May 2026 Volume 8
ENERGY
DON'T BLAME AI FOR RISING ELECTRIC BILLS By Bernard L. Weinstein
I n recent years, U.S. electricity demand has experienced a significant surge, driven by colder winters, hotter summers, as well as the growth of data centers, artificial intelligence, and other power-hungry industries. Continued rapid demand growth is projected for at least the next decade. As a consequence, the cost of power has been escalating rapidly for most American households. Indeed residential electricity bills have outstripped the overall consumer price index (CPI) for the past two years. In 2025 alone, average residential bills jumped by 12%. Not surprisingly, higher bills have prompted a consumer backlash and a search for culprits. Some blame the Trump Administration’s energy policies while others point their fingers at grid operators and local utilities. For example, many older power plants have been retired while grid reliability has become a problem in some parts of the country. But the principal villain is alleged to be the rapid growth of artificial intelligence (AI) and the huge power demands required by the server farms that back up AI. Without question, AI and its components consume lots of electrons. Ten years ago, data centers required about 10 gigawatts of power. Last year, demand jumped to 40 gigawatts— consuming about 3% of America’s power generation capacity. According to a recent analysis by Goldman Sachs, data centers will constitute 40% of power demand growth through the end of the decade. But with sizable utility scale generation currently under construction, AI will only be drawing upon 5% of the nation’s power resources in 2030. Given AI’s relatively small share of power demand today and in the future, the argument these facilities are primarily responsible for rising electricity costs doesn’t hold up. A new study by Charles River Associates gives further credence to the argument AI is not the principal driver of higher power costs. In fact, the CRA study found that in 34 states, residential electricity rates have actually grown more slowly than the overall cost-of-living since 2020. Large rate increases have been concentrated in the Northeast, due to wholesale market price increases, and in California due to wildfire-related costs. The only states where data center demand was found to have significantly pushed up power prices were located in the PJM
regional interconnect, a transmission grid that covers a vast region spanning from the Mid-Atlantic to the Midwest. In his recent State of the Union address, President Donald Trump announced that most major tech companies had agreed to sign a “ratepayer protection pledge” under which future data centers would have to supply their own power needs, primarily through on-site generation. This initiative should be especially beneficial to residential ratepayers in the PJM service area; but a broader re-examination of the PJM markets model is also needed to help bring down electricity prices. In short, the prevailing narrative that AI is driving up electricity costs is largely inaccurate. Instead of railing against tech companies, politicians and regulators should focus their attention on accelerating transmission build out, hardening the grid, facilitating the permitting of new base-load power plants, and expanding natural gas pipelines. Additionally, since electricity prices have risen in PJM more than other parts of the country, state and federal officials need to rethink the efficacy of PJM’s market model and whether this framework is conducive to meeting future power demands, especially for AI investments. These steps can help ensure electricity remains affordable and reliable in the years ahead.
Bernard L. Weinstein is emeritus professor of applied economics at the University of North Texas, retired associate director of the Maguire Energy Institute at Southern Methodist University, and a fellow of Goodenough College, London.
Reprint from: https://www.realclearenergy.org/articles/2026/03/06/ dont_blame_ai_for_rising_electric_bills_1169067.html
FIA MAGAZINE | MAY 2026 10
Made with FlippingBook - Online catalogs