コンピュータが多すぎる：Ingrid Burrington が Paris Marx に聞く『Hyperscale』

all these investors are jumping on that. And of course, even if they don't make money, the investors can still make a significant profit when the company goes public because for some reason, these companies can be worth crazy amounts of money on the stock market, even if they don't make any actual money. And of course, we're probably going to see that happen very soon with Anthropik or OpenAI, unless the bottom falls out of the AI industry before they get to go public, right? It's just, it makes no sense. Hello and welcome to Second World Save Us, made in partnership with The Nation magazine. I'm your host, Paris Marks. And this week, before we get to my guest, who is actually sort of me, but the interviewer is the guest sort of this time. If you remember from last time I released the book, this is what I like to do when a book comes out.

Anyway, I'll talk about this in a second. Basically, the big news is that my book Hyperscale, after months of telling you about it, is finally out. It officially came out Tuesday in the US, Canada, the UK, and I guess a number of other countries as well. And it's just kind of wild to think that this thing that I spent two years working on, researching, writing, editing, you know, all those pieces that go along with it is finally out into the world now. You can read it. So many other people can read it. And that's both a bit daunting, but also exciting, right? Of course, it has to be exciting in some ways. No, believe me, it really is. But you know what I'm going to say? If you have not picked up a copy yet, you really should do so. I put a lot of work into this book. If you like my work, if you like this podcast, if you like my writing, I think you're really going to like my book. And it's about a topic that is just so topical right now. I could not have predicted two years ago that as soon as my book came out, data centers were going to be a massive issue that not only people across the United States,

but in so many countries around the world, we're going to be talking about discussing. And of course, you know, so many people pushing back on it, try to stop them being rolled out in their communities. The backlash is real. And my book, Hyperscale, explains many of the reasons why that is happening and how we probably shouldn't have been surprised that when the tech industry started to try to build out a ton more data centers, that people were not going to be happy about it because we could see that already happening in places where they were building a lot of data centers already. So grab a copy of my book, Hyperscale. There will be a QR code here on the screen somewhere if you're on the video version. If you want to go get more information and buy a copy, you can obviously go to hyperscalebook.com where you can get more information. And of course, I would be remiss if I didn't also mention the fact that I'm going to be in a ton of places over the next couple of months to talk about my book. And maybe you will want to come out to hear more about it. I know I mentioned this last week, but I'm going to quickly tell you

again that tomorrow, if you're listening to this episode, the day it comes out, I will be in Montreal. On Monday, October 26th, I will be in London, then in Bristol. November 3rd, I'll be in Boston, then Baltimore and Pittsburgh, Oakland and San Francisco, Los Angeles. Then on November 13th, I'll be in Toronto, then Hamilton, Ottawa, Vancouver, Calgary, Edmonton, Waterloo, Ontario. And then on November 28th, I'll be over in Dublin, and I'll have a number of other days in Europe to announce very soon. And certainly I'll let you know on the show as well if you're in Europe and you want to see me over there. Basically, that's all to say I'm going to be in a lot of places. You should come out and see me. You should get a copy of my book and I would be more than happy to sign it for you if you do come to one of those events. So with that said, I know I've gone on way too long about the fact that my book is out and you should buy it. Now we should pivot to actually talking about the book itself. And that is why I invited my friend Ingrid Burrington to basically come on the show, to take over hosting the show for a week, and to interview me about the book.

If you've never heard of Ingrid or encountered her work, you're really missing out because she has been writing about internet infrastructure for over a decade and put a lot of this landscape on the map for American readers in particular when she did an 18-story online series for The Atlantic in 2015. That series certainly gave me a lot of insight into how this all works. She also wrote a book called Networks of New York, which was originally published in 2016, but was revised and republished this year. And you can still buy copies of it. I'll include a link in the show notes. She also hosts a cool and interesting podcast called Rip Corp that, you know, you should also check out. So when I was thinking of someone to come on the show to interview me about hyperscale, Ingrid seemed like the perfect person. She's someone who I was talking about early in the process of putting this book together. She's someone who is very knowledgeable about this whole space. And so I figured, of course, she should come on to talk to me about Hyperscale. And she's going to have to come back, obviously, in the future to talk to us more about her work that she has been doing for quite some time.

