Why I think panic about local impacts of data centers is just a panic

A request for counterexamples

In the last year of following increased local resistance to data centers being built, I’ve listened to lots of recorded testimony at town halls, and read through countless comments and articles where people argue for why data centers are so uniquely evil and shut down anyone defending them as shills for AI companies.

Throughout all this, the only claims I’ve seen that actually stood up to scrutiny were:

• Data centers will meaningfully add to America’s CO2 emissions.

• Elon Musk’s data centers specifically added local air pollution to already heavily polluted residential areas. This is the clearest case of a data center I would’ve personally wanted to prevent from operating.

• Data centers built right up against people’s homes create noise pollution.

• Data centers can sometimes in specific circumstances raise electricity bills somewhat, though this hasn’t been the norm so far.

For everything else, whenever I look into where people are actually getting their ideas about the hundreds of other data centers being built, the source always leads back to some confused misreading of local reporting, a wild calculation error, a bad game of telephone, or a wildly misleading article.

Almost everyone talking about data center water impacts shares the same story of the Georgia couple whose wells ran dry, and never mentions this was definitely due to some issue with construction and not the normal operation of a data center. People regularly claim everyday people’s water access has been harmed by the normal operation of data centers, but no one can ever point me to where that is. Most stories I read about droughts being “exacerbated by data centers” don’t add the portion of water data centers are actually using, and when I look into it, it’s almost always vanishingly small compared to the area in general. Talk about data centers “poisoning water” seems to have been invented out of thin air.

Most people commenting on data center impacts on electricity prices bring up a (I think bad) Bloomberg article that people easily mistake to read that data centers raise electricity prices as much as 267% near where they’re built, despite this reflecting a rate that residents don’t actually pay at all and only one of several big factors that determine electricity bills, this mechanism not actually having any causal effect on the residential rate increases the article goes on to describe, and the fact that the authors are starting from a baseline in April 2020 which was a unique low point for electricity prices as COVID lockdowns were in effect, and the nominal price of electricity rising a lot in general in the last 6 years, mostly due to inflation. In reality there’s a surprising negative correlation between the states with the most data center capacity and states where electricity prices have increased the most.

Data center heat exhaust harms seem entirely speculative. The only one that looks really dangerous so far is the very largest data center ever proposed, in the place most likely to trap heat (Utah desert valleys), and the only study done on the air effects of data center heat exhaust found it was comparable to the normal slight temperature increases of living near other dense human activity in a city.

The most popular video made about data centers this year was full of wild misrepresentations of every study it was citing. No claims from it held up.

People commenting on the air pollution from data centers in general almost never talk about how they compare to other things, and when they do they find that, say, the largest cluster of data centers in the whole world consuming as much energy as a large city together emits about as much harmful pollution into the county as a single gas plant. Not nothing, but definitely not some unique environmental catastrophe. The main pollution harms of data centers are I suspect actually offsite coal plants far away they may draw from.

Data centers for the most part pay huge amounts of taxes almost everywhere they’re built, with the exception of maybe Michigan. More on that in a future post.

The so-called massive environmental costs of training cutting edge AI models seem to fade into the background of other things we create for hundreds of millions of people to use over months. My best guess is that the emissions are the same as a day of normal manufacturing of Coke cans.

The idea that America doesn’t “have enough land” for buildings optimized down to the literal atom that will collectively take up about a third of the land we currently use to grow Christmas trees seemed so obviously silly that I felt ridiculous posting my land use article, and yet that’s the one that’s drawn the most scorn and backlash by far.

After a year of basically feeling like I’m living in this weird dream reality where every single thread I follow leads me to a dead end instead of a reliable source, I need to basically beg my critics: I need a specific case for why data centers are going to be especially harmful along these axes compared to other regular industry, instead of just fading into the background of the normal externalities of other large light industries, because I just can’t find it at all. I need something, anything, that actually justifies this public outcry. I need some kind of basic overview of water harms that doesn’t just gesture at “millions of gallons of water” without context. I need some realistic estimates of air pollution when data centers with onsite backup generation are following normal EPA rules where they don’t just fade into the background of other normal sources of air pollution. I’m just not seeing it.

I’ve received one good account of why data centers can throw off the ways water rights work in specific areas, which I’ll publish and respond to soon. That specifically struck me as good pushback. But besides that, I want to send up a signal flare to try to make the case to me beyond just saying “Everyone knows these things are a disaster.” Show me your thread of reasoning that leads back to something solid that I can grip onto and understand.

People will often cite “experts” but these experts are often random professors just echoing common wisdom that’s developed about this rather than actually giving specific solid numbers for why we should be especially worried here. I have yet to find a serious expert in the field giving a comprehensive reason why we should be especially worried about data centers. There are experts who have raised legitimate concerns about specific data centers, like climate experts and physicists concerned about the warming effects of Kevin O’Leary’s massive Utah data center, the largest ever proposed, but these never generalize into giving us much alarming info about the buildout more broadly.

On the climate criticism specifically, I’m seeing two camps:

• People worried that data centers will add huge amounts of emissions despite AI being useless.

• People worried that data centers will add huge amounts of emissions because AI will be really economically useful and powerful and continue to grow.

The first view just doesn’t make sense to me. If AI is a fad, it will not rise to be whole percentage points of American emissions. At some point people will wake up before we spend that much energy on it, because that much energy is expensive.

I myself take pretty weird wild futures of AI seriously, because I think it’s clear that AI’s capable and likely to get much more capable relatively quickly. In those worlds, AI can potentially rise to be huge portions of our energy grid and emissions. But almost no one I talk to about climate seems to share my worries about this. Their objection is that emissions are wasted on AI because it’s useless rather than that we’re doing something potentially incredibly dangerous that could at best lead to explosive economic growth. The people I know worried about AI risk will often imply that everyday people protesting data centers are also secretly worried, yet I barely see this coming up in public discussion.

I haven’t had time to finish deep dives on every last issue with data centers. I need to write my big climate post, finish a series on pollution, write something on heat exhaust, and explore noise. But I do already know a lot about each, and for each I am just not seeing anything that justifies the level of public outcry we’re seeing. And yet when I venture into polite society and say this, I’m often treated as basically a conspiracy theorist or a shill for big AI promoting what amounts to climate denial. All these ideas that I know individually are overblown work together as a kind of shield. If I poke at any one, the other person leans back on all the others. They can’t possibly all be wrong, that would be ridiculous! And yet for any one I poke at, they can’t lead me back to anything that seems solid at all.

This is a weird position for me to be in! I’ve always been a good predictable lib and don’t find myself on the opposite side of almost all media coverage often. It’s driven me a little crazy and does make me worry I’m missing something everyone else seems to see. But after a year of following this with no really satisfying answers, it’s become hard not to start to think I’m the one with my feet on the ground here.