Data Center Energy Demand Pushes Power Bills Higher in 2026

Bills rise faster than electricity use

US households are paying more for each kilowatt-hour of electricity in 2026, and they are not using much more of it. Federal data for July 2026 show the average residential price up 4.9% from a year earlier, to 18.31 cents per kWh, while residential sales grew only 1.0%.26 Across all customer classes, the average price rose 4.4% and sales rose 1.9%.26 Over a longer period the rise is steeper. One analysis puts the national average monthly residential price up 15% in nominal terms from January 2025 through July 2026. For the first seven months of this year, prices ran nearly 40% above the same months of 2019.20

The Energy Information Administration expects the average residential price to reach about 18.2 cents per kWh for 2026, up from 17.3 cents in 2025 and 16.5 cents in 2024.24 The same agency projects US electricity use will set new records in 2026 and 2027. It names AI-focused data centers and electrification as the drivers, with demand climbing from 4,195 billion kWh in 2025 to 4,288 billion kWh this year.22 Earlier this year the EIA projected that commercial-sector demand, the category that includes data centers, would pass residential demand in 2026 for the first time on record.27

My reading of the coverage is this: data centers are a real and measurable cause of higher bills. The effect is concentrated, though. It shows up most clearly in particular grids and particular utilities, and it is just one of several forces pushing prices up. Calling bills nationally "exploding" because of AI goes further than the evidence. Brushing data centers off as a scapegoat goes too far in the other direction.

PJM is where the cost is clearest

The strongest evidence comes from PJM Interconnection, the grid operator for 13 states and Washington, D.C., including Northern Virginia's data center corridor. PJM's capacity auction pays power plants to be available years ahead. It cleared at $28.92 per megawatt-day for 2024/25, jumped to $269.92 for 2025/26, and then hit the cap at $329.17 for 2026/27.1 The 2027/28 auction cleared at $333.44. The 2028/29 auction, held in July 2026, cleared at the price cap for the third auction in a row.8 That latest auction also fell about 6.8 gigawatts short of PJM's reliability requirement.5

PJM's independent market monitor, Monitoring Analytics, has repeatedly linked these prices to data centers. It estimated that data center demand caused 63% of the price increase in the 2025/26 auction, about $9.3 billion that customers recover through their rates.1 In the December 2025 auction, data center load made up $6.5 billion of $16.4 billion in costs. Across three auctions, forecast data center load above existing levels accounted for $21.3 billion.3 The monitor's president now puts the total added capacity cost at $29.4 billion over the last four auctions.9 NPR reported a similar figure of about $29 billion spread across PJM's 67 million ratepayers.13

The effect reaches wholesale energy prices as well as capacity payments. PJM's average wholesale price rose 47% year over year in the first seven months of 2026, to nearly $117 per megawatt-hour. The monitor estimated prices would have been about $10 lower without data center demand.20 For households, analysts estimated capacity costs alone would add about $18 a month to average bills in western Maryland and about $16 in Ohio.1 One estimate cited by NPR says Maryland customers pay roughly $168 to $216 more per year mainly because of the data center boom.13 State-level EIA data line up with these findings. In July, Ohio and Maryland posted some of the largest year-over-year price increases in the country, at 15.2% and 14.8%.11

The auction caps are hiding part of the problem. The market monitor calculated that the December auction would have cost $9.9 billion more without the cap negotiated with Pennsylvania's governor.3 PJM's board has now proposed a backstop auction with a cap of $555 per megawatt-day, well above the $325 cap in its last base auction.9 The Department of Energy has also given PJM emergency authority to cut power to data centers and other large customers during grid emergencies.8 These are not the moves of a grid operator that expects the strain to ease.

