain unproven.

OpenAI has attracted extraordinary amounts of capital, but its infrastructure ambitions are larger still. A private company can raise equity at a high valuation and continue to operate at a substantial loss for years. That does not make every long-term lease obligation safe. Equity investors accept uncertain returns; lenders and guarantors need predictable payment capacity.

If Nvidia supports a project whose main economic purpose is to buy Nvidia equipment, the transaction can create circularity even without improper accounting. Nvidia’s chip sales may be real, and OpenAI’s lease payments may be contractually valid, yet the economic demand is partly enabled by Nvidia’s balance sheet. Investors then need to distinguish customer-funded revenue from supplier-supported revenue.

The risk grows when similar arrangements multiply. Nvidia’s quarterly filing already disclosed billions of dollars of guarantees associated with partner facilities, large equity investments, and substantial future investment commitments. A $250 billion backstop would represent a different order of magnitude. Even if the Ohio exposure is phased, it could establish a precedent for other strategic customers to request comparable support.

The guarantee may solve the wrong bottleneck

Capital is only one constraint on AI infrastructure. Power generation, turbines, transmission equipment, skilled labor, water, permits, transformers, networking components, and construction management can all delay deployment. A stronger credit wrapper cannot manufacture a gas turbine or shorten every interconnection study.

The more ambitious the campus, the more likely that several bottlenecks interact. A delay in power generation can leave completed data halls idle. A delay in chips can leave power assets underused. A model-architecture change can alter cooling or networking requirements. A financing structure that assumes synchronized delivery may prove fragile.

Demand could grow while returns disappoint

AI usage does not have to collapse for the project to underperform. Compute demand can rise rapidly while prices fall faster. Competition among model providers may transfer much of the economic benefit to customers. Open-source models, efficiency improvements, specialized chips, and better software could reduce the amount of premium Nvidia compute required for a given task.

OpenAI might fill the campus and still struggle to earn an adequate return if inference prices decline, customer acquisition remains expensive, or model development consumes the productivity gains generated by new hardware. In that scenario, the physical infrastructure is busy but the tenant’s cash flow remains weak.

That distinction is central. Utilization measures activity. Creditworthiness depends on revenue, margins, liquidity, and access to capital. A lender can be repaid by a low-margin business if cash flows are stable, but a rapidly expanding company with high utilization can still default if its funding model breaks.

Ten Risks That Matter More Than the Headline

1. OpenAI credit and refinancing risk

The most direct risk is that OpenAI cannot meet lease or project obligations. OpenAI has raised enormous sums, and major technology groups have strategic reasons to support it. Those facts improve access to capital but do not create an investment-grade public credit record.

Long-lived data-center obligations may outlast a fundraising cycle, product generation, or partnership structure. If equity markets become less receptive, OpenAI may need debt, strategic capital, asset sales, or revised contracts. A guarantee is valuable precisely because lenders recognize that possibility.

Refinancing risk matters even before default. Construction debt may need to be replaced with permanent financing after completion. If interest rates rise, credit spreads widen, or projected cash flows weaken, refinancing can become more expensive. Nvidia could be asked to extend support or face a claim under the original guarantee.

2. Construction and completion risk

Large data centers frequently encounter cost inflation, contractor shortages, design changes, and schedule slippage. The Ohio project adds power plants, transmission, environmental remediation, and federal land coordination to the conventional data-center construction challenge.

A guarantee covering lease obligations may begin only after completion, or it may support debt during construction. The distinction is crucial. Completion risk is often highest before an asset generates revenue, when partially built infrastructure has limited resale value.

Fixed-price contracts, performance bonds, contingency budgets, milestone-based funding, and experienced contractors can reduce this exposure. They cannot eliminate it at a project measured in gigawatts and tens of billions of dollars.

3. Power-price and fuel-supply risk

Natural gas provides dispatchable power, but it creates a continuing commodity exposure. A long-term fuel contract can stabilize prices while introducing counterparty and basis risk. Pipeline capacity may require expansion. Severe weather can disrupt supply or cause regional price spikes.

