The artificial-intelligence boom is creating an energy problem that could reach far beyond crowded power grids. A new Sierra Club analysis projects that planned data-center construction could push U.S. electricity-sector carbon dioxide emissions to roughly 2.1 billion metric tons by 2030—about 42% above 2025 levels and enough to wipe out roughly 15 years of emissions reductions.
The analysis estimates that the additional facilities could add approximately 618 million tons of carbon dioxide pollution by the end of the decade. That increase is comparable to the annual electricity-related emissions of more than 128 million U.S. homes. The calculation assumes that planned AI projects move forward and that utilities rely heavily on new fossil-fuel generation to serve them.
The biggest concern is not limited to data centers that connect directly to the existing grid. Sierra Club estimates indicate that “behind-the-meter” or off-grid projects—facilities supplied by their own generators—could account for more than 460 million tons of the additional pollution. Grid-connected generators, meanwhile, could add about 154 million tons, raising baseline power-sector emissions by roughly 10%.
That scenario reflects a rapidly expanding electricity appetite. The Lawrence Berkeley National Laboratory’s 2025 update estimates that data centers could consume 11.8% of all U.S. electricity by 2030, with possible outcomes ranging from 9.5% to 15.3%. The International Energy Agency likewise says data centers are positioned to drive almost half of U.S. electricity-demand growth through 2030.
Natural gas is emerging as the quickest way for developers and utilities to add dependable capacity. The IEA says natural gas already supplies more than 40% of the electricity used by U.S. data centers and could provide more than 130 additional terawatt-hours of annual generation by 2030. On-site gas turbines can also let projects bypass lengthy grid-connection queues, but they may operate with lower efficiency and fewer of the pollution controls associated with large centralized plants.
The forecast is not a guarantee that every announced facility will be built, and future emissions will depend on construction timelines, efficiency improvements, renewable generation, storage, and whether utilities approve new gas infrastructure. Still, the Sierra Club’s estimate is higher than several earlier projections, underscoring how quickly AI-related demand has changed the energy outlook. For technology companies, the pressure now extends beyond purchasing enough electricity: they will also face growing scrutiny over whether that power comes from clean generation or locks communities into decades of fossil-fuel dependence.
