**Categories:** International, Energy **Tags:** Google, Constellation Energy, nuclear energy, data centers, US power grid **Excerpt:** Google's agreement covers around 3.6 GW of electricity on the PJM grid. The genuinely new share is 890 MW, enabled by power increases at eleven existing reactors, while a separate contract covers 2.7 GW of generation already available. **Slug:** google-constellation-36-gw-electricite-nucleaire-pjm **Image alt:** Technicians carrying out turbine maintenance at a US nuclear power plant.
Google and Constellation Energy have announced a set of contracts covering around 3.6 gigawatts of electricity on the PJM grid, which serves a large part of the eastern United States. The agreement responds to strong growth in energy needs associated with data centers and artificial intelligence. Its significance lies as much in its structure as in its volume: one portion is intended to finance new nuclear capacity, while the remainder relies on generation from plants already operating.
The main component is a twenty-year power purchase agreement intended to enable the addition of 890 megawatts of nuclear capacity. Constellation plans to invest more than $4.3 billion in equipment and upgrades at eleven units in Illinois, Pennsylvania and New Jersey. These power uprates aim to produce more electricity from existing reactors without building a new plant.
A second agreement, lasting fifteen years, covers 2,700 megawatts from Constellation's current nuclear fleet. Together, the two volumes total around 3,590 MW, or nearly 3.6 GW. This distinction matters: presenting the entire contract as new nuclear generation would overstate the additional capacity actually created by the agreement.
Investment in the existing fleet
Power increases, or uprates, involve modernizing equipment and improving the efficiency of operating plants to raise their authorized and available capacity. They can bring dispatchable generation online faster than building a new facility, which requires longer procedures, construction and timelines. They nevertheless require safety studies, regulatory approvals and appropriate maintenance periods.
Under the announced plan, the first additional capacity is expected from 2028. The schedule will depend on progress at the different units and the necessary approvals. Constellation's plan would also involve thousands of direct and temporary jobs, particularly for modernization work. These benefits matter for the states concerned, but should not obscure the question of the final cost of the work and how it is allocated.
For Google, long-term nuclear supply can help reduce exposure to market fluctuations and provide electricity available at all hours. For the energy company, a major buyer's commitment provides commercial visibility that supports investment. This model illustrates how large technology companies are becoming more direct participants in energy financing.
The PJM grid challenge
Data center demand is increasing pressure on PJM, one of the largest electricity markets in the United States. New connections require lines, substations and sometimes additional generation. Grid construction timelines can become as decisive as power plant availability. An electricity contract alone does not guarantee that power can be delivered to sites when it is consumed.
Cost allocation is therefore central to the debate. Regulators and grid operators must ensure that a large consumer's specific needs do not automatically translate into higher bills for households and small businesses. Conversely, network investments can benefit several categories of users if financing rules correctly reflect the services provided.
The contract also combines energy accounting and physical availability issues. Electricity flows through a shared grid, where generation and consumption must remain balanced in real time. Long-term purchasing helps finance or hedge volumes, but does not mean that every electron consumed by a data center physically originates from a specific reactor. The system must ensure availability through all connected assets.
AI as an accelerator of energy contracts
The agreement also includes cooperation involving Google Cloud and its artificial intelligence tools to improve plant and grid management. Using software to optimize maintenance, forecast needs or facilitate infrastructure operations could improve operational efficiency. These gains will nevertheless depend on data quality, cybersecurity and validation of sensitive decisions.
This contract confirms that digital growth is changing how electricity is purchased. Companies now want to secure large volumes over long periods, sometimes by participating directly in investment. This evolution can help finance low-carbon, dispatchable assets. In return, it requires greater transparency about costs, timelines, avoided emissions and genuinely new capacity.
Google and Constellation's decision does not, on its own, resolve the US system's constraints. It nevertheless shows how predictable commercial demand can help extend and modernize the existing nuclear fleet. The indicators to track will be the actual commissioning of the additional 890 MW and PJM's ability to connect new data centers without shifting disproportionate costs onto other users.