Do Data Centers Raise Your Electric Bill?

A strange week in Washington

On September 29, President Trump met with technology executives and told them data centers "are going to be very popular." The next day, Senate Democrats blocked legislation from Sen. Jon Husted of Ohio, a Republican in a close race against the Democrat Sherrod Brown. The measure would have required states and some utilities to consider a standard under which data centers pay the extra costs their power demand creates. That same day, Rep. Rich McCormick, a Georgia Republican, told viewers of the conservative network Newsmax what a data center would do for them: "they're going to keep your electricity rate low, they're going to pay for your teachers your schools your roads, most likely your property taxes, you're gonna have more jobs."

In an October 1 article in The Hill, Sudiksha Kochi reported that Republicans were nervous. An August memo from the National Republican Senatorial Committee had warned that "data centers are the anchor hanging around Husted's neck." Democrats had seized on the issue. In Ohio, Brown has highlighted the data-center tax breaks Husted supported as lieutenant governor and suggested he was "the point man to bring these data centers to Ohio." Amy Koch, a Republican strategist, said: "It's hard to be nuanced with a president who has no nuance." Whatever the president predicts about everyone liking data centers, she said, "no one's going to like data centers," but "there … is a need for data centers, and let's do it in a way that's smart."

The usual party lines do not hold. In June, Zoya Teirstein and Kate Yoder of Grist, a nonprofit environmental news outlet, reported that Texas Gov. Greg Abbott, a Trump ally, had urged his legislature to rein in data-center development. They also reported that New York Gov. Kathy Hochul, a Democrat, would not say whether she would sign her state's one-year moratorium. In August, Sophie Hurwitz of the progressive magazine Mother Jones quoted remarks that Texas's Republican agriculture commissioner, Sid Miller, made to the Texas Tribune: "Republicans should be on this issue and we're not." He went on: "The Democrats are, and they're right on the issue."

The polling explains the nerves. Grist reported Gallup's finding that about 70 percent of Americans oppose a data center being built near them, including 63 percent of Republicans. Other survey data in the same report found conservative Republicans more likely to object than moderate ones, 53 percent against 44. "I'm not sure I've ever seen a chart where conservative Republicans are closer to liberal Democrats than liberal [and] moderate Republicans are," Anthony Leiserowitz, who directs the Yale Program on Climate Change Communication, told Grist. The opposition is also winning: Grist counted at least 75 projects, worth roughly $130 billion, stalled or blocked in the first three months of the year.

Everyone is sure about the electric bill

The two loudest camps split the electric-bill question between them. One says data centers lower electric bills, the other says they raise them, and each treats its answer as settled.

McCormick's promise, "they're going to keep your electricity rate low," is the builders' answer. The president's version was larger. Sean Moran of the Daily Caller, a partisan outlet on the right, quoted his Truth Social post from late August: "If they want to be successful and rich, with far lower taxes and jobs all over the place, let Data Reign."

Rep. Alexandria Ocasio-Cortez, a New York Democrat, gave the opposite answer in March. It came in a press release from Sen. Bernie Sanders's office that announced their legislation for a national moratorium on new data centers. Among the harms of unregulated AI, she said, "We have seen ... data center construction inflate electric bills in communities across the country."

Measured against the buffet, McCormick's promise survives in part. He began by telling viewers "if you don't want a data center, that's fine, don't have one," and a data center can keep rates low in a place like North Dakota. What does not survive is the same promise made to any neighborhood, without conditions. Ocasio-Cortez's statement also survives in part: rising electric bills in data-center regions are real, and "communities across the country" can be read as many places rather than everywhere. But the sentence invites a reader to hear a national rule where the evidence shows a regional one. Each speaker took half of a mixed case and presented it as the whole, as most people who answer the electric-bill question do. Hardly Thinking calls that habit G12: Fixed Pie or Growing Pie, and it runs through the rest of this fight.

Each camp also has a place it points to. For the builders it is Loudoun County, Virginia. In June, Shirleen Guerra reported for The Center Square, a free-market news service, that the county's data centers "generate 38% of Loudoun's General Fund revenue and nearly half of its property tax revenue." Her report covered an event held by the advocacy group National Taxpayers Union. For the opponents it is a porch. A June op-ed in Civil Eats by the environmental-justice advocate Anthony Nicome opens on a Virginia resident who hears "the relentless hum of a newly built AI data center beside his house," where farmland and a community garden were "converted over the last few decades into concrete and server racks." Before any evidence arrives, that opening has made the reader a neighbor rather than a planner.

