On the menu today: If it were up to me, the national dialogue would focus on real and indisputable evils — the regimes of Iran, Russia, China, and North Korea; terrorist groups at home and abroad, along with those who traffic in drugs, sex, and human beings, both domestically and internationally; and other indisputable menaces to all we hold dear. But there’s a consistent and surprisingly effective effort to convince you that the biggest threat to your community is the plans for a new AI data center on the other side of town. Read on.
Democrats’ Data Center Obsession
Back in 2024, I observed that when some of America’s biggest tech companies realized that they needed significantly more electrical power to run their data centers in the decades to come, they decided that restarting decommissioned nuclear plants was the best, most cost-effective, and most reliable option. And with the seeming snap of their fingers, a slew of those closed nuclear plants were scheduled to start operating again in the coming years.
And it wasn’t just Republican governors like Glenn Youngkin of Virginia eager to re-embrace nuclear power; Democrats like Pennsylvania Governor Josh Shapiro, Washington Governor Jay Inslee, Michigan Governor Gretchen Whitmer, and Virginia Senators Mark Warner and Tim Kaine all jumped on board. It was a case of the right policy finally being enacted after decades of foot-dragging and fearmongering, but more than a little frustrating that years of conservatives winning the policy argument and being right on the facts didn’t move the needle on the issue; it was Microsoft, Amazon, Google, and other big companies simply saying, “We want this.”
We should have known that eventually the progressive wing of the Democratic Party would wake up and galvanize opposition; now an increasingly loud swath of Americans, mostly on the left, seem to hate data centers the way they used to hate your SUV, your Big Mac, and, well, you.
Vermont Senator Bernie Sanders and New York Representative Alexandria Ocasio-Cortez want a federal ban on the construction of new data centers, because “data centers will deepen our reliance on fossil fuels during a climate crisis.” (Note that nuclear power is not a fossil fuel and does not produce carbon dioxide.)
AOC, fresh off botching a yes-or-no question on whether the U.S. should commit U.S. troops to defend Taiwan if China were to invade, said in a written statement, “Congress has a moral obligation to stand with the American people and stop the expansion of these data centers until we have a framework to adequately address the existential harm AI poses to our society. We must choose humanity over profit.”
The environmental concerns about AI data centers are varied. In The Dalles, Ore., “residents of the riverfront community are wondering if Google is also changing their weather”:
Steam venting from cooling towers on Google’s local data centers sometimes collects along the Columbia River on chilly mornings. Residents have become accustomed to the phenomenon, which is often visible to drivers along Interstate 84 as they pass through The Dalles.
The steam may be a bigger issue for pilots flying in and out of the Columbia Gorge Regional Airport, just across the river from The Dalles. Some blame the “Google cloud” for dense concentrations of fog that sometimes prevent aircraft from taking off or landing, occasionally forcing flights to divert to Portland — 80 miles to the west.
In Loudon County, Va. — nicknamed the “Data Center Capital of the World” — locals complain about a constant hum that they compare to a hovering helicopter. The Atlantic writes that AI data centers are part of an emerging “dirty, dystopian world.”
There also is widespread concern about the water use of data centers; a widely cited anecdote is that writing a single email using artificial intelligence uses a whole bottle of water. That is not the case; the popular understanding of data centers’ water use is apparently the result of a widely cited environmental activist’s miscalculation, one that writer and researcher Andy Masley is on a crusade to dispel:
Using Claude, some back-of-the-napkin math, a guiding desire for things to make sense, and his background as a high school physics teacher (“explaining how much a watt hour is turns out to be really useful,” he told us), Andy identified a concerning pattern: the country’s top papers, from Bloomberg to The Washington Post, were stoking fears with almost certainly wrong or woefully misleading statistics about data centers’ water use.
So he wrote about it — at length. In these dispatches, which are definitely not designed for the modern attention span, Andy defines terms, aggregates data in charts, and walks people through his math and general thought process for tens of pages at a time. He didn’t expect the essays to blow up. Mainly, he just wanted the writing on-hand when people got mad at him at parties. But maybe because the message is so contrarian — data centers are not taking all the water, they’re not projected to, and sometimes they’re actually improving water access — his Substack has gone gangbusters (he’s up to 5,200 subscribers from roughly 0). Just in the last few weeks, he’s exposed major errors in a leading book on the subject, earned the nickname “AI water guy” in the Pirate Wires Slack channel, made his case on The New York Times’ tech podcast (a lesser accomplishment, to be sure), and pretty much single-handedly started a national conversation on AI water doomerism.
A recent paper claimed that in “the places where data centers are built, land surface temperatures increase by an average of 2°C (3.6°F) after the data center begins operations.” Masley is not only unconvinced; he writes, “It is the single worst writing and research on AI and the environment that I have read. . . . The methodology is completely off-the-wall goofy and the study does not at all show what the authors claim . . . it’s almost certain that it’s instead entirely due to the fact that the surface of any building, a data center or Wal Mart or house or Starbucks, is hotter than the grass that was there before.”
Despite Masley’s efforts, the left’s contention that AI data centers are an environmental menace and the root of evil in the modern world is gaining traction in public opinion.
A recent Washington Post poll found the “share of Virginian voters who would be comfortable with construction of a new data center in their community has plunged to 35 percent, according to the Post-Schar School poll conducted late last month. . . . The poll shows a striking drop since the same question was asked in 2023, when 69 percent of voters said they would be comfortable with a new data center in their community. That shift suggests Virginians now consider data centers almost as undesirable as nuclear power plants, which a 2023 Post-University of Maryland poll found just 33 percent of voters nationwide would be very or somewhat comfortable seeing built in their community.”
