NVIDIA guarantees the PORTS-Pike campus in Ohio
On August 17, 2026 NVIDIA announced it is guaranteeing a data center campus being built in Ohio to exclusively host its own AI compute. SB Energy will build and operate the facility and OpenAI will lease it for 20 years. The total target is 8 IT-GW with capacity arriving from 2028.
The campus, the capacity and the parties
Per the NVIDIA release dated August 17, 2026, the campus sits on and around the decommissioned Portsmouth Gaseous Diffusion Plant in Pike County, Ohio and is named the PORTS-Pike Technology Campus.
Capacity comes in two stages. An initial 4.25 IT-GW of AI factory capacity was secured with an option on the remaining 3.75 IT-GW, for a total target of 8 IT-GW. Per the release that capacity will be fed by 10 GW of new power generation.
The division of roles is clear. In the words of the release, SB Energy will build, own and operate the data center under a 20 year lease to OpenAI. NVIDIA, as the exclusive AI compute provider, is securing land, power and shell capacity and providing credit support on those elements.
On investment, NVIDIA is putting $1.5 billion into SB Energy. At least $4.2 billion is allocated to regional grid infrastructure. The community benefits fund reaches $80 million with a $40 million contribution from OpenAI. Capacity comes online in phases from 2028.
The figure absent from the official release: $105 billion
The most striking number in the deal does not appear in the NVIDIA release itself. Per reporting based on the 8-K filed with the SEC, NVIDIA provided residual value guaranties with a cumulative cap of $105 billion backing leases of roughly 4.25 GW of IT load. We separate that figure here because it is absent from the official announcement.
The same filing describes how the guaranties work. They take effect as individual leases commence, which is expected from 2028, and leases can run up to 20 years from individual commencement.
The risk sharing is structured this way: if OpenAI defaults or runs into payment difficulty, NVIDIA covers the shortfall up to the guaranteed minimum value. NVIDIA has options to take over, re-lease, sell, terminate or defer remedies for up to a year, and OpenAI has agreed to reimburse NVIDIA for amounts paid.
The executive statements in the release point the same direction. NVIDIA founder and CEO Jensen Huang says AI is becoming infrastructure, the foundation for intelligence in every industry. SB Energy co-CEO Rich Hossfeld says infrastructure is vital for the AI economy. OpenAI CEO Sam Altman says they want the people who live there to feel the benefits too, through good jobs. The release gives no exact employment figure, saying only that the campus will create tens of thousands of Ohio jobs.
The bottleneck is not the chip, it is land and power
The clearest thing this deal says is that the scarce resource in AI has moved. The debate was long about chip supply, while what is being secured here is land, power and building shell. A chip maker providing credit support on those three elements shows where the bottleneck has shifted. That assessment is ours.
Second, the financial structure itself. A hardware vendor guaranteeing its customer lease obligations moves the relationship from supply toward partnership. That is efficient for both sides if demand persists, and means risk accumulates with the guarantor if demand weakens.
Third, the scale. Feeding 8 IT-GW with 10 GW of new generation shows these facilities are now planned like power plants rather than industrial sites. The cost base of AI is being tied to the price of electricity.
Fourth, the calendar. Capacity begins in 2028 and leases run up to 20 years, so the effect of what was announced today shows up in the second half of the decade rather than this year. It does not explain today model prices, but it sets the capacity ceiling for the period ahead. That comment is ours.
What it means for businesses in Türkiye
The assessment below is not in the sources, it is our reading. The release contains no information about Türkiye.
There is no direct action. This story happens at the infrastructure layer and contains nothing an SMB here does today. The indirect consequence matters though: the cost of the AI tools you use is tied to capacity investments like this and to the price of electricity.
The practical response sits in cost planning. Treating your AI spend as a variable line item whose input is energy is more realistic than treating it as fixed. Gemini 3.7 Flash introductory pricing ending on December 31 the same week is the concrete example: prices move by period rather than only downward.
The second point is dependency. Capacity being tied up among specific players through long term agreements can reduce the flexibility to find alternative providers over time. Not tying a critical workflow to a single provider is a commercial precaution rather than a technical one. That inference is ours.
The UNALSOFT take
In the systems we build for clients on the agentic AI side we separate two things from the start: business logic and provider. Which model a workflow runs on has to be changeable, because as this story shows the cost base depends on variables outside our control. Energy prices, capacity agreements and introductory periods set your budget without being your decision. That is why we prefer flows built provider independent. A system that can be swapped when one provider gets expensive is more valuable even while that provider is still cheap.
Is your AI workflow provider independent?
Let us review your cost and dependency risk together.