Last July, a Houston-area homeowner posted a complaint that resonated across the Sun Belt: “I keep the thermostat at 78 and it still cost me $290.” Average summer residential electricity rates in the U.S. have climbed in six of the last seven years, according to the Energy Information Administration’s Electric Power Monthly. By the summer of 2026, those bills could climb further, driven by a force most households never see on their invoices: the explosive growth of data centers built to power artificial intelligence. While no utility has yet isolated the data-center share of residential rate increases, the EIA’s own modeling of rising wholesale electricity costs suggests households in the most affected regions could see summer bills rise by a meaningful single-digit percentage above 2025 levels, on top of increases already baked into approved rate cases.

The EIA now projects the sharpest four-year increase in national electricity demand since the early 2000s, measured from roughly 2022 through 2026, and it identifies AI-driven computing as the primary accelerant. Data center load is the dominant new source of consumption growth in both the agency’s Annual Energy Outlook 2025 and its rolling Short-Term Energy Outlook. Federal regulators, meanwhile, are raising pointed questions about who will pay for the grid upgrades all that new demand requires.

Data centers are rewriting the demand curve

A single large AI data center can consume 100 megawatts or more of electricity, roughly enough to power 80,000 homes. Clusters of these facilities, concentrated in Virginia, Texas, and Georgia, are adding tens of gigawatts of projected load to regional grids that were already running tight during peak summer months.

The scale is new, but so is the intensity. According to the International Energy Agency’s 2024 analysis of data center energy use, a modern AI-optimized facility can consume several times more electricity per square foot than a conventional enterprise data center, which itself dwarfs the energy footprint of a typical office building. That density is what makes AI-era load growth qualitatively different from anything grid planners have dealt with before.

The EIA’s Short-Term Energy Outlook traces how that demand feeds into wholesale power markets. Natural gas still fuels the largest share of U.S. electricity generation. When data centers bid up consumption during the hottest weeks of the year, gas-fired plants run harder, wholesale auction prices rise, and those costs eventually land on residential bills. The agency’s modeling draws a direct line from AI-related load growth through fuel markets to the retail rates households actually pay.

Whether summer 2026 rates cross into record territory nationally depends on natural gas prices, regional grid conditions, and regulatory decisions still unfolding as of May 2026. The upward pressure is well documented. The record-breaking threshold is plausible but not yet confirmed.

Regulators are worried about cost-shifting

The Federal Energy Regulatory Commission has framed large-load growth from data centers as both a reliability risk and a consumer-protection problem in its forward-looking reliability and market assessment for 2026. The commission has stressed that large electricity users must pay their fair share for transmission upgrades, new substations, and backup capacity, rather than shifting those costs onto residential ratepayers.

FERC commissioners have repeatedly warned in public meetings that the costs of serving massive new data center loads should not fall disproportionately on families who had no say in where those facilities were built. Former Commissioner Allison Clements, whose term ended in 2025, was among the most vocal on this point during her tenure, and the commission has carried that position into its current agenda.

A key flashpoint is co-location: the practice of connecting a data center directly to a power plant or substation. When a tech company negotiates a co-location deal, the terms can determine whether the surrounding community helps pay for reinforced transmission lines and additional generation capacity. FERC’s 2026 agenda includes heightened scrutiny of these arrangements, but the commission has not yet issued binding rules or taken public enforcement action specifically targeting data center interconnections.

That gap matters. Without clear federal mandates, cost allocation falls largely to state utility commissions, which vary widely in their willingness and capacity to push back against well-resourced corporate applicants.

What the forecasts still cannot tell us

Federal projections paint a convincing directional picture, but important pieces are missing. The EIA does not publish granular, state-by-state load forecasts tied specifically to data center construction, so it is difficult to say precisely how much harder the grid will be squeezed in Northern Virginia’s “Data Center Alley” compared to, say, rural Oregon.

Utility-level detail is similarly thin. No major provider has filed a public rate case that isolates the residential bill increase attributable to data center demand alone. Rate filings typically bundle multiple cost drivers, from fuel prices to storm-hardening investments, making it hard to assign a specific dollar figure to the AI effect before bills actually arrive.

Major tech companies, including Microsoft, Google, and Amazon, have made public commitments to power their data centers with clean energy and invest in grid infrastructure. Those pledges are real, but neither the EIA’s demand forecasts nor FERC’s reliability analyses currently show a measurable link between renewable co-location at data centers and lower residential rates. Until that evidence materializes, the promises remain forward-looking, not proven.

What to watch before the air conditioners switch on

The federal record as of May 2026 supports a clear conclusion: AI-driven electricity demand is rising faster than at any point in a generation, and the infrastructure spending required to meet it will be enormous. The unresolved question is not whether costs will increase, but how they will be divided.

Three developments in the coming weeks will determine whether summer 2026 bills set new highs or whether policy catches up in time to cushion the blow:

• State rate cases. Filings before public utility commissions in data-center-heavy states like Virginia, Texas, and Georgia will reveal how much of the infrastructure tab utilities plan to pass to residential customers.
• FERC action on co-location. Any rulemaking or enforcement targeting data center interconnection agreements would signal that federal regulators are moving beyond rhetoric.
• Summer wholesale prices. Real-time auction data from regional grid operators like PJM and ERCOT during June, July, and August will show whether AI-era demand is translating into the price spikes forecasters expect.

For families budgeting for the months ahead, the safest assumption is that summer electricity bills will be higher than last year, potentially meaningfully so. The scale of the increase will depend on where you live, how your utility allocates costs, and whether regulators act before the hottest months arrive.