US power-sector emissions are on track to hit 2.1 billion metric tons by 2030, up 42% from 2025 and enough to erase 15 years of cuts, according to Sierra Club projections reported by Bloomberg. The expected jump is tied to surging electricity demand from data centers, including off-grid sites that could consume roughly as much power as Mexico.
The projection lands as tech companies race to add computing capacity for AI, and as the Trump administration moves to underwrite new power supply. On Tuesday, the Department of Energy said it would lend Constellation Energy $5 billion to expand nuclear generation, with most of the new output already committed to Google, Amazon and Meta.
Why AI data centers are driving US power emissions
Sierra Club's estimate points to a reversal after years of decline in US power emissions. Off-grid data centers — facilities that build their own generation rather than drawing from the public grid — are a major part of the increase in the projection, according to Bloomberg.
Hyperscalers including Google, Amazon and Meta have been searching the grid for extra capacity to support AI data centers, and they are paying heavily for it. Their commitments run into tens of billions of dollars over the coming years, with power now treated as a competitive advantage.
Constellation Energy nuclear loans and the data centers behind them
The Department of Energy is providing $5 billion in loans to Constellation Energy to expand nuclear generating capacity at eight power plants by 1.1 gigawatts, TechCrunch reported. Constellation said it plans to "uprate" 11 existing reactors — an industry term for expanding capacity, often by using more highly enriched uranium and upgrading equipment for higher power levels.
The company said it had already committed to spending more than $7 billion to add 1,080 megawatts of nuclear capacity for Google, Amazon and Meta. The loan commitment has yet to be finalized and "could lower financing costs" for the projects, Constellation said in a press release.