Electricity is not a minor operating expense at this scale. A one-cent-per-kilowatt-hour change applied to 87.6 terawatt-hours would equal approximately $876 million a year at a theoretical full 10-gigawatt continuous load. The full campus may never operate at that level, but the calculation shows why energy economics can overwhelm seemingly small pricing assumptions.

4. Technology-obsolescence risk

AI hardware improves quickly. A data center designed around one generation of accelerators may need electrical, cooling, networking, or rack-density changes for the next. Nvidia’s platform leadership reduces compatibility risk, yet it also raises the pace of replacement.

The useful economic life of servers may be much shorter than the life of buildings and power plants. Financing must match those different durations. A 20-year infrastructure asset supporting hardware replaced every few years creates residual-value and upgrade-funding questions.

The risk is not limited to a rival chip. Algorithmic efficiency can reduce compute requirements. Smaller specialized models can replace general-purpose models for many tasks. Inference may migrate to lower-cost hardware. Any of those changes could lower the value of capacity built to current assumptions.

5. Customer and ecosystem concentration

Nvidia already depends on a relatively small number of large direct customers for a substantial share of sales, even though those customers serve many end users. A guarantee tied to OpenAI adds another form of concentration: credit exposure to a customer whose demand also supports Nvidia’s revenue growth.

Concentration can be rational when the customer is strong and the relationship creates durable advantages. It becomes dangerous when supplier, investor, lender, and strategic-partner roles reinforce one another. A problem at OpenAI could then affect chip demand, investment values, guarantees, and broader AI sentiment simultaneously.

6. Accounting and disclosure risk

Investors will need enough disclosure to understand the arrangement. Relevant information includes the maximum exposure, current funded amount, duration, collateral, probability of loss, fees received, beneficiaries, and relationship between the guarantee and Nvidia product sales.

Accounting rules may require recognition of a guarantee liability at fair value when issued, followed by additional loss provisions if payment becomes probable. The exact treatment depends on contract terms. Even when no large expense is recognized initially, the economic exposure can be meaningful.

Disclosure also affects how investors judge revenue quality. If guaranteed financing enables related product purchases, Nvidia may need to explain whether the transactions are separate, whether collectability is probable, and whether any consideration paid to the customer affects reported revenue. There is no evidence in the available reporting that Nvidia plans improper recognition, but the structure will invite scrutiny.

7. Governance and conflict-of-interest risk

Nvidia, SoftBank, and OpenAI have overlapping investment and commercial relationships. Each has an incentive to support the others’ success. That alignment may enable ambitious projects, but it complicates negotiations over price and risk.

Independent boards, conflict committees, external advisers, and transparent terms can help. Outsiders still may struggle to determine whether a financing guarantee is the best use of Nvidia’s capital or a strategic subsidy intended to protect prior investments and future chip sales.

8. Regulatory and environmental risk

The project will require compliance across energy, environmental, utility, land-use, and potentially national-security regimes. Natural-gas generation brings air-emissions permitting and climate-policy exposure. The former enrichment site brings cleanup obligations and public concern about redevelopment.

A change in administration, court challenge, local opposition, or revised environmental standard could alter timing and cost. Federal backing can accelerate coordination, but it also makes the project more visible and politically sensitive.

9. Opportunity-cost risk

A guarantee can consume risk capacity even when it consumes little cash. Credit-rating agencies, banks, directors, and shareholders may treat contingent obligations as constraints on future borrowing, acquisitions, dividends, repurchases, or investments.

Nvidia could earn a fee, warrants, priority supply commitments, or other compensation. The opportunity cost should be compared with alternative uses of the same balance-sheet capacity. A guarantee is attractive only if its expected strategic return exceeds the expected loss, administrative burden, and foregone flexibility.

