OpenAI’s PORTS-Pike Bet Shows How the AI Boom Is Turning Into an Energy Boom

PORTS is emerging as a test of how much physical infrastructure the next generation of artificial intelligence will require: 10 GW of data-center capacity, 10 GW of new power generation, billions of dollars of transmission investment and an entirely new energy system built around computing demand.

The project is being developed in Pike County, Ohio, around the former Portsmouth Gaseous Diffusion Plant, a Cold War-era uranium enrichment site now being redeveloped for AI and advanced computing. The U.S. Department of Energy has described the planned facility as the world’s largest AI data center.

A 10 GW AI campus needs its own energy economy

SB Energy, a SoftBank Group company, plans to develop 10 GW of new power generation, including at least 9.2 GW of natural-gas capacity, to support a data-center complex of the same 10 GW scale.

The energy investment alone is enormous.

The $33.3 billion power component is linked to Japanese funding under the U.S.-Japan Strategic Trade and Investment Agreement. Separately, SB Energy plans to invest $4.2 billion with AEP Ohio in transmission infrastructure.

That puts disclosed power and grid commitments at roughly $37.5 billion before the full cost of the AI data-center campus itself is included. Ohio University previously noted that the investment in the AI data center had not yet been disclosed.

The grid buildout is almost as important as the data center

PORTS will require new 765-kV high-voltage transmission lines and four substations, according to the Department of Energy. Long-lead electrical equipment has already been secured.

SB Energy has committed to funding the transmission upgrades rather than passing their cost to Ohio electricity customers. Excess generation and transmission capacity is also expected to be made available to the regional grid.

The power system therefore isn't simply an accessory to the data center. It is becoming one of the project's largest capital investments.

That distinction matters for the broader AI trade. If future frontier-scale campuses follow the same model, AI demand could increasingly translate into orders for gas turbines, transformers, switchgear, transmission equipment, pipelines, substations and engineering services, not only GPUs.

35,000 workers — but only 2,500 permanent jobs

The latest PORTS figures also reveal where much of the employment generated by AI infrastructure could appear. The campus expects 35,000 construction jobs versus 2,500 long-term operating positions, a ratio of approximately 14 construction jobs for every permanent operating job.

Earlier DOE estimates were considerably smaller, calling for more than 10,000 construction jobs over four years and more than 2,000 operational jobs, plus tens of thousands of indirect jobs in manufacturing and services.

The newer 35,000 figure suggests the employment footprint has expanded alongside the scope and duration of the buildout.

Those jobs extend beyond construction of server halls. DOE lists power plants, interstate gas pipelines, transmission systems, substations and data centers among the infrastructure requiring skilled workers. Permanent roles are expected across electrical systems, facilities, maintenance, engineering and technical operations.

From uranium enrichment to AI training

The choice of location is unusual but strategic. The Portsmouth site was once part of the U.S. nuclear weapons and enrichment infrastructure. DOE is now leasing federal land there for the AI campus while environmental cleanup continues. One initial lease covers approximately 189 acres intended specifically for AI and cloud data-center development.

The project has also received FAST-41 coverage, making it the first project listed under the program's high-performance computing, advanced computer hardware and software sector. The designation is intended to coordinate and accelerate federal permitting and environmental reviews.

That may ultimately be the larger economic experiment at PORTS: whether enormous concentrations of cheap power and AI compute can become anchors for new industrial clusters.

SoftBank sees infrastructure as the foundation of the AI race

The project also fits into the much larger capital buildout surrounding OpenAI. OpenAI and SoftBank are already lead partners in Stargate, which was launched with plans to invest up to $500 billion over four years in U.S. AI infrastructure, beginning with an initial $100 billion deployment.

The numbers coming out of Ohio explain why AI infrastructure is increasingly becoming an energy story.

A frontier computing campus measured in 10 gigawatts cannot be built by buying GPUs alone. It requires power plants, pipelines, transmission lines, substations and thousands of construction workers before the first large-scale training workloads can run.

PORTS-Pike therefore offers a different way to read the AI capex boom. The semiconductor industry may supply the intelligence hardware, but an increasingly large share of the money needed to scale AI could flow through the power and industrial economy surrounding those chips.

With 35,000 construction jobs, 10 GW of planned generation and at least $37.5 billion already identified for power and grid infrastructure, PORTS-Pike is starting to look less like a conventional data center and more like a new industrial complex built around AI compute.