Do Data Centers Raise Your Electricity Bill? The State-by-State AI Power Fight — and the Utility Stocks Caught in It

Do data centers raise your electricity bill? The honest answer is sometimes, partly, and it depends on your grid — and almost certainly not by the number that’s gone viral. The most-shared figure, a 267% jump, is real but widely misunderstood: it measures wholesale power prices at specific points on the grid, not the total bill a household pays. Fact-checkers have rated the “your bill went up 267%” version of the claim Mostly False.

Here is the clear-eyed version — how your bill actually works, what the data does and doesn’t show, the state-by-state fight to make data centers pay their own way, and the utility stocks whose entire AI-power thesis hangs on how regulators rule.

How your electricity bill actually works

Before any scary percentage makes sense, you need to know what’s in the bill. A residential electricity bill is the sum of four parts: generation (the wholesale cost of the power itself), transmission (moving it across high-voltage lines, including grid-capacity charges), distribution (your local utility’s poles, wires and upgrades), and taxes and fees. Crucially, the wholesale generation piece is only about 30-50% of a typical bill (an estimate from Yale economist Kenneth Gillingham, cited in the fact-checks below). That single fact defuses most of the panic: a big percentage move in the wholesale price cannot move your total bill by the same percentage.

Data centers can push on the bill through three channels: raising wholesale prices at nodes near concentrated demand; driving up transmission and capacity costs (PJM’s capacity costs rose 174% for the 2025-26 delivery year, partly on data-center demand); and triggering utility grid spending that lands in distribution charges. But weather, aging infrastructure and ordinary utility capital spending push bills up too — which is exactly why pinning the increase on data centers specifically is contested.

The “267%” number, explained — and corrected

The figure everyone is sharing comes from a Bloomberg analysis published in September 2025, which examined roughly 25,000 pricing nodes across seven regional grids and found that wholesale prices at some nodes near data centers rose as much as 267% between April 2020 and April 2025 — and that more than 70% of the nodes with increases sat within 50 miles of significant data-center activity. The underlying data is sound. The problem is what people do with it.

When a U.S. senator said in June 2026 that residents near data centers had seen electricity bills “go up by as much as 267%,” [PolitiFact]( rated it Mostly False (June 12, 2026), and a WRAL fact-check reached the same verdict days later. The reason is the mechanic above: 267% is a wholesale figure, and wholesale is only a third to a half of a bill. Actual five-year residential increases in the hardest-hit areas were far lower — roughly 94% in Washington, D.C., 74% in Maryland, 73% in Maine and 58% in New York. Real, and painful, but not 267%.

So do data centers actually raise bills? Yes — just less, and not alone

Strip away the exaggeration and a genuine signal remains. PJM’s 174% jump in capacity costs is a concrete, data-center-influenced increase flowing into bills through the capacity component. And a peer-reviewed study led by Jeremiah Johnson of NC State (Environmental Research Letters, May 2026) projects that data-center and crypto demand could raise U.S. electricity costs by a national average of 6-29%, and as much as 57% in some regions, by 2030 — though those are modeled projections, not realized increases. Nationally, retail prices have climbed roughly 40% since 2021, the fastest stretch on record, but analysts caution data centers “shouldn’t get all the blame,” with weather and aging grids doing much of the work. The honest framing: data centers are a driver of higher bills in specific regions, not the driver everywhere.

The state-by-state fight

This is why 2026 became the year states started fighting back. Per the legislative tracker [MultiState](https://www.multistate.us/), more than 300 data-center-related bills have been filed across more than 30 states this session — up from about 200 a year earlier, and a clear shift from courting data centers with incentives to regulating them. The sharpest tool is the moratorium: 14 statewide-moratorium bills across 11 states — New York, Vermont, Maryland, Georgia, Virginia, New Hampshire, Minnesota, Oklahoma, South Dakota, Wisconsin, Michigan and Illinois.

But here is the fact most coverage buries: as of now, zero statewide data-center moratoriums have actually been enacted anywhere in the U.S.

• Maine came closest — its moratorium passed the full legislature, then was vetoed, and the override failed.
• New York passed a one-year moratorium on large (20 MW and up) data centers through both houses in early June (the Assembly 102-39; the Senate 43-17 on the official roll call, often reported as 44-16). As of publication it is awaiting Governor Hochul’s signature — neither signed nor vetoed. If she signs it, New York becomes the first state to enact one.

