OpenAI has secured a 20-year lease for an 8-gigawatt AI campus in Ohio, backed by a $105 billion residual value guarantee from Nvidia and exclusive chip supply rights.
OpenAI has signed a 20-year lease with SoftBank subsidiary SB Energy for the PORTS-Pike campus in Ohio, marking the largest data center project announced to date. The agreement secures approximately 8 gigawatts of IT capacity for OpenAI, while the Wall Street Journal estimates the site’s gross capacity at 10 gigawatts once cooling and supporting infrastructure are included. Located partly on a former U.S. Department of Energy uranium enrichment facility, the campus will draw power from a 9.2-gigawatt gas plant owned by the U.S. government and financed by Japan under a trade agreement. Nvidia is backing the venture with a guarantee of up to $105 billion.
Nvidia CEO Jensen Huang has identified “LPS” as the next strategic resource. The acronym stands for land, power, and shell, referring to the site, the electricity supply, and the building itself. According to Huang, these fundamentals have replaced chips and networking gear as the primary bottleneck in AI infrastructure expansion. Because AI labs such as OpenAI are scaling faster than their balance sheets can sustain long-term capital commitments, Nvidia is intervening with financial support.
Under the arrangement detailed by the Wall Street Journal, Nvidia is not guaranteeing OpenAI’s rental payments. Instead, it backs the residual value of the completed data centers in the first construction phase, which covers 4.25 gigawatts of IT capacity. Should OpenAI terminate the agreement, SB Energy must first secure a replacement tenant and attempt to sell the facilities before Nvidia covers any shortfall in asset value, with its liability capped at $105 billion. In exchange, Nvidia secures exclusive rights as the chip supplier for the initial half of the campus and commits $1.5 billion in investment to SB Energy.
Huang projects that each system generation will require roughly 1.5 million GPUs, generating between $150 billion and $200 billion in revenue. Across all planned locations, he estimates OpenAI’s compute commitments through 2030 will reach approximately 12 gigawatts of Nvidia hardware. If Nvidia exercises its option to develop the remaining 3.75 gigawatts in Ohio, the total portfolio would expand to roughly 16 gigawatts, valued at approximately $600 billion. OpenAI maintains that it only pays for operational capacity, with the first 800 megawatts scheduled to come online in 2028.
The scale of these off-balance-sheet commitments extends across the broader industry. A Wall Street Journal analysis reveals that nine major technology firms, including Alphabet, Meta, Microsoft, and Nvidia, collectively hold around $3 trillion in primarily AI-related obligations that do not appear on their financial statements. Accounting standards dictate that leases are only recorded once payments commence, and purchase commitments are recognized only upon delivery. Consequently, unactivated leases now total $1.2 trillion—four times the amount recorded a year ago. Within just three months, Alphabet’s purchase commitments surged from $332 billion to $811 billion.
These agreements are largely non-cancellable, and both Alphabet and Amazon have recently reported negative free cash flow. Morgan Stanley analysts caution that the sheer volume of deferred liabilities makes it increasingly difficult for investors to assess the true debt burden of these corporations. Originally, Nvidia had sought to backstop the entire project with a $250 billion guarantee but reduced the commitment following a 5 percent decline in its stock price and mounting investor pressure. The final $105 billion cap falls short of the most recent $120 billion estimate and applies strictly to asset value rather than the full lease obligation.