I think the data center backlash is mostly about data centers

Andy Masley，摸着石头过河。点击阅读 Andy Masley 的 Substack 专栏，已有数千订阅者。关注

This Heatmap poll has generated a lot of discussion on Twitter in the last few days:

People from basically all political sides are interested in owning the data center backlash, and are all saying it’s being caused by what they’ve always been on about. It looks too powerful and intense to be about data center buildings themselves, so it must secretly be about something else, and commenters tend to assume that secret other motivation conveniently lines up with how they themselves think about politics and AI. People who are anti big-tech will say it’s a backlash to big tech. Some say it’s a broad rejection of economic growth as a measurement of human wellbeing.

My fellow AI safety enthusiasts will sometimes say it’s a reaction to rapid unsafe progress in AI that threatens to destabilize society or destroy the world. People in the “AI is bad because it’s bullshit and fake” camp will say it’s a reaction to AI flooding the world with slop, and Americans recognizing we’re investing in something stupid.

I follow a general rule that anything narratively satisfying is suspicious, and also that everyone’s thinking is less structured and coherent than we like to imagine. For me, it would be so much more satisfying if this bizarre massive uprising based on ideas that often collapse under mild scrutiny were all secretly about my own worry that AI could be incredibly dangerous incredibly soon. The idea that this is all happening because people did actually just get fired up about often goofy interpretations of isolated large numbers is much more disappointing, but reality is often disappointing. You must try your best to dwell in anti-paranoia when you can.

I think any movement that wants to either stop the backlash or capitalize on it will be making a mistake if they assume secret motives are driving it. However, because people’s beliefs are often opaque even to themselves, it seems possible for politicians to capitalize on the backlash and meme people into thinking this was “always about” some other unrelated specific thing, so if you’re trying to do that, this seems like a fruitful time to twist the narrative, but I don’t think you should go out assuming it’s actually true. I’ll argue here that based on the limited info we have, it looks a lot like the data center backlash is not about AI safety or job loss or AI being a scam or social media or big tech having too much power or anything else like this, it’s about data centers and the claims people believe about them.

Of course, the data center backlash being all about object-level claims and facts about how data centers interact with the places they’re built would be very convenient for me personally, so take it with a grain of salt. Maybe I’m no better than all those other commenters, but I think my case here is pretty solid. I’ll go through other explanations offered. This will be more speculative than a lot of my posts, I could be getting this one pretty wrong.

It’s probably not about big tech

It’s tricky to disentangle worry about big tech from worry about AI. To try to separate them, I’m going to consider the hypothesis that Americans are lashing out against data centers because they don’t like the tech companies building them in general, and want to find a way to limit their power over society, and data center opposition is a tool they think will have some effect. This hypothesis seems to imply that when the average person hears “A data center is being built by AMAZON” it’s Amazon that really sticks out to them as the bad thing, and they start to formulate how to stop the data center. The data center is a coincidental building that could as well be something else.

What would we expect to see if this hypothesis were true? A few things:

  • Americans would dislike the tech companies building the data centers about as much as they dislike the data centers. If opposing a data center is just a tool to harm a tech company, you’d think about as many people would hate the tech company as the number trying to oppose data centers to harm the company.
  • Americans would also oppose other things being built by and for the same companies building data centers.

Neither of these seem to be the case.

Americans love big tech, as long as you ask about specific companies

If you poll people on how much they like “big tech” you’re going to get negative responses, but I think any vague powerful sounding entity like this is also going to poll negatively. If instead you replace them with the big four hyperscaler companies building out the biggest AI data centers, Americans are mostly pretty positive on them. The four companies building the biggest data centers right now are Google (technically Alphabet), Amazon, Microsoft, and Meta.

Amazon is one of the institutions Americans view most favorably, competing with “the military”

Even Facebook is +13% in approval, with just as many people very favorable to it as very unfavorable. More Americans are favorable to Facebook than NATO or the European Union. A more recent poll from The Argument, which didn’t include Meta, also found strong approval for the other three hyperscaler companies actually building data centers.

So the specific story that voters are looking to “get back” at the hyperscaler companies for what they’ve done to society using the only lever they know how to use looks wrong. I would think that if you’re so motivated to punish big tech companies that you want to block their attempts to build things near you, you would also at least say you “strongly disapprove” of the companies on a survey, yet the number of people who strongly disapprove of Google is 5%, Amazon 8%, and Microsoft 7%. Meta’s the outlier at 22%, but is still +13% compared to data centers’ -60%. So far this doesn’t look like a promising explanation.

