Data center criticism is based on bad science and scare quotes

If you are following the growing artificial intelligence industry at all, you’ve probably come across some facts and figures about how AI is going to use up all of the country’s water, while also using up all the power and causing prices to spike for regular non-AI using humans, and also that the giant data centers required for this industry are taking over all the usable land, and will cause pollution to soar, and/or lead to a wide range of other negative societal effects. I don’t think of myself as an apologist for AI — although I have written about how it can be useful for important things like medical research, and how using it as a tool for writing isn’t always a terrible thing — and it’s possible that anti-data-center arguments have some merit purely as an anti-billionaire measure, but at the same time, many of the arguments I’ve noted above simply aren’t rational. They may contain numbers that sound terrible — billions of gallons of water, megawatts of power, etc. — but when looked at rationally they tend to collapse.

According to a recent Gallup poll, seven out of 10 Americans are opposed to the construction of data centers for artificial intelligence in their local area, and almost half of those surveyed were strongly opposed. In the same March survey, 53% of Americans say they oppose building a nuclear energy plant in their area, far less than the 71% opposed to data center construction. Since Gallup first asked the nuclear power plant question in 2001, the high point in opposition has been 63%. Half of opponents mention data centers’ excessive use of resources, including almost 20% mentioning either their use of water or energy. Sixteen percent mention a related environmental concern of pollution, including noise pollution and air and water pollution. Most of the remaining opposition stems from general or specific concerns about artificial intelligence.

If you want to keep up with the revolt against data centers, there’s a dedicated website called Data Center Watch, which says it is run by a “boutique research firm tracking the growing opposition to data center development.” According to the site, local residents blocked or delayed about 20 projects around the US in the second quarter of last year, representing nearly $100 billion in proposed investment. Residents across the US are attacking and/or trying to block data center projects, according to a report from the Wall Street Journal:

Unnerved by the potential rise in electricity costs and land use from the Howell Township, Mich., project, neighbors in the roughly 8,500-person community outside Detroit knocked on doors, hung up fliers and flooded Facebook groups to rally support for blocking the effort. Across the county line, residents in Saline Township have fought a data-center proposal tied to Oracle and OpenAI. That roughly $7 billion project would potentially require enough electricity to power at least 750,000 homes. It recently got conditional approval to move ahead with the effort despite pushback from locals.

Note: In case you are a first-time reader, or you forgot that you signed up for this newsletter, this is The Torment Nexus. Thanks for reading! You can find out more about me and this newsletter in this post. This newsletter survives solely on your contributions, so please sign up for a paying subscription or visit my Patreon, which you can find here. I also publish a daily email newsletter of odd or interesting links called When The Going Gets Weird, which is here.

Bad atomic bomb analogies

Robert Bryce, an author, filmmaker, and public speaker who says he has been writing about energy, power, and politics for four decades, has a post called Tracking the Great Data Center Revolt, in which he describes how he has been reporting on land-use fights over various energy projects for over 15 years, speaking to dozens of people from all over the world — farmers, ranchers, politicians, and ordinary citizens — and documented nearly 1,200 rejections or restrictions of solar, wind, and battery projects. But Bryce says he has “never seen anything like the raging backlash against data centers.” Since January 1, there have been more rejections or restrictions of data centers in the US than in all of 2025, and the number of rejections in 2026 is already nearly equal to the total number of wind projects rejected worldwide in 2025.

At least some of the opposition to data centers seems to be based on a generalized dislike of billionaires, which in some cases seems totally fair. Kevin O’Leary of Shark Tank, for example, is a notoriously unlikeable person — a reputation he seems to revel in — so it’s not surprising that residents of northwestern Utah would be opposed to his data center project. According to some estimates, it’s designed to reach a 9-gigawatt power capacity, which would make it one of the largest “hyperscale” data centers in the world, and would consume twice as much power as the entire state of Utah currently uses. O’Leary didn’t endear himself to anyone when he suggested that opponents to the center were being funded by China, which he said his private investigators discovered — an allegation that two of the opponents he singled out made fun of.

According to Rolling Stone, Utah is facing a critical water shortage driven by the warmest winter in over a century, with snowpack levels at the lowest ever recorded, and scientists say that heat and emissions generated by a colossal data center like the Stratos project would “wreak havoc on an area already severely impacted by climate change.” The magazine explains that there’s concern that the amount of water needed to cool the new facility could further drain the Great Salt Lake, and wildlife biologists say the heat generated by the center could also disrupt the movement of migratory birds and deer. Patrick Belmont of Utah State University, says it’s like “putting a hairdryer that has the energy consumption of New York City in the middle of a fragile desert ecosystem,” an analogy that doesn’t really make any sense but at least shows some creativity.

