Nvidia $105B OpenAI Ohio Data Center Financing

The financing arrangement that hovered over the market for three weeks as a report is now a filing. Nvidia has agreed to back as much as $105 billion in financing for a giant new OpenAI data-center campus in southern Ohio, according to a securities filing disclosed on Monday, August 17, 2026. The company separately agreed to invest $1.5 billion in SB Energy, the SoftBank-backed developer that will build, own, and operate the site.

The disclosure formalizes — and substantially resizes — a deal that The Wall Street Journal first described in late July as a guarantee of roughly $250 billion. The final structure is smaller and more specific, but the shape is the same: the world’s most valuable chipmaker is standing behind the borrowing of its largest customer, and investors spent the day arguing about what that means.

What the filing says

Under the terms disclosed, Nvidia will provide a financial guarantee supporting the construction and lease of the campus rather than writing a direct check for the full amount. The credit will support an initial 4.25 gigawatts of computing capacity, with an option for an additional 3.75 gigawatts — up to roughly 8 gigawatts in total if both tranches are built out. That is a step down from the 10-gigawatt figure attached to the earlier reporting, and it caps Nvidia’s headline exposure at about $105 billion, below the $120 billion ceiling the company had signaled after investors pushed back on the original number.

The mechanics matter. The guarantee is tied to a 20-year lease OpenAI is signing for the facility. Under that lease, Nvidia’s obligation is generally triggered only by a shortfall — the difference between a guaranteed minimum value of the lease and whatever is recovered through a replacement tenant or a sale of the asset. In other words, Nvidia is not funding the campus outright; it is promising to cover lenders and the developer if OpenAI walks away and the space cannot be re-let at comparable value. It is closer to a co-signer than a shareholder, the same credit-enhancement logic that made the July report so striking.

The separate $1.5 billion equity investment in SB Energy is the piece analysts fixated on. SB Energy, majority-owned by SoftBank Group, will build and operate the data center. Nvidia is therefore taking a stake in the developer of a facility that OpenAI will fill with Nvidia GPUs — a single deal in which Nvidia helps fund the landlord, guarantees the tenant’s lease, and books revenue on the chips that go inside.

The Ohio campus

The site is the PORTS-Pike Technology Campus in Pike County, near Piketon in southern Ohio — the same location flagged in the earlier reporting, built on the grounds of a former uranium-enrichment complex with existing power infrastructure and transmission access. Developable, power-connected land at gigawatt scale has become one of the scarcest inputs in the AI build-out, and the Pike County site’s legacy grid connections are a large part of its appeal.

The first 800 megawatts is expected to come online by 2028, with the remaining capacity phased in over the following years as power, transformers, and the physical build allow. OpenAI said the project will support roughly 35,000 construction jobs through 2032 and about 2,500 long-term positions once the campus is running.

For OpenAI, the campus is another slab of the compute base it is assembling across the country, alongside its self-built Project Camellia site in Georgia and the 10-gigawatt systems partnership with Nvidia unveiled last year. The company has mapped out compute spending that runs into the hundreds of billions of dollars, a scale no operating cash flow can cover and no unrated borrower can finance alone — which is precisely why the Nvidia guarantee exists.

The market’s verdict: “circular”

Analysts reached for the same word within hours of the filing: circular financing. The critique is now familiar. Nvidia’s customers need capital to buy Nvidia’s chips; Nvidia helps supply or guarantee that capital; the capital is spent on Nvidia’s chips, which shows up as Nvidia’s revenue, which underpins the valuation and cash position that let it extend the next guarantee. When the chip vendor becomes the backstop for its own demand, critics argue, reported growth can outrun genuine, independent end-demand — and risk that would normally sit with diversified lenders concentrates back onto the vendor.

Stacy Rasgon of Bernstein Research wrote in a note that “the action will clearly fuel ‘circular’ concerns.” The market agreed in real time. Nvidia shares fell about 5%, ceding the title of most valuable company back to Apple. The price of credit-default swaps on Nvidia bonds — effectively insurance against a Nvidia default — recorded their sharpest intraday jump since the contracts began actively trading late last year, according to ICE Data Services figures cited by Bloomberg. That is a notable move for a company still sitting on one of the strongest balance sheets in the world, and it signals that bondholders are beginning to price the contingent liabilities Nvidia is accumulating across its web of AI deals.

Defenders make the opposite case. Vendor financing is old and ordinary — aircraft manufacturers, telecom-equipment makers, and enterprise-hardware giants have all guaranteed customer purchases for decades — and standing behind debt for a customer whose product is in obvious demand is a rational way to accelerate a market. Both readings can be true at once: the underlying demand for compute can be real and the financing structure can amplify fragility if that demand ever softens.

A wider pattern

The Ohio guarantee does not stand alone. It sits within a lattice of AI-era financing that has grown as consequential as the models themselves — from Nvidia’s reported role in arranging hundreds of billions in Wall Street AI financing to the exotic debt structures rivals have adopted, including Amazon’s $25 billion AI bond sale and the growing pile of off-balance-sheet obligations hyperscalers are using to fund data centers without parking the debt on their own books.

What makes the current phase distinct is the concentration. A handful of balance sheets — Nvidia’s above all — increasingly underwrite the demand that shows up as their own revenue. That works smoothly as long as compute demand keeps climbing and the campuses fill. The open question is what the structure looks like in a downturn, when a guarantee stops being a line in a filing and becomes a bill.

What it means

The filing turns a rumor into a template. A concrete, disclosed structure — guarantee plus developer equity plus a 20-year lease — is now on the public record, and other AI labs and chip vendors will study it as a model for financing gigawatt-scale campuses.

Who wins. OpenAI wins access to cheaper, larger capital than its own credit could command, letting it lock up scarce power and land ahead of rivals. Nvidia wins a mechanism to convert balance-sheet strength into durable chip demand, pulling forward revenue that might otherwise wait on customers’ ability to borrow. SoftBank and SB Energy win a marquee anchor tenant and a well-capitalized partner for the build.

Who carries the risk. The risk migrates onto Nvidia. A guarantee is a contingent liability — cheap unless something goes wrong, expensive if AI demand disappoints or OpenAI’s economics falter and the company is left standing behind debt tied to a half-utilized campus. The 5% stock drop and the CDS spike are the market’s first attempt to put a number on that tail.

What to watch next. First, disclosure: how Nvidia accounts for the guarantee and the SB Energy stake in its next filing, and how much of its cumulative contingent exposure across all its AI deals it breaks out. Second, the ratings agencies and lenders — how they price debt backed by a chipmaker’s guarantee rather than a borrower’s own cash flows. Third, the phasing — whether the first 800 megawatts actually lands by 2028, given how often power and transformer timelines slip. And fourth, the read-through: if a vendor backstop plus developer equity becomes the standard way to fund AI campuses, the industry’s growth and its fragility will increasingly rest on the same few balance sheets. The deal is signed now, not merely reported — which makes those questions urgent rather than hypothetical.