And that I think that you will be very interested in. But for now, we're talking about Hyperscale. I hope that you're really going to enjoy it. Obviously, I hope that you're going to pick up a copy this week if you haven't already. Thanks so much. Enjoy this week's conversation. Ingrid, welcome to Tech Won't Save Us. Paris, thank you for having me and thank you so much for writing this book. And yeah, it means a lot that you asked me to come do this. Absolutely. I really appreciate that you accepted. And I also feel a bit self-conscious that the first time you're on the podcast is actually to interview me and not for me to interview you. but clearly that's something we're going to have to change in the future. You know, we'll, yeah, we'll, we'll, we'll come back to that. Absolutely. Guests will know that Ingrid will be back next year to talk about your work and not just to talk about my work. I guess there's sort of, it always, I was still silly asking a writer like why did you write a book? But a lot of the groundwork for, for this book really started a couple of years ago with the day of vampire

series and really a lot of the work that the show has been doing for the last few years. and I'm curious about the decision to translate that into a book, you know, and what a book does in the world that a podcast doesn't do. Absolutely. It certainly reaches, or I don't know, does it reach more people these days? I'm not even sure. Maybe it doesn't, but I still felt that it was important to do it, right? And maybe it's worth saying as well, you know, part of the reason that I wanted you to come and interview me about the book is because we had been talking about this, you know, and my ideas for the book and how it was going to look from like a very early stage. You know, you were someone who was very early to seeing, you know, my thinking about how I was putting this together and kind of getting some advice from you on that. So, you know, that's why I think that you're kind of the perfect person to come talk to me about it on the podcast. And so, yeah, the decision to do the book is kind of like, well, you know, one, I was I was looking to do another book and I was looking for a subject, you know, that was part of it because, of course, I had done a book on transportation that was published back in 2022.

And so, you know, by, you know, 23, 24, it was kind of like, OK, you know, what am I going to do next? You know, what is the next thing that I really want to dig into? And at first I was thinking about doing something on the tech billionaires, you know, the Elon Musks, the Jeff Bezos, how they think about the world, that kind of a thing. But I just wasn't like, I don't know, I didn't have the motivation to like dig into that enough, I guess. like it just wasn't interesting enough to me and I felt like you know I had done so much work on Elon Musk in the past in particular and I felt like I was starting to get a bit bored with Elon Musk at the same time that like unfortunately everybody was getting much more interested in him and turning against him and I was like I've been against this guy for so long and now you're all catching up and I just don't care anymore not that I don't care but you know I was starting to see these stories about data centers in different parts of the world you know whether that was in Oregon or in Ireland or in Chile. And, you know, it seemed like this was starting to become an issue, right, as more of these were constructed and, you know, they were making these demands on different communities that, you know,

there was something here that it felt like wasn't getting the attention that it deserved and was also poised to become a bigger issue, especially as, you know, 2324 is obviously after ChatGPT and we're starting to get the, you know, the kind of early indications that the industry is in the process of at least planning to build out a lot more data centers in response to that. And so I start to feel that, okay, one, we're seeing the impacts that these data centers can have in communities where there are a lot built or where they're not kind of properly located. Maybe the resources aren't there to provide for them. And now if the industry is planning to build out so many more of them, then it feels like a lot more communities are going to be, you know, kind of bearing the brunt of this build out. and you know that that seems to be how it has played out unfortunately yeah well and I think in my opinion about sort of what books do there's also like a temporal effect that they can have that's very different from a podcast you know like it's so much easier for a podcast or recorded

