Planned US AI data centers could raise power-sector emissions 42%
A Sierra Club estimate puts power-sector emissions at 2.1 billion metric tons in 2030, though projections differ over renewable energy and construction plans.
Planned AI data centers in the US could add 618 million metric tons of carbon emissions from electricity generation between 2025 and 2030, according to a Sierra Club report obtained by Bloomberg. The environmental nonprofit projects a 42% increase, enough to erase 15 years of emissions reductions in the power sector.
Under that estimate, electricity generation would produce 2.1 billion metric tons of emissions in 2030. Sierra Club's projection of the AI expansion's impact is between two and four times the amount it would have forecast a year and a half earlier.
The outlook remains uncertain because renewable energy could meet much of the additional demand, and some planned facilities may never be built. Rhodium Group projects that power-sector emissions in 2040 would remain 24% below 2025 levels even under its scenario with the highest emissions.
Some developers are arranging their own electricity supplies rather than waiting for grid connections. Amazon's planned data center in Texas would draw power from a 7.65-gigawatt natural gas plant with permission to emit up to 33 million tons of CO2 annually. That could make the plant the country's largest single CO2 emissions source.