10. Macroeconomic and rate risk

The financing spans a period in which interest rates, inflation, construction costs, and capital-market appetite can change materially. Higher rates increase the cost of debt and lower the present value of distant cash flows. A recession can weaken enterprise AI spending even as long-term adoption continues.

AI infrastructure may be strategically important and still cyclical. Semiconductor and data-center markets have repeatedly experienced periods of overbuilding. The extraordinary scale of current commitments increases the consequence of even a modest forecasting error.

How a Credit Guarantee Could Appear in Nvidia’s Financial Statements

A guarantee is different from a loan. Nvidia may not transfer cash to OpenAI or the project company when the contract is signed. Instead, it promises a lender, lessor, or other beneficiary that specified payments will be made if the primary obligor fails.

At inception, the guarantor generally evaluates the fair value of that promise. A fee received, warrants, commercial rights, or other consideration may be part of the measurement. Depending on the contract and accounting rules, Nvidia may record a liability representing the stand-ready obligation even when management believes default is unlikely.

Later, if a loss becomes probable and estimable, the company may need a larger provision. Cash outflow occurs if the guarantee is called and Nvidia must pay. Nvidia might then gain a claim against OpenAI, collateral, project assets, or replacement-tenant revenue.

The balance-sheet presentation alone will not answer every investor question. A modest initial liability can coexist with a very large contractual maximum because fair value incorporates the probability and timing of loss. Footnote disclosure is therefore essential.

Four numbers investors would need

• Maximum contractual exposure: the largest amount Nvidia could be required to pay.
• Current exposure: the amount associated with completed or financed phases at the reporting date.
• Recognized liability: the accounting value carried on the balance sheet.
• Collateral and recovery rights: assets or claims that could reduce ultimate loss.

Those numbers should not be conflated. A $250 billion maximum may coexist with a much smaller current exposure. Conversely, a small reported liability does not mean the maximum is irrelevant.

Revenue recognition deserves separate attention

The chip purchases reportedly would be financed separately from the $250 billion lease and project guarantee. That separation reduces one obvious concern, but the arrangements remain economically connected. The Ohio campus exists to house computing equipment, and Nvidia benefits when OpenAI can deploy more of it.

Investors will want to know whether chip sales occur at market terms, whether payment is unconditional, and whether Nvidia provides any rebates, financing, guarantees, or investments linked to those purchases. The appropriate accounting depends on the complete contractual package, not the title of one agreement.

The safest analytical approach is to avoid assuming either that all revenue is circular or that none of it is. The degree of supplier support should be quantified as contracts are signed and disclosed.

The Competitive Context: Nvidia Is Defending More Than One Chip Sale

OpenAI’s infrastructure strategy is not exclusively tied to Nvidia. The company has announced plans involving custom accelerators developed with Broadcom, large cloud arrangements with Oracle, and partnerships that can provide access to different hardware platforms. Amazon, Google, Microsoft, and Meta also are investing in proprietary silicon to reduce costs and dependence on merchant accelerators.

That diversification gives OpenAI bargaining power. It also gives Nvidia a reason to support projects that lock in its platform at large scale. A campus designed around Nvidia systems can create long-term demand for accelerators, networking, software, and services. The strategic value may extend well beyond the first equipment order.

Nvidia’s advantage is not limited to processor performance. Its CUDA software ecosystem, networking products, systems engineering, and developer adoption reduce the friction of deploying large clusters. Customers evaluate time to train, reliability, utilization, and software compatibility, not merely the purchase price of a chip.

Custom silicon can still alter the economics. A hyperscaler or model developer willing to optimize software for its own accelerator may achieve lower costs for recurring workloads. Broadcom’s announced collaboration with OpenAI for 10 gigawatts of custom accelerators shows that OpenAI wants alternatives. Nvidia may view the Ohio guarantee partly as a way to preserve a large share of future deployment before those alternatives mature.