The research disagrees

Outside PJM, the evidence is mixed. A September study by Cameron Scalera, Valentina Bosetti and Filippo Pecci covered about 1,200 utility territories and 2,700 data centers. It found that an average-sized hyperscale facility raised residential prices about 4%, and that households saw larger increases than business customers.15 Scalera said that around ten data centers are associated with an average price increase of 11% to 13%.15 The study's most telling result is that in territories served by nonprofit utilities, the effect was close to zero.1512

A 2026 study from the Electric Power Research Institute reached the opposite conclusion for 2015 through 2024. It found that data center growth slightly lowered average retail rates, because fixed grid costs were spread across more kilowatt-hours sold.15 Republican Senate candidate Alan Armstrong has seized on that kind of finding to argue that data centers are being made a scapegoat for an underbuilt grid.15

The two findings can both hold. Load growth can lower unit costs when new customers pay for what they use, and raise them when they don't.11 Spreading fixed costs worked when data center growth was modest and the grid had spare capacity. Scalera's team found smaller price increases where surplus capacity existed.15 That spare capacity has now run out in markets like PJM. In those markets new load forces scarcity pricing and new construction, not cheaper use of the wires already in place. The result for nonprofit utilities points to the same thing: how a utility is structured and how it divides costs among customers matter as much as how many megawatts arrive.

Other causes of higher bills

Not every price increase traces back to servers. The EIA expects wholesale prices to average $52 per megawatt-hour in 2026, 11% above 2025. It attributes that mainly to winter storms and hot summers, not data centers.22 Gas delivered to power plants cost 36% more in 2025. Utilities spent $78.6 billion on transmission and distribution in 2023, and that spending reaches rates with a lag.11 In the Midwest, advocates point to aging infrastructure damaged by storms as a major cost, alongside new data center load.16 The states with the largest price increases also undercut a single-cause explanation. Hawaii led July's increases at 25.4%. Maine, California and New York had some of the largest gains since 2019, and none of them is the center of the AI buildout.1120

Texas complicates things further. It hosts the most large AI sites in one ranking, yet business electricity prices there were almost unchanged.17 The national figures also contain a contradiction. The EIA projected summer wholesale prices below last year's because natural gas was cheaper, while residential retail prices kept rising. Transmission and capacity charges did the work.23 So bills can rise even when energy itself gets cheaper, and that is the route through which data center infrastructure costs reach households.

How big the growth gets

No one disputes the scale of the growth. Lawrence Berkeley National Laboratory and the Brattle Group estimate that US data center demand more than doubled from 2018 to 2024 and could roughly triple again by 2028.15 The IEA expects data centers to account for about half of US electricity demand growth through 2030.25 Global data center consumption rose 17% in 2025, and AI-focused facilities grew about 50%.28 PJM expects 5 to 7 GW of new data center demand each year but only 2 to 3 GW of new generation.7

The forecasts are uncertain, though. Lawrence Berkeley's projection for US data center use in 2030 ranges from 521 to 843 TWh. In ERCOT, observed data center load has come in at roughly half of what developers requested.11 Much of PJM's capacity cost comes from facilities that are forecast but not yet running. About $6.2 billion of the December auction's data center costs were tied to sites expected online by 2027/28.6 That creates a risk ratepayers should take seriously. If speculative projects never get built, households will have paid for capacity no one ended up needing. RMI, a clean energy think tank, warns that overbuilding gas plants could add up to $118 a year to the average household bill.20

Political pressure is building

The backlash is already changing policy. In September the House voted 417-3 for a bill that would require state regulators to consider making data centers pay for the grid upgrades built to serve them.19 Virginia began taxing data center electricity use at $0.011 per kWh on July 1. Ohio and Illinois paused their tax breaks. Texas ordered an audit of grid connections, with ERCOT facing about 474 GW of connection requests, roughly 90% of them from data centers.19 Lawmakers filed more than 300 data center bills in over 30 states this year.12 Local opposition stalled at least 120 projects worth nearly $200 billion in the first half of 2026.20 Seventy percent of Americans say they worry data centers will raise their bills.13

The central fight is now over who pays. The physics of demand growth is settled. Utilities, regulators and Congress are moving toward separate rate classes and cost-allocation rules for large loads. Whether those rules take effect before the next round of capacity auctions will decide how much of the AI buildout ends up on household bills.