Skip to content
Artificial intelligence. Clearly. RSS feed
AI NEWS AI

News. Evidence. Perspective.

Inside the 10-gigawatt gas, grid and compute system planned at Portsmouth

Ohio’s planned Portsmouth campus links unprecedented compute, generation and transmission targets. The useful question is not whether one 10 GW headline is impressive, but whether every dependent system arrives in the right sequence.

A planned artificial-intelligence campus at the former Portsmouth uranium-enrichment site in Ohio is so large that calling it a data-centre project hides most of the engineering. The public plan joins a 10-gigawatt computing campus to 10 gigawatts of new generation, at least 9.2 gigawatts of it natural gas, plus a new pipeline, high-voltage transmission and substations. Each layer has its own construction schedule, approvals and failure modes.[1][2]

The number is difficult to picture. Ten gigawatts is a power level, not an annual energy forecast. If a 10-gigawatt load ran at that level for every hour of a non-leap year, it would consume 87.6 terawatt-hours. The same arithmetic puts 9.2 gigawatts of gas generation at 80.6 terawatt-hours if it ran continuously. Neither result predicts utilisation, outages or dispatch; both show why this cannot be evaluated as one building with a large electricity bill.

The facts

Portsmouth’s linked infrastructure plan

Portsmouth’s linked infrastructure plan
MeasureValueContext
Planned data-centre development10 GWThe announced maximum campus scale; not evidence that all capacity will arrive at once.Source 1
Planned new generation10 GWAt least 9.2 GW is described as natural-gas generation.Source 1Source 2
Transmission investment$4.2 billionSB Energy has committed to fund the planned AEP Ohio work, subject to approvals.Source 1Source 3
High-voltage buildNew 765 kV lines and four substationsThe electrical path is part of the project, not an invisible utility service.Source 1
Expected first power2029AEP Ohio’s stated expectation, not a guarantee of full campus completion.Source 3

The critical path is the slowest connected system

A server hall can be commissioned only when several other systems are ready: a firm site, cooling and water arrangements, electrical equipment, fuel delivery, generation, transmission, networking and enough compute hardware. A delay in one can strand investment in the others. That makes the sequencing more important than a single headline completion date.

The Portsmouth fact sheet says long-lead electrical equipment has already been secured and describes 765-kilovolt lines and four substations. AEP says power is expected to begin flowing in 2029. Those statements narrow one part of the schedule, but they do not establish that all 10 gigawatts of generation, transmission and compute will be available in 2029. Serious progress reporting should publish capacity by phase rather than repeat the ultimate target.[1][3]

At a glance

What a useful progress dashboard would separate

What a useful progress dashboard would separate
LayerQuestion that mattersEvidence of readiness
ComputeHow much IT load can operate?Commissioned megawatts, tested racks and cooling capacitySource 1
GenerationHow much power can run when required?Commercial-operation dates, fuel contracts and tested outputSource 2
GridCan power move safely in both directions?Energised lines, substations, interconnection rights and operating limitsSource 3
Site and cleanupWhich land is actually available?Lease milestones, remediation boundaries and permitsSource 4Source 1

“Excess capacity” needs an operating definition

The parties say excess generation and transmission capacity will be made available to the grid and that the developer will fund the new transmission. That commitment matters, but the public value depends on operating details: when capacity is excess, whether it is firm during peak conditions, which fuel and transmission constraints apply, and how costs and revenues are allocated. A nominal surplus at one hour is not the same as dependable capacity during a regional shortage.[1][3]

The IEA expects global data-centre demand to rise rapidly while emphasising substantial uncertainty in AI uptake, efficiency and energy-system bottlenecks. Portsmouth illustrates that uncertainty in physical form. A very large final target may be plausible while the delivered capacity curve remains staged, conditional and vulnerable to long lead times.[5]

The project is consequential precisely because it is more than a data centre. Its success will be measured not by the biggest number in the announcement, but by whether power plants, pipelines, transmission, substations, cleanup and computing capacity arrive in a coherent sequence—and whether the promised separation between project costs and ordinary customer bills survives real operation.

Follow the evidence

Sources & further reading

  1. Portsmouth Site U.S. Department of Energy · Accessed 8 Oct 2026
  2. Key Questions on Energy and AI International Energy Agency · 16 Apr 2026 · Accessed 8 Oct 2026