Google's Nuclear Deal Mixes Reactor Upgrades With Supply

Google and Constellation Energy agreed on October 6 to support 890 megawatts of additional nuclear capacity at operating reactors in the PJM electricity market. They also signed a separate agreement covering 2,700 megawatts from Constellation’s existing generation fleet. Those figures sit under one power deal, but they describe different things: a plan to produce more electricity and a contract involving plants already in service.

The 20-year agreement for the new capacity is intended to fund upgrades at 11 Constellation-owned nuclear units in Illinois, Pennsylvania and New Jersey. The other agreement runs for 15 years. Google gets a long-term arrangement for its electricity needs; Constellation gets a commercial basis for work on its reactors and for keeping existing generation in the PJM market. The companies expect the first increase in reactor output by 2028. That is a target for the first upgrade, not a date when all 890 megawatts will be operating.

PJM is the regional grid at the center of both agreements. Electricity from an upgraded reactor would enter that grid rather than travel along a dedicated line to a Google data center. This is a useful way to read the deal: Google can help pay for generation in the market where it needs power, while other users also draw from the market. How much new power ultimately arrives, and when, depends on the upgrades being completed.

One Deal, Two Kinds of Capacity

Constellation calls the 890 megawatts an uprate. The company plans to modernize equipment at plants that already operate, including turbines, steam generators and digital controls, so the reactors can deliver more electrical output. This is an attempt to extract additional capacity from existing sites. It does not require building a new reactor, but the extra output remains a project to deliver, rather than electricity already on the grid.

The 2,700-megawatt portion starts from a different position. It is a long-term supply arrangement involving Constellation’s operating PJM fleet. The arrangement is not tied to a specific generation source. Constellation describes this part as providing revenue certainty for existing assets and keeping their energy and capacity in the PJM market. That could be valuable: retaining a working plant can matter when electricity demand is growing. But a contract for output from existing assets does not add 2,700 megawatts of generating capacity by itself.

The 2,700 megawatts come from unspecified sources. Readers should resist treating that entire portion as newly produced nuclear electricity. The stronger, narrower claim is that the deal supports planned nuclear uprates and separately contracts for supply from Constellation’s existing fleet.

The companies put Constellation’s planned investment at more than $4.3 billion and say the work will create roughly 7,200 construction jobs while sustaining about 4,400 existing jobs. Those are company projections attached to the agreement, not completed expenditures or a count of jobs already created. They do, however, make clear why Constellation wants a long contract before undertaking upgrades across 11 units. A buyer’s commitment can make a capital project easier to plan; it cannot do the engineering work on the buyer’s behalf.

Google and Constellation also agreed to incorporate ways for Google to reduce non-critical electricity use during periods of grid stress. That provision addresses the demand side of the arrangement. Its practical value will depend on how much use Google can shift at the times when the grid needs relief. The agreement names a tool for managing pressure on PJM alongside the plan to add supply, rather than treating a power purchase as the whole answer.

What a Restart Shows About an Upgrade

Constellation has already made a different nuclear wager with another technology customer. In 2024, Microsoft signed a 20-year agreement for power associated with the planned restart of Three Mile Island Unit 1 in Pennsylvania. The 835-megawatt unit had shut down in 2019; Constellation’s chief executive attributed the closure to poor economics. Its proposed return requires work on major plant systems and regulatory approval to operate again. The company’s 2028 restart date was its plan in 2024, not a measure of the project’s current progress.

The name Three Mile Island carries history of its own. Unit 2 suffered a partial meltdown in 1979. Microsoft’s agreement concerns Unit 1, the separate reactor that later closed for economic reasons. Keeping those units distinct makes the precedent more informative: the Microsoft project concerns the return of a retired generating unit, with a substantial recommissioning process between a contract and any new output.

Google’s project begins with reactors that are operating and seeks to raise their output across multiple units. Microsoft’s contract was tied to bringing a closed unit back into service. Both deals use long-term technology-company demand to support nuclear projects in PJM, and both attach a buyer to generation that could serve the wider grid. Their starting conditions create different obstacles. A restart needs authority to load fuel and run a previously closed reactor; an uprate needs equipment changes to produce the promised extra output from working plants. Calling both simply “nuclear deals” conceals the work each one must complete.

The history also explains the commercial shift. Three Mile Island Unit 1 closed when Constellation said its economics no longer worked. Five years later, a technology company agreed to buy power associated with its proposed return. Google’s agreement applies that long-contract logic to a fleet that is still running: support upgrades where new output is possible, and contract separately for generation already on the market. The common thread is a customer willing to make a commitment spanning decades. The physical result depends on the plant work each commitment is meant to support.

For anyone assessing whether Google’s electricity purchases will ease pressure on PJM, the first figure to watch is the uprates’ delivered output. The second question is whether the existing plants covered by the supply agreement remain in service. The contract addresses both questions, but only the reactor upgrades promise additional generating capacity.