When Americans are asked about tech companies in the abstract, they think they need “more regulation”

but I’m not really sure what this tells us. If I asked anyone on the street “do you think major companies are regulated too much or too little?” I’m willing to bet they’d default to too little regardless of what they otherwise believe about specific companies. A lot of the polling about very general industries like this leaves me worried that if you ran polls on any industry calling it “big __________” it would also get mostly negative approval. I suspect Americans have a negative view of “big ag” but a positive view of “farmers” and (outside of Monsanto, because they mostly only hear about it in negative stories) neutral to positive views on a lot of food companies. How do you feel about Archer-Daniels-Midland?

The fact that the hyperscalers each get net positive favorability I think at least complicates the story that the data center backlash is just a concealed way of lashing out at them.

The other thing that complicates the story is that this doesn’t seem to be happening with other buildings associated with tech companies.

Americans are kind of fine with other buildings associated with big tech

If someone felt Amazon had too much power in society and wanted to stop them by putting up blocks to large buildings Amazon was trying to build, what specific type of building would be most likely to be blocked? Obviously Amazon warehouses/fulfillment centers. And yet Amazon distribution centers are some of the very most popular buildings in America.

If you want to “strike back at the power Amazon has over our lives” I don’t see a better way of doing that than opposing a distribution center. This directly harms Amazon’s ability to affect your community’s buying habits.

It’s interesting that the chip manufacturing plant polls worse than the traditional manufacturing factory, with an additional 11% of people opposing it, but also a lot more people being unsure. Maybe that additional 11% do actually oppose AI specifically or big tech in the abstract? It’s also interesting that an AI data center does meaningfully worse than a traditional data center, which also polls much worse than other types of buildings. It seems like this could be some combination of people disliking AI specifically, or people (correctly) assessing that the really wildly large new data centers going up are mostly for AI, that these are often orders of magnitude larger than traditional data centers, and therefore that they’re uniquely dangerous for environmental reasons. I’ll explore this more in the next section. Either way, data centers seem to be uniquely unpopular compared to other buildings associated with the same tech companies building them, though we could really use more polling on this. The way they stand out seems to fly in the face of the idea that they’re just a way to get back at tech companies in the abstract. Why are Americans then not also reacting to the buildout of other similar buildings? One response could be “they benefit from Amazon warehouses in a way they don’t from data centers” but it seems a little weird to imagine that Americans are lashing out against tech companies’ influencing more of their lives unless the tech company is doing something that will specifically make its services much more directly present in their lives.

So I come away from this data pretty uncertain, but much more skeptical about takes like this specifically:

It’s probably not mainly about AI

I’ve scrolled by a lot of very confident takes that the reason people are reacting so negatively to data centers is that they can sense that they’re the tip of the spear of some potentially horrific alien future with advanced AI. That’s how I feel about them personally!

But polling really doesn’t seem to imply this is mainly a reaction to AI. It’s obviously true that a lot of people are motivated by their beliefs about AI, but this doesn’t look like the main story to me. Whenever Americans are given the option to say “Among other issues with data centers, I also separately either think AI is a useless waste or a dangerous technology and that’s influencing how I think” they mostly opt not to say that. I think this is partly due to the average person just not actually thinking about AI that much, and separately having seen tons of news stories about data centers harming local communities.

Polling implies AI’s playing a small part in the backlash

Here’s some polling on why Americans say they oppose data centers. This from Fox News in July asked voters what their “primary reason” was for opposing data centers:

This seems to be a clear story of voters worried about the immediate impacts of the buildings themselves, mostly not AI more broadly. But maybe this poll excludes people’s worry about AI? Just because it’s not their first concern doesn’t mean it’s not playing a part.

An earlier Gallup poll in May asked respondents to just write out why they were opposed to data centers however they wanted. Gallup then organized them into lots of different general worries. Here are the results:

Something unfortunate here is that as far as I can tell, Gallup didn’t give us a way of clarifying how much overlap there is between “negative views of AI” and “specific AI concerns” and how they classified them, so people’s concern about AI could be as high as 27% or as low as 14% (if the entire 13% of people with specific AI concerns were also categorized as having “negative views of AI”). Still, when people opposed to data centers were given the option to clearly say why, at least 3/4ths didn’t mention AI at all. This leads me to think the backlash isn’t mainly about AI, at least not consciously.

Most recently, a survey was done that found that when voters learned that a data center would recycle the same water for 10 years instead of drawing new water, their support jumped by 31 points (!!!!!!!). Voters say they’re worried about water, and when they learn water won’t be a big deal they stop worrying. points out that the survey results are deceptively framed in the document I shared, and that the 31 points actually mean “31% of people become more supporting than they were before” which could mean anything. I retract this point.

Americans have mixed opinions of AI

To be clear, Americans seem to have skeptical though mixed views on AI.