Why the campus could lock in an architecture

Data-center design decisions have lasting consequences. Power distribution, liquid cooling, rack density, network topology, storage, and software orchestration are optimized around expected equipment. A site can be upgraded, but switching architectures at scale is not free.

If Nvidia systems anchor the first 800-megawatt phase, later expansions may benefit from operational familiarity and shared infrastructure. Engineers, contractors, and software teams learn one platform. Spare parts, monitoring, and maintenance processes become standardized. That creates a practical form of lock-in even without an exclusive contract.

OpenAI will resist dependence that weakens its negotiating position. The most likely outcome is a heterogeneous infrastructure portfolio: Nvidia remains central, while custom and rival accelerators handle selected workloads. The Ohio project’s financing terms could influence how that mix develops.

AMD and other merchant competitors

Advanced Micro Devices is the most visible merchant alternative for large AI accelerators. Its products can pressure Nvidia on price and give customers a second source. Intel and specialized startups also pursue parts of the market, while cloud providers offer their own chips as services.

A financing advantage can become a competitive advantage. If Nvidia can use its balance sheet to make a project bankable, a technically credible rival without similar financial capacity may find it harder to win the same deployment. That could strengthen Nvidia’s market position while attracting antitrust and policy attention.

The competitive question is therefore broader than whether Nvidia’s next chip is faster. It is whether the company can package hardware, networking, software, supply commitments, investment capital, and credit support into an infrastructure solution that rivals cannot match.

A Timeline of the Ohio Project and the Reported Nvidia Talks

1. January 21, 2025: OpenAI, SoftBank, Oracle, and MGX announce the Stargate initiative, with an ambition to invest up to $500 billion in U.S. AI infrastructure over four years.
2. September 22, 2025: OpenAI and Nvidia announce a letter of intent covering at least 10 gigawatts of Nvidia systems. Nvidia says it intends to invest up to $100 billion as capacity is deployed.
3. September 2025: OpenAI says its announced Stargate sites, including Abilene and five additional locations, represent nearly 7 gigawatts of planned capacity and more than $400 billion of investment over three years.
4. Early 2026: SB Energy advances the PORTS Technology Campus at DOE’s Portsmouth Site in southern Ohio, with early construction activity and a long-term 10-gigawatt objective.
5. February 12, 2026: AEP Ohio reports 17,861 megawatts of binding and pre-tariff large-load commitments across its territory, highlighting the scale of regional data-center demand.
6. February 27, 2026: Reuters reports that OpenAI has arranged a $110 billion funding round involving Amazon, Nvidia, and SoftBank. Subsequent commitments bring the reported total to $122 billion at an $852 billion post-money valuation.
7. Spring 2026: Federal and corporate officials promote the Portsmouth redevelopment and associated power plan, including Japanese financing for natural-gas generation and new transmission investment.
8. May 20, 2026: Nvidia reports fiscal first-quarter revenue of $81.6 billion and data-center revenue of $75.2 billion, demonstrating the cash-generating base behind its strategic financing capacity.
9. July 26, 2026: The Wall Street Journal reports that Nvidia and OpenAI are in talks over a guarantee of approximately $250 billion for lease and project financing tied to the Ohio campus.
10. July 27, 2026: Reuters and CNBC provide additional coverage. Nvidia shares fall roughly 5% as investors weigh credit exposure and circular-financing concerns.
11. July 28, 2026: At this article’s research cutoff, the parties have not announced a final agreement or disclosed definitive terms.

The chronology shows that the guarantee discussion did not emerge from nowhere. It follows a series of increasingly large infrastructure, investment, and hardware commitments. The Ohio proposal is a potential extension of that strategy, not a standalone project.

Four Plausible Outcomes for the Negotiations

Scenario 1: A phased guarantee is signed

The most straightforward outcome is an agreement that covers only committed phases and expands as milestones are met. Nvidia could guarantee a portion of lease or project payments, receive fees or warrants, and secure preferred-supplier status or minimum purchase commitments.

