OpenAI secures giant Ohio data center backed by Nvidia
OpenAI has leased an eight-gigawatt Ohio data center backed by a $105 billion Nvidia guarantee, highlighting the extreme financial maneuvers required to secure future AI infrastructure.
OpenAI has committed to a 20-year lease at the PORTS-Pike campus in Ohio, a facility managed by SoftBank subsidiary SB Energy. Situated on a former U.S. Department of Energy uranium enrichment site, the project will supply OpenAI with 8 gigawatts of IT capacity, which expands to a gross total of 10 gigawatts when accounting for cooling and infrastructure. Electricity will be sourced from a 9.2-gigawatt gas plant owned by the American government and funded by Japan. The initial 800 megawatts of capacity are scheduled to go online in 2028.
Nvidia is backstopping the venture by guaranteeing up to $105 billion of the residual asset value during the first phase of construction, which covers 4.25 gigawatts of capacity. Under the agreement, Nvidia secures exclusive chip-supply rights for this first phase and will invest $1.5 billion in SB Energy. Nvidia CEO Jensen Huang noted that physical assets like land, power, and building shells have overtaken chips as the primary bottleneck in AI scaling. Huang projects that OpenAI's compute demands through 2030 will require 12 gigawatts of Nvidia hardware, potentially rising to 16 gigawatts worth $600 billion if Nvidia exercises its option on the remaining 3.75 gigawatts in Ohio.
This massive deal highlights a broader industry shift toward off-balance-sheet financing. According to financial analyses, nine major tech corporations, including Alphabet, Meta, Microsoft, and Nvidia, hold roughly $3 trillion in unrecorded AI obligations. Unstarted leases have quadrupled over the past year to $1.2 trillion, while Alphabet's purchase commitments skyrocketed from $332 billion to $811 billion in a single quarter. Analysts warn these hidden liabilities obscure true corporate debt, especially as Alphabet and Amazon recently experienced negative free cash flow.
For AI practitioners, this unprecedented infrastructure buildout ensures the long-term availability of compute for next-generation models. However, it also shows that AI development is consolidating around entities capable of securing scarce physical resources. As physical infrastructure becomes the ultimate gatekeeper, smaller developers may find themselves increasingly dependent on the cloud platforms of these heavily leveraged giants.
This is our own summary of reporting by The Decoder



