In his new book “Hyperscale,” journalist Paris Marx investigates the impact of data centers and generative AI, taking readers inside the data centers themselves and explaining how computing infrastructure that goes back decades recently became a political landmine.
The book is well-timed, to put it mildly, as data center backlash seems to have been one of the motivators for the massive blue wave seen in the recent midterm elections. And those debates are likely to continue as Democrats wield more governmental power.
In an interview with TechCrunch, Marx described the data center buildout as “fundamentally unsustainable” and “completely out of line with a livable planet,” while also arguing that these facilities make for an obvious target amidst broader anger at the tech industry. After all, if a data center is being built in your community, “it’s very physical, it’s very material.”
“It doesn’t matter what side of the political spectrum that you’re on,“ Marx said. “You are going to feel the effects of that in your quality of life, in your power bill, potentially in your access to water, potentially in what you hear around your community, if these air conditioning units on the roofs of data centers are making a lot of noise.”
He also suggested that data center opposition is only the most visible piece of a wider movement targeting “the physical manifestations of the tech industry, […] whether it’s the opposition to Flock cameras rolling out across the United States, or the smart glasses that Meta is trying to put on people’s faces, or even the self-driving cars that Waymo is rolling out.”
During our conversation (which I’ve edited for length and clarity), Marx and I discussed data centers’ environmental impact, how tech companies are responding to community resistance, and what it’s actually like inside a data center.
I liked your description of trying to get access to the inside of a “hyperscale” data center and then eventually taking a tour [of one]. Could you talk a little bit more about what that experience was like? How did you get in? What surprised you when you actually got in?
Getting inside of a hyperscale data center was not the easiest task in the world. I spoke to some researchers who have been doing research on data centers for a number of years longer than I had, and they told me that before the generative AI moment, and before all the attention that has been put on data centers recently, that it was quite a bit easier to do so.
I sent emails to a number of companies asking if I could get inside, often didn’t get a response, other times got a no. I ended up, after visiting the outsides of some data centers in places like northern Virginia and Ireland and spots like that, I actually went down to New Zealand to go inside of a data center itself. And the thing that struck me most was the level of security.
Obviously, you would expect that this is an infrastructure that is really securitized, because the data inside needs to be kept safe, obviously [they] have these big fences and all the security cameras around the data centers. But to go inside, I had to send my passport in advance. Then when I got there, they had to check the passport, and then as we were going through the data center itself, every single door in the data center was locked unless you had a key card to unlock it.
Then, when you get inside of the server room itself, where all this computer hardware is stored in tall racks that look like cabinets — you can imagine, in your mind, that this would be kind of a sinister location, with some eerie lighting or something like that. But in truth, it’s just quite utilitarian. It’s a warm room with some air conditioning blasting into the servers from vents that are nearby.
It was really loud as well, because you have all these fans running from all these servers, it was really kind of piercing. To be inside of this room, if you were going to be in there for a while, you’d probably want earplugs or something like that.
Before I read your book, I thought of data centers as something that’s important, but a little bit boring. How did you decide they were interesting and rich enough to sustain an entire book?
I felt the same. I was aware that data centers had been out there, but it was not something that I was paying particularly close attention to. But then there were a number of stories that of tipped me off to the fact that there might be more here, and that it was worth looking deeper into what was going on — the first being hearing about the water pressures created by data centers in Chile around the capital of Santiago, where companies like Google and Microsoft were building data centers, and communities were concerned about their access to water and started organizing around that.
And then hearing about how much electricity data centers were consuming in Ireland, to the point that now they consume more than 20% of the country’s electricity, and that was obviously pushing up power prices. It was putting pressures on the grid, such that the grid operator actually put a moratorium in place for grid connections for data centers around the Dublin area, back in 2022. This was even before the generative AI moment that caused a surge in construction of data centers.
And then the other one was in The Dalles, Oregon, where Google had actually built its first company-owned data centers about 20 years back. In that case, its water deal was coming up for renewal. It wanted to build more data centers there, so it needed more water. And when the residents started asking, “How much water is Google actually using?” a journalist at The Oregonian went to the city and essentially asked, “Hey, how much water is Google using? Our readers would like to know this,” he was hit with a lawsuit instead of being given that information. Because Google did not want it to be public. And then, of course, when it did come out about a year later, Google revealed that it was using 29% of the city’s water, which seemed quite significant.