Some of this fight is not about electricity at all, and people on opposite sides say so. In August, Ginny Hogan wrote in the labor-left newsletter Labor Politics that "construction isn't the root of the problem. The root of the problem is AI." The Information Technology and Innovation Foundation, a technology-industry think tank, made the same diagnosis from the other side in July: "water is a convenient hook on which to hang more general opposition to data centers (and AI)." At least three quarrels travel under one name: who pays for new power, whether AI should be slowed at all, and whether a town or the nation gets to decide. The second is a separate argument, and a legitimate one. The first and the third are where the reasoning slips.

Two numbers from one report

The facts about electricity costs can be chosen to sound small or large. In August, Ethan Howland of Utility Dive, a trade publication, reported the latest findings of the independent monitor of PJM. PJM is the grid that serves the mid-Atlantic and parts of the Midwest. Two figures in the monitor's findings often travel separately.

From January through July of this year, data-center demand made up 9 percent of the wholesale price of power on that grid. It did so through the payments that keep power plants available, and the figure does not count any effect on the price of the electricity itself or on transmission. Over the grid's last four auctions for those payments, data-center demand added $29.4 billion.

Both figures are true. They measure different things over different periods. A writer who wants the effect to sound modest cites the first; one who wants it to sound enormous cites the second. Neither is a household electric bill, though such costs reach households. Caleb McCullough of PolitiFact concluded in June that they have "been passed on to consumers, especially in areas with high data center activity." Amazon's cloud chief, Matt Garman, pointed to the same regional split in a post that argued data centers are not to blame for rising rates. Todd Bishop of GeekWire summarized the post: rates "have risen in some" states with many data centers, "and fallen in others." The useful question about any such figure is what it measures, and for which place.

Three pies in one post

The president's August post did not stop at "let Data Reign." It continued: "The good news is that there are plenty of other places that want them. If we kill the Golden Goose, you will only have yourselves to blame. China could not be happier with this anti Data Center movement."

The post describes three different pies. For the town, the pie grows for everyone: lower taxes, jobs all over the place. Between towns, the pie is fixed: whatever one town refuses goes to "plenty of other places." Between nations, it is fixed again, and a refusal is China's gain. The last two claims pull against each other. If a refused project moves to another American town, the refusal is one town's loss and its neighbor's gain, not China's. None of the three pies is the electric bill. The post promises lower taxes and jobs and says nothing about who pays for the new plants and lines. That is the question a town has to settle.

A charitable reading comes from the builders' own side. A July briefing at political.org, a site that argues for building, cites industry tracking of "more than 140 local groups having blocked or delayed over $60 billion in U.S. data center projects in just over a year." If refusals pile up faster than willing places can absorb them, each one moves a project, but together they slow the country. The same briefing also warns that bans "may simply redirect development rather than curtail it," which is the post's "plenty of other places" again. The two claims cannot both carry full weight. Which of them does depends on a question neither document asks: Is the national build-out held back by a shortage of willing places, or by something else, such as chips, power plants, transmission lines or money?

Some Republicans say the China argument does not answer the local question. "At times, Republicans have relied exclusively on America's competition with China as an explanation for why data center construction is important," Mark Bednar, who once ran communications for Kevin McCarthy when McCarthy was House speaker, told Sean Moran of the Daily Caller. "While that may be true and central to policy-making and long-term planning, it's been an insufficient talking point to voters."

The county everyone borrows

Loudoun County's numbers are real. Shirleen Guerra's report for The Center Square also says that one speaker described "permanent employment at individual facilities" as "relatively modest," and that the speakers said the benefits should be weighed "alongside concerns about electricity demand, water use and land use." But the report's evidence came mainly from one county, the one with the most data centers on Earth. No second county was offered to show what a typical county might expect.

Loudoun's own budget page gives a fuller picture. It describes the county as holding "the largest concentration of data centers in the world." It says the property-tax rate has fallen, but that "homeowners' individual tax bills have varied due to increased property values." Even in the strongest case, a lower tax rate did not mean a lower tax bill for every homeowner. The page also warns that data-center revenue "will plateau in the next five to 10 years."

The opponents reach for Loudoun too. Anthony Nicome's Civil Eats op-ed sets its porch there. The "last few decades" of change it describes began well before the AI boom, so the scene carries Loudoun's long data-center history into the case against this year's projects. The two uses differ. The tax argument offers Loudoun as evidence for a general rule; the op-ed uses it as a setting. But both sides lean on the county with more data centers than anywhere on Earth. Its vividness is what makes it a poor guide for a rural county weighing its first data center.

