Featured image: Electricity transmission pylons illustrate the shared grid served by generation. Contextual photograph; not a Constellation nuclear site. Photo: Matthew Henry / Unsplash. Unsplash licence.
A large power contract can secure electricity for a customer, add generation to a grid, or do both. Google and Constellation’s latest agreement makes those different functions visible in the numbers.
On 6 October 2026, the companies announced a 20-year agreement supporting 890 MW of additional nuclear capacity and a separate 15-year supply agreement for 2,700 MW from operating generation. Together they amount to 3,590 MW, or 3.59 GW. The total should not be described as 3.59 GW of newly built nuclear capacity. Read the official announcement.
New output and existing supply perform different jobs
The 890 MW is planned additional output across 11 Constellation nuclear units in Illinois, Pennsylvania and New Jersey. The companies say the first increase is expected by 2028. The additional 2,700 MW agreement concerns existing generation in PJM’s fleet. The two parts of the commercial deal.
PJM operates a regional electricity system in the United States. Generation enters a shared network that serves many users. Buying supply and increasing the network’s total generation are related commercial activities, but they measure different changes.
The distinction is useful whenever a corporate announcement combines several agreements. A headline total can obscure how much supply is newly added, how much was already operating and when the added portion is expected to arrive.
Our assessment is that the new-capacity figure is the most relevant starting point for evaluating the expansion claim. The existing-supply agreement can provide commercial support, but its megawatts should be counted separately when discussing additional generation.
An uprate requires engineering and a safety case
Increasing output at an existing nuclear plant is called a power uprate. The US Nuclear Regulatory Commission’s background explanation describes changes that allow a reactor to produce more thermal energy, with the resulting steam driving electricity generation.
The regulator says operators must show that the changed configuration remains safe. Its review covers the technical analyses supporting the request. A purchase agreement supplies a commercial reason to invest; it does not itself grant a reactor permission to run at a higher licensed power level.
The Department of Energy’s UPRISE initiative provides broader policy background. It includes increasing existing-reactor output, restarting dormant facilities and pursuing other capacity additions. The programme’s launch in March 2026 predates today’s deal.
This portfolio approach has a different construction path from building a new reactor on a new site. It still involves equipment, project sequencing, operating requirements and regulatory decisions. The first-delivery target is a milestone to monitor rather than electricity already available today.
Timing matters as much as the headline capacity
The International Energy Agency’s 2025 Energy and AI analysis estimated data-centre electricity use at about 415 TWh in 2024 and projected approximately 945 TWh in 2030 in its base case. Those are a historical estimate and a scenario respectively, not a measured 2030 outcome.
The same analysis highlights a mismatch between relatively fast data-centre development and the longer planning and construction periods of the energy system. A TWh measures energy used over time; a MW measures a rate of supply or demand.
For this agreement, the practical comparison is between the schedule of added generation and the schedule of added electricity consumption. The full contract capacity on an announcement date cannot establish that those schedules already align.
The technology alliance also needs measured outcomes
The companies additionally announced a five-year Google Cloud and Gemini Enterprise technology alliance for energy operations. Its stated aims include planning, generation optimization and infrastructure protection. These are announced deployment intentions.
Evidence of success would require defined measurements, such as demonstrable operational improvements under stated conditions. An AI partnership is not, by itself, proof of faster regulatory approval or lower household electricity bills.
The deal’s significance is the pairing of long-term demand with a specific planned addition to an existing generation fleet. The next tangible evidence will be approved projects, completed upgrades and output delivered to the grid. Those milestones will show how much of the announced expansion becomes operating capacity.
Leave a Reply