This structure would make the $250 billion figure a long-term ceiling rather than immediate exposure. It would be easier for Nvidia to defend if each phase has independent economics, collateral, and replacement-tenant potential.

For markets, the key question would shift from whether the deal exists to how much current exposure Nvidia recognizes. Detailed disclosure could calm concerns if the initial amount is modest and protections are strong.

Scenario 2: The project is resized or shared

The parties could reduce Nvidia’s role by adding other guarantors, lenders, tenants, or cloud partners. OpenAI might occupy only part of the campus, while Microsoft, Anthropic, or another buyer takes capacity. A shared structure would reduce concentration and improve asset utilization.

Resizing would not necessarily signal failure. A 10-gigawatt site can remain strategically important even if its first several phases are smaller than originally envisioned. A diversified tenant base may make financing easier and the project more resilient.

Scenario 3: Nvidia supports chips but not the real estate

Nvidia could decline the broad lease guarantee while participating through equipment financing, investments, supply commitments, or partnerships. This would preserve sales opportunities while limiting exposure to construction and power assets.

That outcome would leave OpenAI and SoftBank to find another creditworthy guarantor for the buildings and energy infrastructure. Banks, private-credit funds, insurers, sovereign investors, or government-supported entities could fill part of the gap, probably at a higher cost.

Scenario 4: The talks fail or the project is delayed

The parties may be unable to agree on compensation, collateral, control rights, or risk sharing. Power and construction constraints also could postpone the first phase. OpenAI could redirect workloads to other Stargate sites or cloud providers.

Failure would not end the broader AI buildout, but it would test whether headline infrastructure ambitions can proceed without extraordinary supplier support. It also could weaken confidence in the ultimate 10-gigawatt target.

What Would Make the Deal More Credible?

Because the negotiations are private, investors currently see a large headline without the terms required for underwriting. Several provisions would make a final arrangement easier to evaluate.

• Phase-by-phase exposure: Nvidia’s guarantee should grow only after clearly defined construction and occupancy milestones.
• Meaningful OpenAI equity at risk: The tenant and its investors should fund a substantial portion before the guarantor is exposed.
• Collateral and step-in rights: Nvidia or lenders should have enforceable claims on project assets and the ability to replace the tenant.
• Shared risk: SoftBank, OpenAI investors, banks, insurers, or other beneficiaries should absorb losses before or alongside Nvidia.
• Market-based compensation: Guarantee fees, warrants, supply rights, or other consideration should reflect the size and duration of the risk.
• Independent governance: Related-party and strategic conflicts should receive board-level review.
• Transparent disclosure: Nvidia should report maximum and current exposure, recognized liabilities, maturities, and connections to product sales.
• Replacement-tenant planning: The campus should be designed to serve other creditworthy users if OpenAI reduces demand.
• Power-cost protection: Long-term fuel, generation, and transmission arrangements should reduce volatility without shifting costs to Ohio consumers.
• Technology flexibility: Buildings and electrical systems should accommodate future hardware generations and more than one accelerator architecture.

No provision can make a project of this scale risk-free. Together, these protections could convert an open-ended corporate backstop into a measured infrastructure commitment.

What Investors Should Watch Next

The first item is a definitive announcement. A press release or filing should identify the guaranteed obligations, beneficiaries, term, maximum exposure, and consideration received. Until then, the $250 billion figure remains reported negotiation territory.

The second is Nvidia’s regulatory disclosure. A material guarantee may appear in a current report, quarterly filing, contractual-obligations note, guarantee footnote, or risk-factor update. Investors should compare the accounting liability with the maximum contractual exposure and funded project phases.

The third is project-level financing. The identities of lenders, debt maturities, interest rates, construction conditions, and collateral package will reveal how much risk private capital is willing to accept without Nvidia support.

The fourth is the Ohio construction schedule. The reported 800-megawatt first phase and 2028 target require progress on site preparation, generation, transmission, buildings, and equipment procurement. Delays in any one of those categories can affect the others.