History, picked for its ending

Writers choose their history the same way. In September, the historian Arthur Herman reached for fracking in the opinion pages of the right-leaning Fox News. He wrote that "we've been here before," and that "in the end, the fearmongers lost, and the energy companies won, as did America's strategic and economic interests." The same month, Chris Lehmann reached for the 19th-century Populists and their fight against the railroads and Standard Oil. He wrote in the left-leaning Nation: "When you squint a bit at the data-center uprising, its economic message is of a piece with the Populists' crusade to rein in the brutal takeover of the American countryside by the twinned forces of the railroad and oil industries."

Comparing a new fight with an old one is a reasonable way to think. The difficulty is that each precedent arrives with its verdict attached. For Herman, the opponents were fearmongers who lost; for Lehmann, they were crusaders against monopoly. Lehmann's comparison settles the electric-bill question inside a single clause: "Just as today's AI lords are making water scarce and electricity more expensive."

The fracking precedent Herman chose can be read other ways. Teirstein and Yoder's June report for Grist holds two more readings of the fracking fight. In its own voice, Grist wrote that "Fracking poses serious risks to local water supplies," which disputes Herman's "fearmongers." Alex Beauchamp, a regional director of the environmental advocacy group Food & Water Watch, drew a lesson about politics rather than water: "This feels to me like the early days of the fracking fight. A lot of Republicans were really up in arms about it in the beginning, and then it slowly became a partisan issue."

Ginny Hogan, in her Labor Politics essay, draws a historical comparison and marks its limit. She reports a law professor calling coal "a very tight analogy," and then writes: "But there's only so far historical analogies can go."

Nothing that would change their minds

At the two loudest ends of this fight, neither side says what would change its mind. Food & Water Watch made its case for a halt on new data centers in a March 4 press release, which quotes the group's executive director, Wenonah Hauter: "It has yet to be determined if – not how – the industry can ever operate in a manner that sufficiently protects people and society from the profusion of inherent hazards and harms that data centers bring wherever they appear." The president wrote in mid-September, according to Anna Helhoski of NerdWallet: "There is a SICK conspiracy going on against AI and Data Centers, and the only one that is happy about it is China. WHOEVER WINS AI, WINS!"

Each position has a reasonable core. A new industry can fairly be asked to show it is safe, and being first in a technology race can matter. But "inherent hazards and harms ... wherever they appear" makes harm a property of the thing itself, so no design, site or cost rule could answer it. "WHOEVER WINS AI, WINS!" makes the race winner-take-all, so no local cost could count against building. Neither statement names any evidence that could change it.

The middle of the fight is different, and larger. The Food & Water Watch release also quotes Anna Kelles, the Assembly sponsor of New York's moratorium legislation, who called it "a responsible pause, not to stop innovation." According to Helhoski, a New York Times/Siena poll found that "among those who oppose data centers, some 56% said they support limits as opposed to a total ban." For those people the disagreement comes back to the buffet: would they accept a data center that demonstrably pays its own grid and water costs, in a place where the grid and the watershed can carry it? That question can be answered for a particular town. "WHOEVER WINS AI, WINS!" cannot be answered at all.

The Pattern Underneath: fixed pie or growing pie

One pattern runs through all of these examples. Economists distinguish a fixed pie, where one party's gain is another's loss, from a growing pie, where both can come out ahead. Most real situations are a mix of the two, and the common mistake is to argue a mixed case as if it were purely one or the other. Hardly Thinking calls the pattern G12: Fixed Pie or Growing Pie. The mistake runs in two directions.

The fixed-pie failure treats every gain as somebody's loss. It appears among the opponents. The March press release from Sen. Bernie Sanders's office describes AI as "a global race to see which country is the first to eliminate hundreds of millions of jobs." That reads as if there were only so much work to go around. It may instead be a forecast that AI will destroy jobs faster than new ones appear; the release does not say which. Chris Lehmann's clause in The Nation treats scarcer water and dearer power as given.

The growing-pie failure treats every situation as one where everyone wins, and waves away the people who do lose. It appears among the builders: in Rep. Rich McCormick's promise of low rates to every neighborhood, in the president's "jobs all over the place," and in Loudoun, the extreme case, offered with no second county beside it.

The two poles, fixed and growing, are not camps, and the same writer can stand at both. Arthur Herman, in his Fox News column, writes that "the only real beneficiary of a moratorium will be China": a race with one prize, in which every delay here is a gain there. On the neighbors' water, the same column turns the other way: "Anti-data center fanatics spread similar falsehoods about data centers using up water, when in fact your average golf course consumes more." On China he is a fixed-pie thinker; on the local cost, a pie-grower. Anthony Nicome's Civil Eats op-ed does the reverse. Its next-to-last sentence reads, "We do not have to choose between innovation and nourishment, technology and equity, or progress and public health," as if the limits it proposes cost no one anything. Its last sentence makes the choice: "We only have to choose to plan wisely and listen to the communities who have always understood that the land beneath their feet, the water they drink, and the food they grow are worth more than anything a data center can offer." One sentence says no one has to lose; the next says nothing a data center offers could weigh against what the community would lose. A writer can stand at both poles on two parts of the same fight. A reader can do the same.