The polling we have on American attitudes toward AI isn’t great right now in general so I don’t feel like I can draw much from it. It would make sense to me that a big part of the data center backlash is that voters don’t feel like they’re benefited from what’s happening inside data centers and therefore are more inclined to criticize them.

But the polling we do have doesn’t show the kinds of dramatic swings you’d expect to explain the sudden and intense data center backlash. The percentage of people who are “more concerned than excited” about AI has stayed the same since about the release of ChatGPT.

Young people’s concern has significantly risen, but older people’s has stayed relatively flat in the past 4 years.

This is interesting, but I don’t think it fits with the sudden massive drop we’ve seen in data center attitudes:

Meanwhile, the share of people who “haven’t heard much about AI” has also stayed about the same, at half the population.

All this combined with the survey data on Americans’ stated reasons for opposing data centers makes me think that attitudes toward AI are probably a meaningful factor for a minority of people opposed to data centers, but I think the backlash overall is driven by people’s thoughts on AI.

It’s probably not China

Is this actually all a Chinese psyop? I really don’t think so. I have a long separate post here looking at the evidence we have, and I come away thinking the data center backlash is basically entirely homegrown.

I think the backlash can be mostly explained by years of media coverage of object-level local environmental and economic harms

For the past few years, basically all news outlets everyday Americans trust were saying over and over that there were these strange new buildings plopping down without the locals knowing and using a disturbing amount of water and energy. Here’s a representative sample of lots of the most popular news articles on data centers between 2023 and 2025:

Some of these are legitimate stories and others are strange isolated statistics or just flat out fake (the idea that data centers raise electricity bills by 267% is this zombie stat that won’t die, the number referred to a specific electric wholesale price at a node near a nuclear plant in New York that had recently closed down, not residential bills, and via this game of telephone we now have politicians like Elizabeth Warren repeating the misconception).

A lot of media coverage has gone on to just share people’s broad concerns without seeing whether they stand up to scrutiny. Many reporters now seem interested in seeing “what everyday people have to say” without clarifying what they’re getting right or wrong. Take this recent clip about the backlash from ABC:

Very very few media organizations are both reporting on locals’ reactions (which note in this video are entirely about the environment) and then contextualizing them with the actual amount of land, water, or power the data center would use, and how that compares to the places they’re built. Because there’s so little incentive to correct this stuff, it goes uncorrected, and millions of viewers at a time see upset sympathetic locals in news stories that leave them thinking all their concerns are probably correct.

All of this has also been amplified a ton on social media, which created lots of common wisdom among everyday people about data centers that they didn’t have much reason to question. I think these ideas are kind of “locked in” now because they’re perceived to be the expert consensus on data centers, so people are more resistant to questioning them for the same reason they’re (rightly) resistant to specific anti-vax arguments.

With no real reason to each individually do some massive deep dive to try to figure out what was actually going on or put the number in context, people naturally just accepted that these were bad for local communities, and this became the default wisdom among educated people. Then the data center buildout accelerated, and suddenly more people found that their communities were also considering data centers. Why would this be happening if all educated people agreed that these were terrible? This was probably pretty annoying and confusing to a lot of people, who have now seen so much consistent negative coverage of data centers that saying they’re fine seems almost like a conspiracy theory. This all just seems like a much more satisfying explanation to me! It fits the limited polling data and evidence we have.

I’m left uncertain, but I lean in the direction that the data center backlash is mostly about object-level concerns about data centers, and the main role AI plays is that people’s opinions of it are too mixed to justify looking into whether their negative beliefs about these large strange buildings are correct. I’ll eventually get around to finishing my history of the data center panic, but I’d recommend that first linked post if you’d like to see more of my thoughts on how this all started.

One last thing to consider is that data centers are so unpopular now that it might just be fun to join in on the hate. A recent NYT opinion column talked about the communion people feel with each other in this fight. “People like hanging out with each other to attack a perceived enemy” on its own might go a long way in explaining the most intense reactions we see. The column itself gives exactly one sentence passively acknowledging why people are mad in its overview of how much this brings people together.

Some people will bizarrely imply that saying this is all about object-level worry “ignores what voters are actually mad about” or doesn’t take them seriously. I think it’s important here to just look at what voters are actually saying, and take that seriously, rather than trying to work out their inner psychology and secret beliefs (that always so conveniently line up with the person writing about them), and that this is the more democratic and fair way of engaging in data center discourse. I can then make my own case that, with nothing but respect to the voters, they are often mistaken about the facts they share. This is I think a much more respectful, direct, and democratic way of engaging compared to imagining that they all secretly agree with me about the very immediate dangers AI could pose.

添加评论
点赞收藏
点踩分享查看原文
评论
?
参与讨论