In Loudoun County, Virginia, the hum of cooling fans never stops. The county hosts more data center square footage than any other jurisdiction on Earth, according to commercial real estate firm CBRE’s market tracking. Local utility Dominion Energy has warned that new campus requests are arriving faster than substations can be built to serve them. What is happening in Northern Virginia is a preview of a national problem: data centers are devouring electricity at a pace that could reshape the U.S. power grid by the end of this decade.
As of 2023, data centers consumed about 176 terawatt-hours of electricity, according to Lawrence Berkeley National Laboratory’s congressionally mandated energy usage report. Set against the roughly 4.178 trillion kilowatt-hours the nation generated that year, per the U.S. Energy Information Administration, that works out to just over 4 percent of all U.S. electricity. Those two federal figures form the most reliable baseline for understanding what comes next.
What comes next, according to the Electric Power Research Institute, is a dramatic acceleration. In a state-by-state analysis summarized in a February 2026 press release, EPRI examined operational facilities, projects under construction, and publicly announced expansion plans across every state, then modeled scenarios through 2030. Its conclusion: data centers could consume between 9 and 17 percent of U.S. electricity generation by the end of the decade. Even the low end of that range represents more than a doubling from the current share. The high end would mean data centers alone rival the residential air-conditioning load across several southern states.
Why the range is so wide
The gap between 9 percent and 17 percent is not a rounding error. It reflects genuine uncertainty about how many announced projects will actually get built. Corporate press releases about new data center campuses often precede final investment decisions by years. Some projects never break ground. Historically, cancellation rates for large data center developments have been significant. At 9 percent, utilities would need to add tens of gigawatts of firm generating capacity beyond what is already in interconnection queues. At 17 percent, the required buildout would likely exceed what the current permitting and construction pipeline can deliver by 2030.
The headline projection of roughly 12 percent by 2028 is the author’s interpolation, not a figure drawn from any single study. It sits near the midpoint of EPRI’s 2030 range and assumes that much of the anticipated growth materializes on an accelerated timeline. Whether construction schedules, supply chain bottlenecks, and utility interconnection backlogs allow that pace remains an open question as of mid-2026. Federal statistics on electricity consumption will offer the clearest way to track actual demand against these projections as new monthly and annual data become available.
The fuel question behind the growth
Where the electricity comes from matters as much as how much is consumed. The International Energy Agency has flagged AI-driven data center demand as a structural challenge for global grids. Natural gas and coal still supply the bulk of power feeding these facilities in many regions. Whether new U.S. data center load will be met primarily by renewables, gas-fired plants, nuclear, or some combination depends on utility procurement decisions still being negotiated across dozens of states.
Some of the largest deals are already pointing toward nuclear. Microsoft signed a 20-year power purchase agreement with Constellation Energy to restart a unit at the Three Mile Island plant in Pennsylvania. Amazon acquired a data center campus adjacent to a Talen Energy nuclear plant in the same state. These arrangements suggest that at least some hyperscale operators view firm, carbon-free power as worth paying a premium for. They remain exceptions rather than the industry norm.
What this means for electricity bills and grid reliability
For the roughly 130 million U.S. households that depend on the same grid, the surge in data center demand raises pointed questions about cost and reliability. When a single data center campus requests hundreds of megawatts of capacity, the resulting infrastructure costs can be substantial. Transmission upgrades, substation expansions, and generation additions needed to serve that campus are often socialized across all ratepayers in a utility’s territory. State public utility commissions in Virginia, Georgia, and Texas have begun scrutinizing whether data center operators should bear a larger share of those costs.
Reliability is an equally pressing concern. Grid operators in high-growth corridors, including Northern Virginia, central Texas, and parts of the Midwest, have warned that interconnection queues for large loads now stretch out for years. Substations designed for incremental residential and commercial growth are being overwhelmed by single-campus requests that dwarf anything in their planning models.
Historically, U.S. utilities expected demand growth of roughly 1 percent per year, driven by population shifts and gradual economic expansion. Large industrial loads like factories or refineries were significant but relatively rare and geographically dispersed. Hyperscale data centers break that pattern. They arrive in clusters, operate around the clock, and flatten daily load curves. That profile reduces the flexibility grid operators once had to schedule maintenance or lean on demand-side management programs.
Three strategies regulators and utilities are weighing in mid-2026
Policy responses are beginning to coalesce around three broad approaches. The first is accelerating transmission and generation buildout. This includes both renewable projects and gas-fired peaker plants that can respond quickly to load spikes. The second is tightening efficiency expectations for data centers themselves. Building codes, voluntary standards, and procurement requirements can push operators toward more efficient cooling systems, higher server utilization rates, and waste-heat recovery. The third is experimenting with location-based incentives or constraints. The goal is to steer new facilities toward regions with abundant clean power or surplus capacity rather than letting them pile into already strained hubs.
None of these strategies will resolve the underlying tension between rapid digital expansion and the physical limits of the grid. But the evidence now on the table, from federal measurements to EPRI’s scenario modeling, makes one thing clear: the planning assumptions that guided U.S. utilities for decades no longer hold. The power sector is being asked to keep pace with an industry that measures growth in months, not years. The gap between those two timelines is where the real risk lives.
More from Morning Overview
*This article was researched with the help of AI, with human editors creating the final content.