A physics professor from Utah State, meanwhile, says the heat generated by the center would add up to 7 or 8 gigawatts of energy in the form of waste heat, which he says would be the “equivalent of about 23 atom bombs worth of energy dumped into this local environment every day.” According to the prof, it would raise local temperatures by five degrees Fahrenheit during the day and up to 28 degrees at night. Even on its face this comparison seems vague and also ridiculous (how big are these bombs? How can heat dissipated by evaporation over time compare to the instant blast radius of a nuclear explosion?). Science writer Andy Masley writes that it gives readers “as much information as reporting how much sunlight will hit you in a day in units of hyper magnified beams.”

Consider the effects of concentrating sunlight in a single place with a magnifying glass. This can do a ton of damage or start a fire. But all this heat energy was already there before, it was just impacting a surface area the size of a magnifying glass, instead of a tiny point. If you concentrate a normal amount of heat energy in a very tiny place, it becomes dangerous. And therefore, if you expand a dangerous amount of heat from a hyper concentrated point to a larger area, it can fade into the background of the normal levels of heat we’re exposed to. A data center spreads its heat exhaust very evenly across the day. Data centers run at basically constant loads 24/7. There aren’t really peaks or low points. In comparison, the Trinity Test released heat in less than a second.

Almonds or data centers?

A new study that has been getting some attention recently says that data centers can create “heat islands” that could raise nearby temperatures by as much as 9 degrees Celsius, which is a a huge amount. Sounds pretty terrible! However, Masley says this study uses a methodology that is “completely off-the-wall goofy,” and in any case doesn’t show anything close to what the authors claim. For example, the study implies that the entire temperature change is due to data-center waste heat, but it’s almost certainly due to the fact that the surface of any building — data center, Starbucks, nursery school, etc. — is hotter than the grass that was there before. In addition to that study (which hasn’t been peer reviewed) there are other studies that are repeatedly cited with dangerous-sounding numbers referring to data centers that are also flawed.

To take just one example, a structural engineer wrote a post about how he was wrong about water usage by data centers, and the main reason was a flawed study from the Lawrence Berkeley National Lab. The study had a massive number for consumptive water use by data centers — 570 million gallons a day — but this includes the evaporation of water from hydro reservoirs, most of which isn’t related to data centers. When that is removed, the usage number drops about 50 percent. A biologist, an economist, an engineer and a geologist from California say that data centers there likely use 20,000 acre-feet of water per year, or about 6.5 billion gallons, while human water use in California is about 13 trillion gallons. Almond growing uses about 4 billion gallons every day. The nation’s golf courses use about 500 billion gallons of water every day.

When it comes to power, estimates are that data centers currently use about 4 percent of the power that the US generates every day, and that if all the centers planned are built that could go up to 12 percent (even the largest data center uses far less power than the average steel plant). The argument is that this pressure will cause — and may have already caused — power costs to rise. But as Matt Yglesias points out, power prices in Virginia, which has the highest concentration of data centers in the US, haven’t risen by as much as they have in Maine, where there are fewer A recent report from an energy and environmental consulting firm found no evidence that data centers are responsible for the increases in power rates in most US states. Instead, it said that such increases are likely caused by grid upgrades and other factors, such as inflation and gas costs.

Among the other issues that people often raise in opposition to data centers is the land use for these massive, sprawling buildings. Masley (not surprisingly) has a post on this as well, and it contains some fascinating figures about the land use of farms in the United States. By 2028, if all the proposed data centers are built, all of the data centers in America will occupy roughly 1,400 square miles — about 0.3% as much as America’s prime farmland occupies, and the buildings themselves will take up about 1/15th of the amount of land dedicated to growing Christmas trees. Of course, a lot of farmland produces food, but there’s a lot of farmland that doesn’t: Corn used for ethanol production takes up 1.5 percent of all the farmland in the US, and is responsible for around 1.5 trillion gallons of withdrawals per year from aquifers and lakes, roughly 40 times as much water as all the thousands of American data centers are forecast to consume in 2028.

As I said at the beginning of this post, I get how at least some of the anti-data-center backlash is a statement about tech billionaires, and I sympathize with that. There might also be some existential angst about AI and how it will cause massive job losses (a topic for a separate post) and other social ills. But it would be nice if those arguments were made more clear, rather than hiding behind selective quoting of bad studies or scare quotes about atom bombs.

Got any thoughts? Feel free to either leave them here, or post them on my website, or you can also reach me on Twitter, Threads, BlueSky or Mastodon. And thanks for being a reader.