The Data Center Did Not Raise Your Power Bill. Ironically, More Demand Makes Electricity Cheaper.
Begin with a belief almost everyone holds. When more people want a thing, its price goes up. We learn this lesson early and we learn it well, because it is true of nearly everything we buy. More shoppers bidding for the same tomatoes, and the price of tomatoes climbs. More families competing for the same house, and the sale price rises. The intuition is so dependable that we apply it without a second thought, and most of the time we are right to.
Electricity is the exception, and the exception matters enormously right now. The public has been told that artificial intelligence data centers, those vast humming buildings packed with servers, are the reason household power bills keep climbing. The story is intuitive, which is precisely why it deserves a second look. The best causal evidence we now have says the intuition is not merely incomplete in this case. It is backwards.
To see why, picture a toll bridge. The bridge cost a fixed sum to build, and it costs roughly the same to maintain whether 100 cars cross it each day or 100,000 do. If only a handful of drivers use it, each one must pay a steep toll to cover the bridge. If thousands use it, the same fixed cost is spread across far more drivers, and the toll per car falls. The cost of the bridge did not change. The number of payers did. A power grid works the same way. The wires, the substations, the transmission towers, and the generating plants represent an enormous fixed cost that must be recovered no matter how much electricity flows over them. Your monthly bill is mostly a share of that fixed cost, not a charge for the next kilowatt-hour produced.
This is the hinge of the whole argument, so it is worth slowing down. In an ordinary commodity market, the price you pay tracks the cost of producing one more unit. In electricity, retail rates track average cost, because the system is a regulated network dominated by fixed investment. The gap between the two is large. For the period from 2014 to 2016, the average wholesale price of power in the Northeast ran around 4 cents per kilowatt-hour, while the average retail rate sat above 10 cents and the national figure near 12 cents. Marginal cost, in other words, was roughly half of average cost. When the extra cost of serving new demand is well below the average cost baked into everyone’s bill, adding demand mechanically pulls the average down. A steady, around-the-clock customer is the toll bridge filling up with cars.
Data centers are that customer. They run near continuously, at utilization rates of 80% to 90%, while the American grid as a whole runs at an average load factor of only about 53%, according to a Duke University analysis. The grid is an airliner that flies nearly empty most of the year. Filling those empty seats with steady load is exactly what lowers the cost per passenger.
Now to the evidence, because an elegant theory is worth little if the numbers refuse to cooperate. A
2026 study by Asa Watten, John Bistline, and Geoffrey Blanford
, published on arXiv, set out to measure the causal effect of data center growth on retail prices. The authors used an instrumental-variables design built on the 1947 Eisenhower interstate highway plan, since fiber-optic corridors tend to follow old highways, which lets them isolate data center siting from the prices that might otherwise have attracted developers. Their finding is plain. Data centers caused average retail electricity rates to fall modestly from 2015 to 2024. Every 10% increase in data center capacity lowered average residential prices by about 0.4%. Translated into lived experience, the typical residential customer lived in a state where data center capacity grew 160% over 2019 to 2024, and that growth cut their rates by 6%. The preferred specification implies that a doubling of capacity lowers residential prices by 3.5%. The statistical test the authors rely on, which is robust to weak instruments, rejects any positive price effect at the 0.1% level.
Notice who produced this result. Two of the three authors work at the Electric Power Research Institute, the electricity industry’s own nonprofit research arm, not an advocacy shop with a thumb on the scale. And they are not alone. Charles River Associates concluded in February 2026 that the data center buildout did not trigger retail rate increases over the past decade. Researchers at Lawrence Berkeley National Laboratory, a federal lab, working with the Brattle Group, found that the states with the greatest load growth saw real retail prices fall over the past five years. When a federal laboratory and the utilities’ own consultants concede the point, the debate is effectively over.
The geography seals it. Virginia is the data center capital of the world, with these facilities consuming more than 20% of the state’s electricity, yet Virginia’s residential rates run roughly 9% below the national average. North Dakota saw electricity demand grow about 35% over five years while its real price fell. If data centers raised bills, these two states should be cautionary tales. They are advertisements.
So why does nearly everyone believe the opposite? Here we reach the heart of the matter, and it is a lesson in human nature as much as in economics. When a bill rises, people look for a cause they can see. Data centers are large, new, and conveniently unsympathetic, so they become the answer to a question they did not cause. The anti-AI activist blames them because he dislikes the technology. The anti-capitalist blames them because he dislikes the firms that own them. The politician blames them because blaming a faceless tech company is easier than explaining his own energy policy. And the utility, most revealingly, blames them because the real story implicates the utility itself.
Consider Georgia Power. Where rates near a project do climb, the cause is frequently a decision the utility made, not a demand the data center imposed. A data center operator would often prefer to build its own generation on its own site, financed with its own capital, drawing nothing from the public grid. That arrangement would add load without adding a penny to anyone else’s bill. Yet the utility has every incentive to block it, because a self-powered data center is a sale the utility loses. Georgia Power would rather supply that power itself, which means stringing transmission lines across miles of countryside to reach the site. Building those lines can require condemning property and tearing down homes that stood in the path. The disruption is real, and it is expensive, and it lands in the rate base. But it is the consequence of the utility’s choice to monopolize generation, not of the data center’s choice to consume electricity. The operator asked to power itself. It was told no.
Texas tells the same story in a different accent. Utilities there are pressing to build new transmission lines out to the Permian Basin, and the reason is straightforward. They want to sell their power into that market, and selling more power requires more wire. To finance that infrastructure they will raise rates, which means consumers pay more in the near term to fund a buildout that may lower costs much later. None of this was demanded by a data center. The data center, once again, would happily generate on site and stay out of everyone’s way. The lobbyists who insist that only the utility may supply the power are the same parties who then point at the data center when the bill arrives. They authored the cost and assigned the blame to someone else.
Strip away the misdirection and the actual drivers of higher bills come into focus. They are the premature retirement of reliable baseload plants, two decades of grid neglect, renewable mandates that raised costs while adding little firm supply, and, in California, wildfire liabilities that pushed average prices up nearly 40% over a few years. The Heritage Foundation and the Manhattan Institute have argued for years that the answer to surging demand is to expand and modernize supply, not to ration it. Energy Secretary Chris Wright has put it more bluntly, calling data centers “the answer to drive down electricity prices,” not the problem.
He is right, and the conservative prescription follows naturally. Do not ban the most valuable new customer the grid has seen in a generation. Require data centers to pay their own way, which the Georgia regulators and the White House Ratepayer Protection Pledge already demand, and then get out of the way and let them build the generation they are asking to build. Abundance lowered the price of power across the whole of the 20th century, as demand exploded and real prices fell decade after decade. It will do so again, if we let it. The villain in this story was invented because the truth was inconvenient to too many people at once. The data center did not raise your bill. The people blaming it did.
If you enjoy my work, please subscribe https://x.com/amuse/creator-subscriptions/subscribe
Alexander Muse is a Fellow at the John Milton Freedom Foundation and publishes daily political analysis at amuseonx.com. Primary sources cited in this piece are linked inline; campaign finance figures are drawn from FEC filings, polling data from publicly released crosstabs, and legal claims from filed pleadings. Corrections are posted to the original URL with a dated changelog. Readers who identify errors are invited to contact the author directly.
That's a new one. More demand doesn't always bring lower prices. A shortage increase prices.
The authors' causal assessment depends heavily on the instrument-validity assumptions. The study examines a historical period that may not resemble the coming AI-driven load boom, and national average effects obscure important regional impacts.