The data-center fight adds a third pie that most fights of this kind lack: the race between nations. Only one country can be first, so a lead in a race is fixed. On that point the fixed-pie instinct is correct. The mistake is letting that one fixed pie answer every local question, as "China could not be happier" does.

Two findings fall mostly on one side or the other. The claim that China is behind the opposition comes almost entirely from the build side, and the evidence against its strong form is good. In September, Loreben Tuquero of PolitiFact reported that the Chinese influence campaigns tech companies had found on social media had little reach. The experts she interviewed called the claim exaggerated. "Evidence of an attempt is not evidence of success," said Jessica Brandt, a senior fellow at the Council on Foreign Relations. (Blaming an unseen enemy for a movement is a pattern of its own, which Hardly Thinking calls G6: The Hidden Hand.) In the other direction, the clearest case of a cost figure that changes what it measures came from an opponent. Sen. Elizabeth Warren, a Massachusetts Democrat, wrote on X that near large data centers "your electricity bills" had risen "as much as 267%." In Caleb McCullough's June fact-check, PolitiFact rated the claim Mostly False: the number described wholesale prices at particular points on the grid, which make up only part of a household electric bill.

The tools, used and abused

The tools on HT are questions a reader can run on any argument, including their own. Each has a fair use and a weaponized one. Two of them do most of the work in this fight.

B4: The Reference-Class Question asks: what is this a case of? Every judgment about something new borrows a comparison, and the comparison often decides the verdict. Loudoun County, fracking and the railroads are all reference classes, and in each case the comparison arrived with the ending the argument needed. Used fairly, the tool names two or three plausible comparisons and notes how much the verdict shifts between them. Ginny Hogan used it fairly in Labor Politics: she reported a comparison with coal and then said where it stops fitting. Used as a weapon, it becomes class-shopping: picking the comparison that flatters a conclusion and calling it obvious.

F1: The Crux Finder asks: what is the one question this disagreement turns on? Most fights are about conclusions, while the real disagreement sits in a fact or a priority underneath. On the electric bill, the crux is whether the grid has room and who pays for what it lacks. In the president's post it is whether willing places are scarce. At the two loudest ends of the fight there is none. None of the people quoted here uses the Crux Finder as a weapon, which would mean demanding that the other side say what would change its mind while exempting one's own position from the same test.

A reader who noticed the missing crux only at the end they already oppose has used the Crux Finder as a weapon. The fair use starts at home. Take your own answer to the electric-bill question, or the claim about data centers you most agreed with, and ask: Which part of this is a fixed pie and which part could grow, and what fact, if it came out the other way, would change my mind? Running a tool on your own side first is a habit HT calls A4: The Mirror Check.

One Washington Post reader did so in a letter published in June: "While I'd like for America to stay ahead of China in the artificial intelligence race, I'll admit that, in this case, I'm a NIMBY." (NIMBY stands for "not in my backyard.") The national race and the local cost sit in one sentence, and the cost is admitted. Few of the officials and advocates in this fight say as much, and few of the rest of us do either.

Reader review

Every HT article ends by asking readers to check it against its own rules. The question is not what you think of data centers but whether this article met the standards it applies to others.

1. Find the passage where this article was hardest on the side you agree with, and the one where it was softest on the side you oppose. Were they judged by the same standard?
2. The article says the effect of data centers on electric bills depends on the place. Find any sentence where it settles the question one way anyway.
3. The article applies B4, the Reference-Class Question, to three comparisons: Loudoun County, fracking and the railroads. Which comparison does it lean on for its own answer about the electric bill, and does it say how that answer would change under a different one?
4. The article uses F1, the Crux Finder, and says the electric-bill fight turns on whether the grid has room and who pays for what it lacks. It faults the loudest voices for not saying what would change their minds. Does it say what would change its own answer?
5. The article opens with a surprising national average and then qualifies it. Did it give the average more weight than its own caveats allow? That is the fault it finds in others' use of numbers.
6. Several sections set a passage from one camp beside a passage from the other. Pick one pair: are the halves equally strong, or was one brought in to balance the other?
7. Choose one quotation that has a link and read it in its source. Does the article represent its context fairly?
8. Name a position in this debate that the article did not examine, and say whether leaving it out changes the picture.