Everywhere else, the moratorium bills are still proposals. The fight is loud; the law, so far, is not.

Who pays when a data center plugs in? The pledges — and the real lever

Two things are supposed to protect ratepayers. The first is voluntary. At the White House on March 4, 2026, seven companies — Amazon, Google, Meta, Microsoft, OpenAI, Oracle and xAI — signed a “Ratepayer Protection Pledge” to “build, bring, or buy” the new power their data centers need and cover the cost of delivery upgrades. It is a genuine signal, but it is voluntary and non-binding — enforced by reputation and politics, not statute. (Separately, and three weeks earlier on February 11, Anthropic made its own commitment to cover the electricity-price impact of its data centers — a distinct initiative, not part of the White House pledge.)

The second protection, and the one with teeth, is FERC. The fight crystallized when FERC rejected an expansion of Talen Energy’s deal to sell power directly from its Susquehanna nuclear plant to a co-located Amazon data center — utilities argued the arrangement could shift up to $140 million a year in transmission costs onto ordinary regional ratepayers, letting the data center operate as a “free rider.” Then, in a pivotal December 2025 order, FERC found the grid operator’s rules inadequate and directed it to build new service options for co-located load billed on net withdrawals — a framework that effectively unlocks data-center co-location while trying to stop the cost-shift. The catch: FERC directed the operator to write the rules; the final tariff terms — including a possible minimum transmission charge — are still being worked out. The framework is set; the fine print isn’t.

The investor angle: the utility stocks caught in the crossfire

That FERC framework is why a handful of independent power producers trade as AI plays at all. The press framed the December order as a “major victory” for the region’s generators — and [Constellation Energy](https://www.beingshivam.com/quote/CEG/) (CEG) and [Vistra](https://www.beingshivam.com/quote/VST/) (VST) are the most directly exposed: both are positioned to sell power to adjacent data centers under the new net-withdrawal rules, with Constellation’s nuclear fleet the marquee asset. [NRG Energy]( (NRG) is a broader integrated-power-and-retail name with general AI-demand exposure, but it was not among the generators named in coverage of the co-location ruling — its leverage is more indirect.

The scorecard above frames the three on exposure and risk. The shared risk is the same unresolved question: the still-open minimum-charge and behind-the-meter cost-allocation rules will decide how much cost co-located data centers can avoid shifting onto ratepayers — which is simultaneously the ratepayer’s protection and the cap on the generators’ upside. For the bullish supply-side version of this trade, see TECHi’s AI data centers as a power-stock story; for the equipment angle, GE Vernova and Vertiv. You can compare the full AI-power group on the TECHi stock screener.

Will the backlash stick?

There is a real chance this is more than a news cycle. Gallup found in March 2026 that about seven in ten Americans oppose a data center being built in their area — and, unusually, the opposition crosses party lines (75% of Democrats and 63% of Republicans). A Consumer Reports survey late last year found 78% of Americans worried that data centers will raise their energy bills. Affordability is a midterm issue, and “your power bill” is the kind of kitchen-table grievance that turns a technical regulatory fight into a political one. That does not mean moratoriums become law — most won’t — but it does mean the cost-allocation rules will be written under political pressure, which tends to favor ratepayers over unlimited build-out.

How to check your own situation

If you want to know whether this touches your bill, three practical steps: look up whether your utility has a pending rate case at your state’s public utility commission (that is where data-center cost allocation gets decided); check whether your state is one of the 11 with a moratorium or “large-load tariff” bill in play; and read your bill’s supply vs. delivery split — the supply line is the part most exposed to wholesale moves, and it is usually well under half the total.

The bottom line

Data centers are a real but partial driver of higher electricity prices — concentrated in specific grids, flowing mostly through wholesale and capacity costs, and amplified online into a “267%” scare the math doesn’t support for household bills. The genuine story is the fight over who pays for the AI build-out: a voluntary corporate pledge, a louder state-legislature pushback that has yet to actually enact a moratorium, and a FERC rule-making that quietly matters more than either. For investors, that rule-making is the whole ballgame for the co-location trade — the same rules that protect your bill cap the generators’ upside. Watch the FERC tariff fine print and your state’s next rate case; that is where this is decided, not in the viral chart.