SB Energy Ohio AI Data Centre Backed by 10 GW of Dedicated Power
SoftBank Group and its SB Energy subsidiary have committed to building 10 GW of new generation in southern Ohio, at least 9.2 GW of it from natural gas, to supply a data centre campus on leased federal land. The SB Energy Ohio AI data centre development carries $4.2 billion of new 765 kV transmission that SB Energy will fund itself, a structure built to keep interconnection costs off local electricity bills.
The campus occupies the former Portsmouth Gaseous Diffusion Plant in Pike County, and the Department of Energy published a fact sheet on the arrangement on 10 September, describing it as a public-private partnership meant to power AI growth in Ohio without shifting costs onto the state's electricity customers. Nvidia, OpenAI and SB Energy first set out the plan on 17 August.
Capacity is measured in two currencies. The 10 GW of generation is expected to yield 8 IT-GW of AI factory capacity, and Nvidia has guaranteed that the PORTS-Pike Technology Campus will host Nvidia compute exclusively. The first 800 MW is due in 2028, drawing largely on existing AEP Ohio infrastructure while the dedicated plant is built.
Powering the SB Energy Ohio AI Data Centre
The generation mix is overwhelmingly fossil. Natural gas accounts for at least 9.2 GW of the 10 GW commitment, leaving roughly 800 MW of headroom for other sources the partners have not specified. For the SB Energy Ohio AI data centre, the fuel choice reflects a scheduling problem: gas turbines can be ordered, sited and commissioned faster than any alternative at this scale, and the campus load arrives on a fixed date regardless of what the grid can spare.
The transmission piece is the more unusual commitment. SB Energy will spend $4.2 billion on 765 kV lines itself rather than route the cost through AEP Ohio's rate base. That removes the interconnection charge that has become the flashpoint in other large-load disputes, where existing customers are asked to underwrite upgrades built for a single tenant. The trade is that SB Energy carries construction risk on assets whose return depends on the campus reaching full load.
SB Energy is building at 765 kV, a bulk transmission class chosen to move large blocks of power over long distances with fewer circuits than lower voltages require. The partnership with AEP Ohio keeps the utility inside the regional grid work even though SB Energy funds the new lines, which matters for a project that still leans on the incumbent network for its first tranche of capacity.
Roughly $33.3 billion of Japanese funding is attached to the natural gas generation component, and a $40 million community benefits agreement covers the surrounding area.
| Component | Committed figure |
|---|---|
| New generation | 10 GW |
| Natural gas share | at least 9.2 GW |
| AI factory capacity | 8 IT-GW |
| 765 kV transmission | $4.2 billion, funded by SB Energy |
| Japanese funding, gas generation | about $33.3 billion |
| Community benefits agreement | $40 million |
| First power tranche | 800 MW in 2028 |
The New Default for AI Sites
Dedicated generation sized to a single AI hub has been the exception rather than the rule. Operators have typically signed power purchase agreements with existing plants or bought renewable credits while drawing from the grid. The Ohio build inverts that sequence: generation and transmission come first, sized to the campus, and the grid connection supplements rather than underpins the load. If the model spreads, the constraint on AI expansion shifts from chip supply to the ability to finance and permit power assets on a two-to-three-year clock.
Ohio is not the only state where large loads are being paired with dedicated supply, but the scale here is unusual. At 10 GW, the generation commitment is larger than the typical new-build pipeline a single developer would carry, and it is being assembled for one buyer. That concentration explains the transmission spending, since the capacity serves a single customer rather than a regional market.
Three Balance Sheets, One Site
Nvidia has moved past the role of compute supplier. It has invested $1.5 billion in SB Energy and is backing up to $105 billion in financing tied to the Ohio campus, a lease guarantee that lowers the risk on OpenAI's occupancy. OpenAI has leased the campus capacity, so the offtake for the power plant is effectively pre-sold before most of the generation exists.
That arrangement spreads the exposure. SoftBank and SB Energy own the land, the generation and the wires. Nvidia carries chip supply and credit support. OpenAI carries the demand. No single party absorbs the full cost of a build at this scale, which is how the project reached a size single-tenant data centre deals rarely approach.
Across the SB Energy Ohio AI data centre programme, the gas plant and transmission together amount to roughly $37.5 billion of power infrastructure for one campus, before any building or compute spending. Nvidia's $105 billion financing backstop is larger still, which puts the power spend at about a third of the guaranteed capital stack. That ratio shows where the binding cost now sits.
The federal role is land. The campus sits on leased federal property left over from the Portsmouth site's previous industrial use, and the Department of Energy has framed the lease as a way to convert an underused asset into AI capacity while protecting household electricity bills. Whether that framing holds depends on the transmission commitment doing what it is designed to do.
Contracted Capacity, Unbuilt Capacity
SB Energy filed to go public on 1 September, and its IPO paperwork discloses 8.8 GW of contracted AI capacity with most of it still unbuilt. That gap is the central execution question. Turbines, 765 kV lines and campus buildings all have to arrive on schedule for the 2028 tranche to hold, and any slip pushes the load back onto a grid the project was designed to relieve.
Engineering capacity is already reserved. Bechtel and Kiewit were engaged on the complex as early as March. Ohio's transmission system remains the binding constraint, and the project's own phasing concedes the point by leaning on existing AEP Ohio infrastructure for the first 800 MW rather than waiting for the new plant to be energised.
Comparisons with earlier hyperscale builds are instructive. A campus of this size drawing from an existing grid would sit in an interconnection queue measured in years, competing with other projects for capacity that may never be approved. Owning generation and transmission removes the project from that queue, which is the practical value of the $4.2 billion outlay.
Delays have a cost that falls unevenly. SB Energy holds the transmission assets and the generation, so schedule slippage hits its own returns first. OpenAI's lease obligations run on the campus timeline, which Nvidia's guarantee backstops. The party least exposed to a slow build is the one whose demand created the load.
Ratepayer politics sit behind both the transmission pledge and the community benefits package. Large-load projects across the Midwest have drawn objections from residential customers who fear picking up the tab for infrastructure built for one tenant. Funding the wires directly is SB Energy's answer, and it is the clause that makes the rest of the package defensible with local officials.
Why This Matters
The Ohio commitment shows how AI infrastructure deals get restructured when power, not chips, is the scarce input. A hyperscale tenant that once negotiated for electricity from an existing grid now finances generation and transmission directly, and the federal government is leasing land to make it happen. For decision-makers weighing comparable sites, the template here is the ratepayer question: whoever pays for the wires decides whether a project survives local opposition. SB Energy has bet $4.2 billion that owning that answer is cheaper than arguing about it.
Sources
Related Articles
- Nvidia SB Energy investment talks target $3B stake in Ohio data center [Update]
- SB Energy IPO puts OpenAI dependence at the heart of SoftBank's AI power play
- SoftBank to Build 5GW of AI Data Center Capacity in France Through €75 Billion Investment
✔Human Verified
Researched and cross-referenced against primary sources by the Bytevyte editorial team. This article was generated with the assistance of artificial intelligence and reviewed by the Bytevyte editorial team.