OpenAI just signed the largest data center lease ever announced — a 20-year deal with SoftBank’s SB Energy for an Ohio campus that will eventually deliver around 8 gigawatts of IT capacity. The site, called PORTS-Pike, sits partly on a former Department of Energy uranium enrichment facility and pulls power from a 9.2-gigawatt gas plant owned by the US government. The first 800 megawatts come online in 2028, and OpenAI says it only pays for finished capacity.
Nvidia’s role here is the story. The chipmaker is capping its exposure at $105 billion, down from an original $250 billion backstop that reportedly got scaled back after a 5 percent stock drop and investor pressure. The guarantee isn’t for OpenAI’s rent — it’s for the residual value of the finished buildings in the first 4.25-gigawatt phase. If OpenAI bails, SB Energy has to find a replacement tenant and try to sell the facilities before Nvidia covers any shortfall. In exchange, Nvidia becomes the exclusive chip supplier for the first half of the site and is putting $1.5 billion into SB Energy.
Jensen Huang has a new acronym for what’s actually constraining AI buildouts: LPS — land, power, and shell. Chips and networking gear aren’t the problem anymore. Huang told the audience that AI labs are growing faster than their balance sheets can support long-term infrastructure contracts, which is precisely why Nvidia is stepping in with financial muscle. He’s projecting OpenAI’s total commitments through 2030 at around 12 gigawatts of Nvidia compute, with the Ohio expansion potentially pushing that to 16 gigawatts worth roughly $600 billion if Nvidia exercises its option on the remaining 3.75 gigawatts.
This is the part where the accounting gets uncomfortable. A Wall Street Journal analysis found nine tech companies — Alphabet, Meta, Microsoft, Nvidia among them — hold roughly $3 trillion in mostly AI-related obligations that never appear on a balance sheet. Leases only get recorded once payments begin, and purchase commitments only when goods are delivered. Alphabet’s purchase commitments jumped from $332 billion to $811 billion in a single quarter. Both Alphabet and Amazon recently reported negative free cash flow. Morgan Stanley analysts are blunt: investors can barely gauge these companies’ actual debt levels anymore. When the biggest constraint on AI is physical infrastructure, the financial engineering around it becomes just as critical as the silicon.
