AI Data Centers Could Erase 15 Years of America's Climate Progress by 2030

A Sierra Club analysis projects that U.S. electricity-sector emissions could reach 2.1 billion metric tons of carbon dioxide by 2030, about 42% above 2025 levels - enough to undo roughly 15 years of emissions reductions. The rapid expansion of AI data centers is a major concern because clean power and grid connections are not keeping pace with demand. To avoid lengthy utility interconnection queues, some developers are building or contracting for power plants on-site, often using natural gas. These off-grid generators could emit more than 460 million metric tons of CO2 annually by 2030, while sources serving grid-connected data centers could add another estimated 154 million tons.

The projections describe a high-impact scenario, not a guaranteed outcome. Turbine shortages, delayed connections, and canceled projects could lower emissions, and earlier estimates were substantially smaller. Still, the underlying incentive to use fossil fuels for fast, reliable power remains. Existing renewable-energy requirements may also take years to reach full effect, potentially after new facilities are already operating at scale. The proposed remedy is to require genuinely additional clean energy and transmission for new data-center capacity, rather than letting developers claim existing renewable power or default to gas. Faster interconnection processes and rules that include off-grid generation in climate oversight are also presented as necessary to keep the AI buildout from undermining emissions progress.