And up in Maine, the state legislature has flat-out banned the construction of new data centers, at least for the next year and a half.
When a leftist feels sufficiently strongly about an issue, violence and threats are likely to follow. Somebody like Noah Rothman could write a whole book on the topic — hey, he just did, and it’s coming out in May! — but just a few days ago, Noah noted that opposition to AI and data centers had become the latest excuse of unhinged, violent lefties to run around trying to kill people:
The attempted firebombing of OpenAI CEO Sam Altman’s San Francisco home should be a wake-up call. Had the attacker, who also reportedly made threats against OpenAI’s offices, hit a window rather than the side of the house, against which his Molotov cocktail merely “bounced” off, the event could have been far more horrific.
Just a couple of days later, two men were alleged to have fired on Altman’s house, which should have set the conditions for a national conversation about the horrifying prospect of revivified, violent Luddism.
But it’s not only leftists, progressives, and Democrats who paint AI data centers as a villain. In his most recent State of the Union address, President Trump backed the narrative that AI data centers were making your electricity more expensive:
Many Americans are also concerned that energy demand from AI data centers could unfairly drive up their electric utility bills. Tonight I’m pleased to announce that I have negotiated the new ratepayer protection pledge. You know what that is. We’re telling the major tech companies that they have the obligation to provide for their own power needs.
They can build their own power plants as part of their factory, so that no one’s prices will go up, and in many cases prices of electricity will go down for the community, and very substantially down. This is a unique strategy never used in this country before.
We have an old grid. It could never handle the kind of numbers, the amount of electricity that’s needed. So I’m telling them they can build their own plant. They’re going to produce their own electricity. It will ensure the company’s ability to get electricity while at the same time lowering prices of electricity for you, and could be very substantial, for all of your cities and towns. You are going to see good things happen over the next number of years.
Are AI data centers driving up electricity rates? A September 2025 analysis by Bloomberg News argued that they were, dramatically:
A Bloomberg News analysis of wholesale electricity prices for tens of thousands of locations across the country reveals the effects of the AI boom on the power market with unprecedented granularity. The locations and prices were tracked and aggregated monthly by Grid Status, an energy data analytics platform. Bloomberg analyzed this data in relation to data center locations, from DC Byte, and found that electricity now costs as much as 267% more for a single month than it did five years ago in areas located near significant data center activity.
But Marc Oestreich, editor of the daily energy newsletter GridBrief, writes at Reason that the conventional wisdom about data centers gets everything backwards — that by creating a predictable and steady demand, data centers had become the ideal customer compared to households, and were effectively subsidizing the expansion expenses for power companies.
Blaming data centers for rising electricity prices is like blaming FedEx for the cost of gasoline. Demand didn’t fail. Supply was boxed in.
Electricity is a capital-heavy business. Most of what consumers pay for isn’t the power itself, but the infrastructure that produces and delivers it: generation, transmission, substations, and distribution. Once that infrastructure exists, the marginal cost of serving additional load is relatively low. What makes electricity expensive isn’t use. It’s underuse.
For this reason, steady demand has historically driven prices down, not up. When more electricity flows across the same wires, fixed costs are spread over more kilowatt-hours. Utilities recover investments more efficiently, and per-unit costs fall. . . .
In many regions, hyperscalers pay industrial rates and cover the full cost of the infrastructure built to serve them, including new substations, upgraded transformers, and transmission expansions that remain part of the grid long after a project is complete. In northern Virginia, large data center customers cover roughly 9 percent of transmission costs, helping keep residential transmission rates below the national average. In Mississippi, revenue from large data-center loads has funded grid modernization without raising household rates.
This is not theoretical. It’s how grids stay solvent.
The campaign to demonize AI data centers is no doubt intertwined with the growing sense that artificial intelligence is going to be harmful to American life in the long run.
AI is a tool, so its value or harm is entirely dependent upon how a person uses it. The internet is what enables you to read this, or get my books delivered to your door, or seek connections and to advance your career on LinkedIn, share cool pictures on Instagram, stay in touch with old friends on Facebook, or yell at strangers on X. The internet also enables human beings to form hate groups and connect with each other, trade all manner of pornography — including illegal types — send anonymous death threats, spread misinformation, and anonymously make personal attacks.
Once you’ve invented a technology, it is very hard to uninvent it; the genie does not easily go back into the bottle, nor the toothpaste back into the tube. If you want to reverse a particular technological development, your best bet is to make it no longer cost-effective. The Concorde was the fastest passenger jet ever built, with cruising speeds faster than Mach 2, once taking passengers from New York City to London in less than three hours. But after 27 years of service, it became clear that the Concorde’s fuel costs were prohibitive and there wasn’t a demand for fares high enough to make the flights profitable.
But I notice one detail in the Smithsonian Air and Space Museum’s writeup of the history of the plane: “The emerging environmental movement was entirely against the Concorde. What was once synonymous with supersonic and luxury soon became synonymous with noise and exhaust emissions.” That sounds an awful lot like the progressive complaints about AI data centers today.
ADDENDUM: You never know what kind of criticism a reader is going to make; in response to my most recent column about presidential libraries, one reader over at that other Washington publication fumed, “Maybe if you had criticized [Ronald] Reagan in the first place, we wouldn’t be here now.”
Putting aside the fact that I’m a Reagan superfan, with my “Bedtime for Brezhnev” poster visible over my shoulder in my home office in every video appearance . . . reader, just how old do you think I am? I was in kindergarten when his presidency started.