That was around when ChatGPT was being released and generative AI was starting to take off. It felt like, okay, there are these pockets of places where a lot of data centers have already been constructed, where they are putting pressures on the communities that are causing people to respond, and it felt like that was going to start coming to a lot more communities. It was a good time to really start to dig in and pay attention.
You don’t just talk about the data centers’ environmental impact, but also, they [become] your way in to talk about this broader tech moment. They represent to you — particularly the hyperscale data centers — represent to you how these tech companies are doing business now.
Absolutely. The hyperscale data centers and just the scale of the construction — whether it’s the the size of the actual facilities themselves or the number that these companies are trying to build out, just feels fundamentally unsustainable.
They’re just demanding too much from the communities where they’re being located, whether that’s because of electricity or because of water use or what have you, or just the fact that they’re building out so many, that when you think about the cumulative impact and you look at the demand for water, but particularly for energy from fossil fuels, that means that it’s just completely out of line with a livable planet and an amount of computation that feels like it makes any degree of sense for us to have constructed.
I think it really goes back to the hubris of the tech industry and of these major tech companies and the tech billionaires that run them, and the fact that they have these particular visions of the future that are based on fantasies, but also on expanding their own power and influence over the world, and they don’t seem to mind that everyone else then has to pay the cost that comes with their ambitions and what they are seeking to do.
And I feel like it’s become particularly egregious with generative AI and the data centers to power them, where many of the benefits that were supposed to come of generative AI do not appear to have materialized — whether those are the economic benefits, the productivity benefits, but also the the benefits of delivery of services like healthcare and education to the public, as we heard Sam Altman saying a lot in the in the early days of ChatGPT.
Paired with that, there has been this massive environmental cost […] but there’s also this massive and growing social cost, when you think about the mental health impacts of chatbot dependence, of all the false information that pervades social media platforms now, of the deepfake crisis, and I could go on and on and on.
It’s not surprising that there has been backlash. But I have been surprised at how widespread it’s been, and also how bipartisan — you talk about Ron DeSantis and Bernie Sanders both being opposed to data centers. Why do you think it’s become such a unifying issue?
Because it’s not an abstract issue, right? I think even when we think about generative AI or social media or something like that, [it] can feel abstract. It’s a service that you access from your phone or from your computer. It’s not something that is tangible in the world.
But when a data center is coming to set up shop in your community, it’s very physical, it’s very material. It doesn’t matter what side of the political spectrum that you’re on. You are going to feel the effects of that in your quality of life, in your power bill, potentially in your access to water,. potentially in what you hear around your community, if these air conditioning units on the roofs of data centers are making a lot of noise.
I think that there was already a pre-existing frustration with the tech industry that came from the quality of the products, the effects that they’re having on our society and our lives. For some people, [they have] frustration with the way that the tech industry has been collaborating with the Trump administration and with right-wing political groups in different parts of the world. But then you add on top of that the actual effects that come of having one of these infrastructures be planned or located in your community, and that anger gets kind of directed into that physical manifestation of this thing that you have an issue with.
Obviously, the data centers are at the forefront of this, and it’s what we pay the most attention to. But we also see this increasingly with other aspects of the tech industry, whether it’s the opposition to Flock cameras rolling out across the United States, or the smart glasses that Meta is trying to put on people’s faces, or even the self-driving cars that Waymo is rolling out.
There’s a lot of frustration with the physical manifestations of the tech industry, because that is something that can really be felt more directly, but can also be challenged more directly. You can delete an app on your phone, but what difference does that really make to the company? It doesn’t often seem like very much.
The tech companies have reacted to this backlash in a few different ways. One is that Microsoft, Amazon, and OpenAI have all talked about [their commitment] that they’re not going to use NDAs anymore in their negotiations with local governments. How big a difference do you think that will make?
I do think it’s going to be good that it will be easier for people to access information on these projects that are being located in their communities. But part of the reason why we’re seeing these shifts is exactly in response to that opposition.
We’re seeing the tech companies propose to pay for more of the electricity infrastructure that will be required for their data centers to reduce the pressure on ratepayers, because people are angry about that. They’re promising to use different water-based cooling systems in their data centers, to say that they’re able to use less electricity.