OpenAI signs 20-year Ohio data center lease with Nvidia backing up to $105 billion
OpenAI has signed a 20-year lease with SoftBank subsidiary SB Energy for the "PORTS-Pike" data center campus in Ohio, giving it around 8 gigawatts of IT capacity; the Wall Street Journal puts the project's gross total at 10 gigawatts once cooling and other infrastructure are counted. The site sits partly on a former US Department of Energy uranium enrichment facility and draws power from a 9.2-gigawatt gas plant that the US government owns and Japan is financing under a trade agreement.
Nvidia is backing the deal with up to $105 billion and becomes the exclusive chip supplier for the first construction phase, which covers 4.25 gigawatts of IT capacity. Per the WSJ, Nvidia is not guaranteeing OpenAI's rent payments; it is backing the residual value of the finished data centers. If OpenAI walks away, SB Energy must first find a replacement tenant and then try to sell the facilities, and only then does Nvidia cover the shortfall, capped at $105 billion. Nvidia is separately investing $1.5 billion in SB Energy. OpenAI says it only pays for finished capacity, and the first 800 megawatts are due online in 2028.
Nvidia CEO Jensen Huang announced the deal using a new acronym, "LPS": land, power and shell. According to Huang, these basics, not chips or networking gear, are now the real bottleneck in the AI buildout, because AI labs like OpenAI are growing faster than their balance sheets can support long-term infrastructure contracts. Huang expects about 1.5 million GPUs and $150 billion to $200 billion in revenue per system generation, and puts OpenAI's total Nvidia compute commitments across all sites at roughly 12 gigawatts through 2030. If Nvidia exercises its option on the remaining 3.75 gigawatts at the Ohio site, the package grows to about 16 gigawatts worth roughly $600 billion. The WSJ had previously reported that Nvidia originally wanted to backstop the entire project with around $250 billion, but scaled the guarantee back after a 5 percent stock drop and investor pressure; the final $105 billion cap is below the $120 billion figure that had circulated most recently, and covers only the asset value rather than the full lease.
A separate WSJ analysis cited in the same report found that nine tech companies, including Alphabet, Meta, Microsoft and Nvidia, together hold around $3 trillion in mostly AI-related obligations that do not appear on their balance sheets, because leases are recorded only once payments begin and purchase commitments only once goods are delivered. Leases that have not yet started add up to $1.2 trillion, four times the figure a year earlier. Alphabet's purchase commitments alone jumped from $332 billion to $811 billion within three months. The contracts are described as nearly impossible to cancel, and both Alphabet and Amazon have recently reported negative free cash flow. Morgan Stanley analysts warn that investors can barely gauge these companies' actual debt levels anymore.
Key facts
- OpenAI signed a 20-year lease with SoftBank's SB Energy for an 8-gigawatt Ohio data center campus (10 gigawatts gross, per the WSJ), with the first 800 megawatts due online in 2028.
- Nvidia is backing the project's residual value for up to $105 billion, capped after scaling back from an original plan of around $250 billion, and becomes exclusive chip supplier for the first 4.25-gigawatt construction phase.
- Nvidia CEO Jensen Huang says land, power and shell ("LPS") have replaced chips as the AI buildout's real bottleneck, and estimates OpenAI's total Nvidia compute commitments at about 12 gigawatts through 2030.
- A WSJ analysis finds nine major tech companies hold about $3 trillion in mostly AI-related obligations off their balance sheets, including $1.2 trillion in not-yet-started leases, four times the prior year's level.
- Alphabet's purchase commitments jumped from $332 billion to $811 billion in three months, and Morgan Stanley analysts warn investors can barely gauge these companies' real debt levels.
Why it matters
This is the largest data center project announced to date, and it shows how AI labs are now financing infrastructure they cannot fund on their own balance sheets: Nvidia is using its cash and chip supply position to backstop the residual value of facilities built for its biggest customer. Huang's framing, that land, power and building shells rather than chips are now the binding constraint, marks a shift in what the industry treats as scarce.
Who it affects
OpenAI gains guaranteed compute capacity without carrying the full financial risk of the build. SB Energy, as SoftBank's subsidiary, is the lessor and must find a replacement tenant or buyer before Nvidia's guarantee pays out if OpenAI exits. Nvidia becomes the exclusive chip supplier for the first phase and takes a $1.5 billion stake in SB Energy. Investors in Alphabet, Meta, Microsoft, Nvidia and other large tech companies are affected indirectly, since the WSJ analysis says the industry's off-balance-sheet AI commitments now total around $3 trillion.
How to use it
There is no product or price here for end users; the deal concerns data center capacity, not a released service. For anyone tracking AI infrastructure economics, the reported terms are the useful reference points: an 8-gigawatt lease over 20 years, a $105 billion residual-value cap covering 4.25 gigawatts of the build, and a possible expansion to 16 gigawatts worth roughly $600 billion if Nvidia exercises its option on the remaining Ohio capacity.
How solid is it
The core deal terms are reported directly and specifically (lease length, capacity figures, the $105 billion cap, the phased structure). The broader $3 trillion off-balance-sheet figure and related numbers (the $1.2 trillion in unstarted leases, Alphabet's jump from $332 billion to $811 billion in purchase commitments) come from a separate Wall Street Journal analysis cited in the piece, not from OpenAI, SB Energy or Nvidia themselves. The claim that Nvidia originally wanted to backstop the project with about $250 billion before scaling back is also attributed to prior WSJ reporting rather than confirmed by the companies.
Risks and caveats
The article itself flags the concern: leases and purchase commitments of this scale are structured so that they largely stay off balance sheets, and Morgan Stanley analysts say investors can barely gauge these companies' real debt levels as a result. The contracts are described as nearly impossible to cancel, while Alphabet and Amazon have recently reported negative free cash flow. Nvidia's own guarantee is not a payment backstop for OpenAI's rent; it only covers the gap in asset value if SB Energy cannot find a replacement tenant or buyer, and only up to $105 billion.