The chip maker that powered the AI boom is now underwriting its real estate. Nvidia on Aug. 17 announced a multi-year partnership with SB Energy to advance the PORTS Technology Campus in Ohio, a data center project where OpenAI will anchor as a tenant, using about 4.25 gigawatts of IT load capacity to run Nvidia’s full-stack DSX AI factory platform.
The financial engineering is as striking as the scale. Nvidia has provided residual value guarantees of up to $105 billion tied to the lease arrangements, and may at its own discretion extend credit support for an additional roughly 3.8 gigawatts of capacity, the company said. The guarantee structure protects lenders and lessors against declines in the value of the underlying assets, letting the project borrow against Nvidia’s balance sheet rather than its own.
The numbers are hard to grasp in isolation. Four-plus gigawatts of IT load is roughly the electrical output of several large power plants, and it would make the campus one of the largest AI computing sites announced to date. OpenAI’s role as anchor tenant means the facility’s economics rest on a single customer whose demand for compute has grown at a rate few could have predicted two years ago.
The arrangement marks a shift in how Nvidia does business. The company built its fortune selling graphics processors that have become the standard engine of AI training and inference, and its full-stack strategy pushed it into networking gear and software. Now it is stepping into capital markets as well, putting its own balance sheet behind the infrastructure its customers need. Analysts said the guarantee transforms Nvidia from supplier to underwriter, a role no chip maker has played at this scale before.
The structure follows a pattern spreading across the industry, in which data center developers and cloud providers finance facilities separately from their own balance sheets and anchor tenants lock in capacity years ahead. What is new is the size of Nvidia’s commitment and the fact that the guarantor is the component supplier rather than a bank or a real estate investor. Lenders get Nvidia’s creditworthiness behind the leases; Nvidia gets a captive outlet for its hardware, software, and networking stack.
The residual value guarantee deserves attention. Such guarantees cover the gap if leased equipment or facilities are worth less than expected at the end of the lease term, and in this case the exposure could reach $105 billion if the assets’ value collapses. Nvidia is effectively saying it believes the AI infrastructure it sells will hold its value, a statement backed by its own engineering projections rather than by third-party appraisals.
Power is the binding constraint the deal is designed to solve. Data center developers around the world are waiting years for grid connections, and the competition for electricity has become the central bottleneck of the AI buildout. A campus of this size will require substantial grid upgrades, and the Ohio site’s location and power access are part of why it was chosen. Whoever secures power first gets to train models first, and the guarantee is Nvidia’s way of making sure its platform is the one running on the power that gets secured.
The risk to Nvidia is concentration. The company’s fortunes already ride on a handful of customers building enormous facilities, and a guarantee of this size ties its balance sheet to the same customers’ ability to pay and to use the capacity they lease. If AI demand softens, or if a tenant walks away, the guarantee would turn into a real liability on Nvidia’s books, a position the company has never carried before.
For now, the market’s reaction was muted, a sign that investors are treating the guarantee as a formality rather than a risk. The test will come later, when the campus is built and the leases mature, and when the question of who pays if the AI boom stumbles finally has an answer written into a contract.
SB Energy’s role is worth understanding. The company, backed by SoftBank Group, has built a portfolio of utility-scale renewable projects and positioned itself as a developer that can bring land, power, and capital together for large industrial customers, and the Ohio campus fits that model. The partnership gives Nvidia a development partner with a track record of completing big projects, and gives SB Energy a tenant and a guarantor that make the project’s financing dramatically easier.
The DSX platform is central to the arrangement’s logic. Nvidia’s full-stack AI factory architecture bundles its processors, networking, and software into a complete data center design, and a campus built around it locks in years of Nvidia hardware purchases across every layer. The guarantee is, in effect, Nvidia underwriting its own demand forecast: it is confident enough in the platform’s value to back the leases that finance it.
Analysts said the deal also reflects how the AI supply chain has reorganized. Chip makers once sold components and walked away; now they co-invest, guarantee leases, and help customers secure power, because compute demand has outrun every forecast and the bottleneck has moved downstream from silicon to sockets. Nvidia’s willingness to put $105 billion behind that view tells investors how seriously the company takes it.
