The surge of artificial intelligence has a physical footprint that reaches far beyond any screen. The warehouses full of computers that train and run modern AI models consume electricity on a scale large enough to reshape demand across whole regions, and utilities are scrambling to keep the power grid ahead of them.
Why AI is so power-hungry
Training a large AI model and then answering millions of queries with it requires enormous amounts of computation. That work runs on specialized processors packed by the thousands into data centers, each chip drawing far more power than a conventional server and generating heat that must be pulled away by energy-intensive cooling systems.
The result is a facility that can draw as much electricity as a small city, running around the clock. As AI adoption has accelerated, operators have raced to build ever-larger campuses, and each new one represents a fresh, concentrated load that a local grid must supply continuously rather than in the daily peaks and troughs of ordinary demand.
The numbers behind the strain
The scale is now well documented. A federal report produced by a national laboratory and summarized by the Department of Energy found that data centers consumed about 4.4 percent of total electricity in the United States in 2023, equal to roughly 176 terawatt-hours. That share had already more than doubled over the prior several years, driven largely by the growth of AI servers.
The projections are steeper still. The same analysis estimated that data centers could account for anywhere from about 6.7 percent to 12 percent of national electricity use by 2028, potentially tripling their consumption. Federal energy trackers, including the analyses published in the government’s daily energy briefings, have flagged this rise as one of the most significant new drivers of electricity demand in a generation.
The demand is also highly concentrated. Data centers tend to cluster in a handful of regions where land, fiber connections, and favorable power arrangements come together, so the strain does not spread evenly across the country. In those hot spots, a cluster of large campuses can dominate a local utility’s load forecast, magnifying the pressure on transmission and generation in ways that national averages obscure.
How the grid feels the pressure
For decades, electricity demand in the United States grew slowly, and utilities planned around modest, predictable increases. The sudden appearance of gigawatt-scale data-center projects has upended that assumption. In some regions, a single proposed campus can request more power than an entire town, forcing grid operators to reconsider how much generation and transmission capacity they need.
That pressure shows up as longer waits to connect new projects, calls for new power plants and transmission lines, and, in some areas, concern that supply could tighten. Because data centers run constantly, they raise the baseline load rather than just the peaks, which changes how utilities must size and operate the system.
The consequences for consumers and climate
The strain carries real stakes for ordinary ratepayers. When demand rises quickly and supply struggles to keep pace, the cost of building new infrastructure can filter into electricity bills, and households in fast-growing data-center corridors have seen upward pressure on rates. Reliability is another worry, since a grid stretched thin has less margin during heat waves and cold snaps.
There is a climate dimension as well. If the new demand is met by burning more fossil fuels, it can slow progress on cutting emissions, and some utilities have delayed the retirement of coal and gas plants to cover the load. The tension between soaring computational demand and clean-energy goals has become a defining energy question of the moment.
What the industry is trying
Operators and utilities are pursuing several responses at once. Technology companies have signed large contracts for wind, solar, and even nuclear power, including renewed interest in small modular reactors and in restarting shuttered nuclear plants, to secure firm, low-carbon supply for their campuses. Engineers are also working to make chips and cooling systems more efficient so that each unit of computing draws less power.
On the grid side, planners are exploring ways to have data centers ramp their demand up and down to help balance the system, along with faster processes for connecting new generation and building the transmission lines needed to move it. Some proposals would have large facilities run partly on their own dedicated power or shift flexible workloads to times when the grid has spare capacity, easing the peaks that most strain the system.
Whether those measures can keep pace with the appetite of the AI boom is now one of the central challenges facing the electricity system. The outcome will hinge on how quickly efficiency improves, how much clean generation can be built, and whether demand keeps climbing as fast as recent forecasts suggest. It is a question that will shape both energy bills and emissions for years to come.
This article was produced with the assistance of AI and reviewed by Morning Overview editors prior to publication.
More from Morning Overview
