An analysis by U.S. environmental groups points out that to meet the massive electricity demand of AI data centers, there has been a surge in the construction of fossil fuel power plants, which could offset the greenhouse gas emission reduction gains achieved by the U.S. power sector over the past 15 years. Based on AI projects included in planning, it is estimated that by 2030, carbon dioxide emissions from the U.S. power sector are expected to reach 2.1 billion tons, an increase of 618 million tons, or about 42%, compared to 2025. This projected scale represents a significant increase from a year and a half ago. To avoid lengthy grid connection wait times, many developers are opting for natural gas power generation, constructing their own power generation facilities in addition to drawing electricity from the public grid. It is projected that by 2030, most of the new emissions will come from these self-generated power facilities not directly connected to the public grid, with emissions potentially exceeding 460 million tons, close to Mexico's annual total emissions. About 80% of the energy from these self-generated facilities is likely to come from natural gas, which will not only drive up emissions and exacerbate local air pollution but has also sparked opposition to data center construction in multiple regions. Currently, there is a huge divergence in forecasts for future emissions from the U.S. power sector, with the gap in electricity generation between the highest and lowest estimates exceeding the combined electricity demand of California and Florida.