Nvidia is putting its balance sheet behind physical AI real estate. In a blog post from CEO Jensen Huang, the company said it will partner with SB Energy to secure land, power and shell — what Nvidia calls LPS — at the PORTS-Pike Technology Campus in Portsmouth, Ohio, with OpenAI signed on as the tenant. The initial deployment is expected to provide 4.25 gigawatts of AI factory capacity, and Nvidia has the option to extend across the site's remaining 3.75 gigawatts.
The economics Huang laid out are unusually explicit. Each generation of Nvidia systems installed at PORTS-Pike could represent roughly 1.5 million Nvidia GPUs and $150 billion to $200 billion in Nvidia revenue. Over a 20-year productive life, the site is designed to support multiple upgrade cycles — meaning the same footprint of land and power keeps generating fresh compute revenue as hardware turns over.
“AI factories are the defining infrastructure of the AI era — where compute transforms energy and data into intelligence that powers every business, industry and country.”— Jensen Huang, Nvidia CEO
Zoom out to the full OpenAI commitment and the number gets larger. OpenAI's existing and planned deployments now represent approximately 12 gigawatts of Nvidia compute through 2030, with headroom to expand to 16 gigawatts if the PORTS-Pike arrangement stretches to the full site. At that scale, Huang pegged the opportunity at roughly $600 billion of Nvidia compute.
Key facts
- 01Nvidia is partnering with SB Energy to secure 4.25 GW of AI factory capacity at PORTS-Pike in Portsmouth, Ohio, with OpenAI as the tenant.
- 02Each generation of Nvidia systems deployed at the site could represent 1.5 million GPUs and $150B–$200B in Nvidia revenue.
- 03OpenAI has committed to 12 GW of Nvidia compute through 2030, expandable to 16 GW, worth roughly $600 billion.
- 04Nvidia will support LPS infrastructure for about 4 GW over a 20-year term, with the guarantee phasing in between 2028 and 2030.
- 05The site has 3.75 GW of remaining capacity that Nvidia may extend the arrangement to cover.
The structure is a departure from how Nvidia has historically sold. Hyperscalers and investment-grade enterprises typically secure their own land, power and shell, sign long-term utility contracts, and then buy Nvidia systems to fill them. That model, Huang said, will continue to represent most of Nvidia's business. Frontier labs are the exception — companies with fast-growing revenue but balance sheets that can't yet underwrite decade-scale infrastructure contracts on their own.
So Nvidia is stepping in. The company is guaranteeing defined portions of lease and power payments, along with a residual-value commitment, for about 4 gigawatts over a 20-year term. The guarantee phases in as data centers come online between 2028 and 2030, and Nvidia's exposure declines as OpenAI makes lease payments and capacity is placed in service.
“Their growth is increasingly constrained not by algorithms or customer demand, but by the availability of compute.”— Jensen Huang, Nvidia CEO
Huang directly addressed the obvious question: is this circular financing, given that Nvidia is effectively backstopping a site that will then buy Nvidia chips? His answer was no. OpenAI pays the lease; Nvidia is using its scale and demand visibility to secure a scarce input, the same way it locks in advanced packaging or memory supply upstream. LPS, in Nvidia's telling, has simply become another critical resource that needs to be reserved with long lead times.
The fungibility argument matters here. If OpenAI ever walks away from the site, Nvidia's position is that CUDA and the breadth of the developer ecosystem make its installed compute versatile enough to be resold to another cloud, enterprise, or lab. That resale story is what turns the guarantee from a bet on one customer into a bet on the platform itself.
“This is the essential economic point: the LPS commitment secures a long-lived AI factory site, while the NVIDIA compute inside can be upgraded repeatedly.”— Jensen Huang, Nvidia CEO
PORTS-Pike will run Nvidia's full-stack DSX AI factory platform, spanning GPUs, CPUs, networking and infrastructure software. It is the flagship example of what Huang described as Nvidia's evolution from selling chips, to selling systems, to now helping assemble the foundations that host those systems. Nvidia framed the move as selective — most LPS hosting Nvidia compute will still be procured directly by customers.
The counterweight is concentration risk. OpenAI is a single tenant, on a 20-year clock, at a site whose value is tied to the assumption that each successive generation of Nvidia hardware will keep producing more intelligence and more revenue than the last. If deployment slips past the 2028-to-2030 window, if power interconnects lag, or if frontier-model economics compress faster than expected, Nvidia's residual-value exposure is the line item that absorbs the shock. Huang was careful to note the guarantee covers defined portions of lease and power — not the full cost of the site — but the direction of travel is clear: Nvidia is now a party to the underlying real estate deal, not just a supplier to it.
The broader signal is that compute capacity has moved from being a chip problem to being a siting problem. Advanced packaging, memory and networking are still constraints, but the choke point Huang is publicly flagging is land, power and shell — the parts of the stack that require utilities, permits, and 20-year commercial commitments rather than fabs. By securing PORTS-Pike, Nvidia is telling investors that the next binding constraint on AI-factory revenue is physical infrastructure, and that it intends to underwrite that constraint for the customers it depends on most. For OpenAI, it means the compute roadmap through 2030 now has a specific address in Ohio. For every other frontier lab, it sets the precedent that Nvidia is willing to co-sign the buildout when the demand is real enough.
Working on something we should cover, or seeing a story we missed? Send leads, documents, or feedback to hello@aichatdaily.com. For sensitive tips, see our secure tips page for Signal and PGP options.
Spotted an error? Email hello@aichatdaily.com with the URL and the issue, or read our full corrections policy.
