Nuclear Power Loans Signal Infrastructure Bottleneck for AI Hyperscalers

The Trump administration is providing $5 billion in loans to Constellation Energy to expand nuclear generating capacity by 1.1 gigawatts across eight power plants, according to TechCrunch. This capital injection underscores a critical shift where energy infrastructure has become a primary bottleneck for hyperscale AI ROI, as tech giants compete for power to support data centers.

Constellation Energy intends to "uprate" 11 existing reactors—a process involving equipment upgrades and more highly enriched uranium—to increase capacity. To deliver power to Google, Amazon, and Meta, the company has already pledged more than $7 billion to increase nuclear capacity by 1,080 megawatts. While the Department of Energy claims the loans support "affordable, reliable power" for businesses and homes, TechCrunch reports there is no evidence the capacity will lower residential rates, and Constellation would not confirm the government's claim.

Most of the additional power from seven of the eight facilities has already been secured by hyperscalers. Specifically, Google is the confirmed buyer for power from four Illinois plants and one Pennsylvania plant. Electricity from the Maryland plant is being bought by Amazon, while Meta is effectively acquiring the entire Clinton power plant in Illinois. The eight plants involved are located in Pennsylvania (Limerick), Maryland (Calvert Cliffs), and Illinois (Braidwood, Byron, Clinton, Dresden, LaSalle, and Quad Cities).

Constellation stated that the pending loan commitment could lower financing costs for these projects. TechCrunch notes that hyperscalers are paying tens of billions of dollars for energy as power becomes a competitive advantage, meaning the government loans could potentially boost the bottom line for both the energy provider and the tech companies.