The fifth is OpenAI’s operating trajectory. Revenue growth, cash consumption, customer retention, model economics, and future funding rounds will determine whether the guarantee remains a remote contingency or becomes a meaningful credit exposure.

The sixth is OpenAI’s hardware mix. Announcements involving Nvidia, Broadcom, cloud-provider chips, or other accelerators can change the expected equipment value of the campus. A 10-gigawatt site does not automatically translate into 10 gigawatts of Nvidia systems.

The seventh is Nvidia’s broader financing portfolio. Additional guarantees, equity investments, loans, or commitments to strategic customers could turn an exceptional transaction into a recurring business model. The aggregate exposure matters more than any single project.

Finally, investors should watch realized AI economics. The decisive evidence will be whether infrastructure produces durable, high-margin revenue for model providers and their customers. Capital commitments are inputs. Sustainable cash flow is the output that ultimately supports them.

Frequently Asked Questions

Is Nvidia giving OpenAI $250 billion in cash?

No. The reported transaction is a potential credit guarantee supporting lease and project financing for an Ohio data-center campus. A guarantee is a promise to cover specified obligations if the primary borrower or tenant fails. It is not the same as an immediate $250 billion cash payment.

Has Nvidia signed the guarantee?

Not as of the July 28, 2026 research cutoff. The Wall Street Journal reported advanced negotiations, and Reuters said it could not independently verify the report. Nvidia, OpenAI, and SoftBank had not announced final terms.

Does the $250 billion include Nvidia chips?

According to the reporting, no. The guarantee would support lease and project debt. Separate discussions could finance as much as $350 billion of Nvidia chips, potentially bringing total project costs above $500 billion when computing equipment is included.

Where would the data center be built?

The proposed campus is at the U.S. Department of Energy’s Portsmouth Site near Piketon in southern Ohio. The property previously hosted uranium-enrichment operations and remains subject to federal environmental cleanup.

How large is 10 gigawatts?

Ten gigawatts equals 10,000 megawatts. At continuous full load, it would use about 87.6 terawatt-hours of electricity in a year. The campus would be developed in phases, so that calculation illustrates ultimate scale rather than near-term consumption.

When could the first phase open?

The reported first phase is about 800 megawatts and is targeted for 2028. Construction, power generation, transmission, financing, permitting, and equipment delivery all can affect that schedule.

What is SoftBank’s role?

SoftBank is a major OpenAI investor, and its subsidiary SB Energy is the developer associated with the PORTS Technology Campus. SoftBank’s economic interests therefore include OpenAI’s success, project development, and the broader Stargate infrastructure program.

Why would Nvidia guarantee a customer’s lease?

The guarantee could lower financing costs, accelerate construction, and secure future demand for Nvidia systems. Nvidia may receive fees, warrants, commercial rights, or strategic benefits. The final economics cannot be assessed until terms are disclosed.

Why are investors worried about circular financing?

Nvidia could support financing that enables OpenAI to build facilities and buy computing equipment from Nvidia. The sales can still be genuine, but supplier support may make demand less independent and can expose Nvidia to losses if the customer or project underperforms.

Could Nvidia afford a $250 billion guarantee?

Nvidia has extraordinary cash generation, but affordability depends on timing, probability of loss, collateral, and aggregate commitments. A maximum guarantee is not an upfront payment, yet the reported ceiling is large relative to Nvidia’s cash holdings and existing disclosed guarantees.

Would Ohio consumers pay for the project’s electricity infrastructure?

AEP Ohio’s special data-center tariff is designed to protect other customers by requiring large users to pay for at least 85% of subscribed demand, demonstrate financial viability, accept exit fees, and commit for an extended period. Actual protection depends on contract enforcement and project creditworthiness.

What is the biggest unresolved question?

