Nvidia is now underwriting the buildings, not just selling the chips
Nvidia has secured land, power and shell capacity in Pike County, Ohio to host compute for OpenAI, and invested $1.5 billion in SB Energy. The structure says more than the size.
On 17 August, Nvidia announced it had secured land, power and shell capacity at SB Energy's PORTS-Pike Technology Campus in Pike County, Ohio, to exclusively host Nvidia AI compute. OpenAI is the customer. SB Energy will build, own and operate the data centre under a 20-year lease.
Strip away the corporate structure and something unusual is going on. A chip company is providing credit support for real estate and electrical infrastructure so its largest customer can afford to deploy its chips.
The structure
- Nvidia provides credit support on land, power and shell buildout to secure an initial 4.25 IT-GW, with an option on the remaining 3.75 IT-GW.
- OpenAI is the customer for the full 8 IT-GW.
- Nvidia invests $1.5 billion directly into SB Energy, joining existing investors SoftBank Group and OpenAI.
- The site runs Nvidia's full-stack DSX AI factory platform: GPUs, CPUs and networking.
SB Energy and SoftBank will build at least 10 GW of new generation to yield those 8 IT-GW of compute capacity, and invest at least $4.2 billion in regional grid infrastructure through a partnership with AEP Ohio structured to protect existing ratepayers.
The campus reindustrialises the decommissioned Portsmouth Gaseous Diffusion Plant, a former uranium enrichment site, spanning private and federal land in Appalachian Ohio. There is an initial $80 million community benefits fund, and the project is projected to create tens of thousands of jobs.
Goldman Sachs and JP Morgan advised SB Energy. Morgan Stanley advised Nvidia.
The bottleneck moved
Jensen Huang's framing was that AI is becoming infrastructure, and that land, power and shell have become vital in the age of AI. That is a description of where the constraint went, not a slogan.
For a decade the limit on AI was chips. Nvidia solved that by making them. The limit now is interconnection queues, transformer lead times, and whether a utility can energise a gigawatt-scale campus this decade. Shipping more GPUs does not fix any of that. Financing substations does.
So Nvidia is doing something chip companies historically avoid: taking balance-sheet risk on the physical plant surrounding its product.
The circularity question
This is the third leg of an arrangement that has drawn obvious scrutiny. Nvidia invests in SB Energy, SB Energy houses compute for OpenAI, OpenAI buys Nvidia systems. Huang has publicly rejected the "circular financing" characterisation and pointed to the scale of underlying demand.
Both things can hold at once. The demand can be real and the financing structure can still concentrate risk. If AI compute demand grows into these commitments, this reads as foresight about the true bottleneck. If it does not, a chip company is carrying twenty-year exposure to Appalachian real estate.
Sam Altman focused on scale, describing enough computing power for millions of people, from finding new medicines to starting businesses. Masayoshi Son framed it around infrastructure built at unprecedented speed.
The number to track
Not the gigawatts. Track whether the 10 GW of new generation gets built on schedule, and whether the AEP Ohio ratepayer protections hold once the campus starts drawing load.
Those two things will show whether AI infrastructure at this scale can be built without the surrounding community paying for it. That question decides whether the next ten of these get approved.