Do data centers raise your electricity bill? What the evidence shows

What does the evidence from PJM show?

PJM is the grid operator for much of the eastern United States. It runs an annual auction in which power plants are paid to be available years ahead, called the capacity market. Prices in those auctions have risen sharply, and PJM's independent market monitor says data centers are the main reason.

For context, the price for 2024/25 was $28.92, so the 2025/26 price was about nine times higher. The price has sat at or near a cap agreed with governors and regulators since the 2026/27 auction. Across the four auctions, the monitor attributes $29.4 billion, which is 46.2% of the $63.6 billion in capacity charges.

Four points change how to read this.

• It is a model, not a measurement. The monitor reruns each auction with a load forecast that leaves out existing and forecast data center load, and reports the difference.
• The monitor has a position. It says data center load should be taken out of the capacity market and bought through its own auction.
• PJM names several causes. For 2026/27, PJM said its forecast peak load rose by more than 5,400 megawatts, driven largely by data center growth, electrification and economic growth.
• Different percentages are not a conflict. About 40% is the data center share of total charges. A figure such as 65.5% is how much higher charges were than in the no-data-center case.

The monitor also says not all of the increase in capacity costs came from data centers. For the first half of 2026, the wholesale cost of power in PJM averaged $114.50 per megawatt-hour, against $76.16 a year earlier. Energy was up 47.7% and capacity up 207.1%. By the monitor's calculation, data center effects on capacity added $11.11 to the first-half price, and most of the rest of the rise was in the energy component.

These are wholesale charges across a region. They are not household bills. PJM said the 2026/27 price, which came in at the cap, would translate to a year-over-year increase of 1.5% to 5% in some customers' bills, depending on how utilities and states pass on wholesale costs. It also said wholesale capacity is a relatively small part of retail bills. In later releases it said load that is hedged through self-supply or bilateral contracts is not exposed to the clearing prices. It did not split the bill range by cause.

What do national studies say about prices?

A widely read national study is the Lawrence Berkeley National Laboratory and Brattle's 2026 update, funded by the Department of Energy. Its authors say it is not definitive, covers prices and not bills, and does not predict the future.

• Prices. All-sector average retail prices rose 5.3% in 2025 before inflation, or 2.6% after inflation. Residential prices rose 5.0% before inflation, and were 33% higher than in 2019 before inflation. After inflation, all-sector prices were 3% higher than in 2019.
• Load growth. State-level load growth was associated with falling all-sector average prices after inflation in recent decades, including 2019 to 2025 in most states. An earlier statistical analysis put the fall at more than a cent per kilowatt-hour in the highest-growth states. The study says the presence of significant data center load does not appear to change that conclusion.
• Residential customers. The link between load growth and residential prices is less clear than for the all-sector average.
• The mechanism. Fixed costs spread over more sales lower the cost per unit. The study also says there are conditions under which growth raises prices.
• PJM is the study's example of the opposite. Its PJM case says accelerated load growth, combined with supply and delivery limits, led to significant price increases, and that capacity prices are one but not the only contributor.

It names other drivers of the 2025 increase: natural gas delivered to power plants was 36% above its 2024 low, a driver in 21 states, and distribution costs were a driver in 22. Generation capital spending rose 22%, transmission costs rose in New England, and storm recovery added costs. Seven PJM jurisdictions had large retail price increases from 2024 to 2025, from 0.8 cents per kilowatt-hour in Ohio to 3.1 in the District of Columbia.

A Nature Energy study published on November 9, 2026 estimated that cumulative price increases attributable to data centers exceeded 10% in the most exposed hubs by 2025, and found that residential customers were allocated a median of 47% of costs. Spezia read only the abstract. The study measures wholesale prices and cost allocation, not household bills.

What do prices look like now?

The Energy Information Administration's national average for residential customers was 16.48 cents per kilowatt-hour in 2024 and 17.23 in 2025. In August 2026 it was 18.16 cents, preliminary, which is +3.4% from August 2025.

The table shows that prices rose in most of these states. It does not show that data centers caused the rise. The sources also differ slightly on 2025: the Energy Information Administration's October 2026 table implies residential prices rose 4.6%, and the Berkeley and Brattle study says 5.0%. Spezia did not find the reason.

Who pays for the grid that data centers need?

This is the contested question. In a March 2025 paper, Harvard's Electricity Law Initiative reviewed nearly 50 regulatory proceedings and said it was skeptical of utility claims that data center energy costs are isolated from other customers' bills. It argued that confidential special contracts keep the public from checking the costs, and that a contract shifts costs when the customer pays less than it costs to serve it. It acknowledged the reverse is possible but thought it unlikely. The Data Center Coalition cited a Virginia legislative audit that found data centers there pay their cost of service, The same audit said data center demand was likely to raise system costs for all customers.

Regulators have responded with large-load tariffs. A Berkeley and Brattle catalogue of 55 tariffs, updated in August 2026, found thresholds from 0.3 to 150 megawatts, with a median of 25. Minimum contract terms run from 1 to 20 years, and the median rose from 5 years for tariffs proposed before 2025 to 12 years for those proposed since. The median minimum bill is 80% of the contracted demand.

The AEP Ohio and Dominion terms are as listed in the Berkeley and Brattle brief. The other rows come from the sources listed below, and Spezia did not confirm the date of each commission order. Georgia's commission says its orders protect existing ratepayers from increased rates and that current ratepayers will not subsidize data centers, and the Georgia utility pledged downward pressure on a typical bill of at least $15 a month for 2029 to 2031. Those are a regulator's and a utility's own claims. The April Berkeley and Brattle price study says there is a risk that the new tariff structures will be insufficient to fully protect other customers.

What have voters and lawmakers done?

Local voters favored limiting data center development in 40 of 41 local ballot contests on November 3, according to Ballotpedia. Some measures were advisory, and they concerned where data centers may be built, not electricity prices. See what the midterms mean for AI rules and data centers and AI laws in force in 2026 for what incoming Democratic leaders say they will pursue on electricity costs and for the state laws on large electricity users.

How do you read a data center bill claim?

Five checks help.

1. Wholesale or retail? A capacity charge is paid by the utilities and suppliers that serve customers, and how much reaches a bill depends on the state, the contract and any self-supply. A bill is what a household pays.
2. A model or a measurement? A counterfactual is a calculation about a world that did not happen.
3. Which region? PJM's experience is not Texas's or Georgia's.
4. Which cause? Gas prices, wires, storms and plant spending move prices too.
5. Who gains from the claim? Utilities, data center groups, regulators and monitors each have a stake.