The central unknown is the allocation of downside risk. Without definitive contracts, investors cannot know how much Nvidia would guarantee at each phase, what collateral it would receive, how OpenAI and SoftBank share losses, or how the arrangement connects to chip purchases.

Historical Comparisons: Vendor Finance, Railroads, Telecom, and Cloud

Supplier-supported financing has a long history. The structure can expand markets when customers face high upfront costs, and it can conceal weak demand when credit substitutes for commercial viability. The difference becomes visible only over time.

Aircraft manufacturers have supported customer financing because airlines buy expensive assets with long useful lives and observable resale markets. Automakers operate captive finance businesses because credit is integral to vehicle demand. Industrial-equipment companies arrange leases and loans because customers often prefer to match payments with the cash flow generated by machinery.

Telecommunications offers a more cautionary comparison. During periods of rapid network expansion, equipment vendors extended credit to carriers that used the proceeds to buy their systems. Some networks were genuinely necessary; others were built ahead of demand. When customers failed, vendors suffered loan losses and saw equipment orders disappear at the same time.

The late-1990s fiber boom is often invoked in debates about AI infrastructure. The comparison is useful but incomplete. Fiber networks eventually became essential, and the internet produced enormous economic value. Many individual companies and investors still lost money because capacity was built too quickly, financed too aggressively, or owned by businesses with unsustainable capital structures.

Railroad expansion in the nineteenth century offers a still older lesson. Infrastructure can transform an economy while generating poor returns for particular owners. Strategic importance and investment profitability are separate questions.

What is different about the AI buildout

AI infrastructure has several advantages over speculative projects of the past. Large technology companies already generate substantial cash, enterprise demand is measurable, and accelerators can be deployed across many workloads. OpenAI has a large user base and significant revenue, even though its spending needs remain greater.

The assets also differ. A transmission line or gas plant can operate for decades. A data-center building can serve more than one generation of equipment. Accelerators depreciate faster, but their output can be sold globally through digital services.

The principal similarity is the financing loop. When suppliers, customers, investors, and developers fund one another, demand can appear stronger and less risky than it would under independent underwriting. The loop can be productive while growth is strong. It becomes dangerous when the participants rely on continued asset appreciation or fresh capital to meet existing obligations.

The cloud-computing analogy

The development of public cloud infrastructure is a more favorable precedent. Amazon, Microsoft, and Google spent heavily before every workload was visible. Their data centers created flexible capacity that customers could rent, and scale lowered unit costs. The result was a durable industry with recurring revenue.

The Ohio project could follow that pattern if its compute is fungible, efficiently utilized, and supported by diversified customers. It could resemble a single-purpose project instead if capacity is tied too closely to one tenant and one technology path.

The lesson is not that large infrastructure spending is inherently reckless. It is that flexibility, customer diversity, disciplined phasing, and durable cash flow determine whether an ambitious platform becomes a cloud-like utility or a stranded asset.

The Broader Economic Stakes for Ohio

DOE and project sponsors emphasize jobs, redevelopment, and community benefits. The official estimates of 10,000 construction jobs and more than 2,000 permanent positions would make the campus a major regional employer. Construction activity also could support contractors, lodging, transportation, and local services.

Permanent data-center employment is more complicated than the headline capital investment. Data centers are extremely capital intensive and generally employ fewer workers per dollar invested than factories with labor-intensive production. The associated power plants, maintenance operations, security, network services, and suppliers can broaden the employment effect, but local officials will need to distinguish temporary construction peaks from lasting payrolls.

The project could expand the tax base and reuse federal land that has carried environmental and economic costs for decades. Its community value will depend on tax agreements, infrastructure obligations, workforce development, water use, environmental safeguards, and whether local residents qualify for the jobs created.

Energy affordability is another distributional issue. A dedicated generation and tariff structure can protect consumers, but a project this large will influence regional planning. New pipelines, substations, transmission corridors, and generation facilities create benefits and burdens that extend beyond the campus boundary.

Southern Ohio’s experience also will shape national policy. If the project delivers power without raising household rates, accelerates cleanup, and creates durable economic activity, it can become a model for redeveloping federal industrial sites. If costs migrate to the public or promised phases do not materialize, it will strengthen opposition to similar projects elsewhere.

Final Assessment

The proposed Nvidia backstop is best understood as a test of how far the AI industry is willing to integrate technology sales, infrastructure development, strategic investment, and credit support. It is not simply a chip order, a real-estate lease, or a government redevelopment project. It combines all three.

The strongest argument for the arrangement is practical. OpenAI wants an extraordinary amount of compute. SoftBank has a site and development platform. The federal government has land and an industrial-policy objective. Nvidia has the technology, cash generation, and strategic incentive to make the project financeable. A phased guarantee with strong collateral and shared risk could unlock infrastructure that produces valuable services for years.

The strongest concern is equally practical. The customer requiring support is also expected to buy the supplier’s products. OpenAI’s valuation and fundraising capacity do not by themselves prove that future AI revenue can support hundreds of billions of dollars of leases, energy assets, and computing equipment. If Nvidia absorbs too much downside, it would convert a high-margin supplier relationship into a concentrated credit bet.

At the July 28, 2026 cutoff, the market has a reported ceiling but not a contract. The missing terms are more important than the headline: initial exposure, phase triggers, guarantee fees, collateral, loss-sharing, replacement-tenant rights, and the connection to chip financing. Those details will determine whether the transaction is disciplined strategic finance or an open-ended subsidy to preserve demand.

The Ohio campus itself should be judged by its first operating phase, not its ultimate 10-gigawatt ambition. An 800-megawatt deployment that opens on schedule, secures reliable power, attracts durable workloads, and produces cash flow would support expansion. Delays, cost overruns, weak economics, or repeated requests for supplier financing would point in the opposite direction.

Nvidia’s reported willingness to negotiate shows how the AI boom is changing the boundaries of the semiconductor business. The company may no longer be able to maximize long-term demand by selling chips alone. The question is whether financing the ecosystem strengthens Nvidia’s platform or makes its balance sheet responsible for risks that customers and lenders are unwilling to carry.

This article is provided for general informational purposes and does not constitute financial, investment, tax, or legal advice.

Sources

• The Wall Street Journal: Nvidia in Talks With OpenAI to Guarantee $250 Billion Financing for Data Center
• Reuters: Nvidia in talks with OpenAI to guarantee $250 billion financing for data center
• Reuters Breakingviews: Nvidia’s financing exposure and circular AI concerns
• The Wall Street Journal: Nvidia stock reaction on July 27, 2026
• U.S. Department of Energy: Partnership to power AI at the Portsmouth Site
• U.S. Department of Energy: Ohio energy and AI project fact sheet
• U.S. Department of Energy: Portsmouth AI data-center development report
• SB Energy: PORTS Technology Campus overview
• Nvidia fiscal 2026 Form 10-K
• Nvidia fiscal 2027 first-quarter Form 10-Q
• Nvidia fiscal 2027 first-quarter results
• OpenAI and Nvidia systems partnership announcement
• OpenAI: Five new Stargate sites and infrastructure commitments
• OpenAI and Broadcom strategic accelerator collaboration
• Reuters: OpenAI funding round involving Amazon, Nvidia, and SoftBank
• Reuters: OpenAI valuation and investor scrutiny
• AEP Ohio: Large-load commitments as of February 2026
• AEP Ohio: Data-center tariff protections
• U.S. Energy Information Administration: Ohio electricity profile
• U.S. Department of Energy: Electricity demand growth and data centers
• International Energy Agency: Energy and AI executive summary
• U.S. Department of Energy: Portsmouth Site history and cleanup

Affiliate disclosure: Businessfinance.news may earn compensation from qualifying actions completed through selected links on this website, at no additional cost to the reader. Affiliate relationships do not influence our editorial reporting, analysis, or conclusions.