interview all your work to kind of like fade into the RSS feed and books have this my friend Max Neely Cohen has described books as time weapons because they extend like beyond you. Right. They're a way of making having that argument kind of exist against a kind of like longer trajectory. We'll talk a little bit more about some of the history stuff you get into in the book. But I'm curious about also sort of the way that you organize the book, which is not entirely chronological. And I think that was the right decision. But I think it would be nice to hear why you made that decision or how you sought through organizing the different sections of the book, the chapters. Of course. And, you know, just to pick up on your point around, like, the temporality of a book, you know, I will say the podcast did come back. It was resurrected on the RSS feed last month, as listeners will know, and as you know well. So I guess that's the way they kind of keep it going, but you kind of have to republish it, right? Otherwise, it does kind of fade into obscurity, and fewer people know that it's there. But even with the first book, you know, I remember when I

was putting it together because at that time, you know, I had only really written articles and done podcasts and things like that. And it was much more immediate, right? Like if anything was going to take any time, it was maybe two or three months for a magazine article to come out. Not kind of the scale of time that a book took, right? Both to write it and then to go through the editing process, which is probably like a year, you know, the editing and printing and all that kind of stuff until it actually comes out. So you finish the book and then there's already this big time lag until it gets published. And, you know, I remember thinking like, is this even going to be relevant? Like when it comes out, you know, will people have moved on from these issues? Will it all be kind of outdated? And, you know, now we're in 2026 and that book came out in 2022, if I'm remembered correctly. And it's kind of like, I don't know, I'm sure like there have clearly been development since then. But I also think that there's still a lot that remains relevant and I still have people bringing it up to me today and are still like recommending it to people and things like that, right? And so that probably shouldn't have been

surprising to me, but it was still surprising to me in some way because I was used to, and because like in this moment, it feels like everything is so short, right? You know, the timeframes are short, like, you know, everything has to be so immediate. And yeah, I guess it was kind of like maybe a learning process for me so that I could see that things can have this longer time scale as well, even though if I'm reading books that are super old, I would still be like, yeah, this is still super relevant, right? Yeah, right. Not me. I can't have a lasting impact into the historical timeline, but all these other people, that's fine. Exactly, exactly. I don't know if you found that with your work as well. Only all the time. Yeah. But yeah, on the question of the book and the layout, it certainly felt like I don't even know how I would have really approached it chronologically. And I feel like if I would have started just with like a big dose of history, it probably would have been less interesting than, you know, kind of interweaving that history into different pieces of it. Part of the approach was

also informed by how I did my first book, right, where I was looking at different modes of transportation in different chapters and, you know, kind of digging into them individually, but also then trying to, you know, craft a larger narrative around what we were seeing, right? So dig into these specific cases or examples, but still have like a bigger story that is trying to be told around that. And so then when I, when I look to approach the data center question and the data center story, you know, I felt like that was still largely the right way to go, right? You know, give people a way into it, you know, by, by giving them some information on data centers, what data centers are, what goes into building data centers, a bit of history as to where data centers come from, and then dig into the specific issues around data centers that people are hearing about, right? Whether that's the electricity usage or the water usage or the climate impact or the wider supply chain. And I think that also makes it accessible to people where, you know, if they can't dig into the whole book right away or they're looking for something specific, they can maybe turn to that chapter and then get an idea of that aspect of what comes with

data centers. And then, of course, through that, of course, you know, I'm looking at those specific sections, but I also have this, you know, wider story that I'm trying to tell as well, right, about the scale of the build-out of computation, about how much computation is required, and about why this feels fundamentally unsustainable and is driven by, you know, particular ideas, narratives, business goals by these tech billionaires that I think are very disconnected from the reality that most people face in their lives, but also what would be best for people if we were thinking about building out technology in a way that actually served the public rather than serving these billionaires and their whims and their interests and things like that. So I guess that was like part of the motivation, but it also just seemed like that would be the most compelling way to structure it, at least, you know, as far as I could think to structure it. Yeah, I also feel like you did a good job, I think, in the way that you organized

it in addressing a lot of the places where I think the most annoying bad face arguments come up, right? Like, because it's, you know, well, but is the water problem really that bad? Or like you know but if you compare it to this or that other thing and the real problem is the supply chain it like you actually you hit all those beats quite nicely and you shut down a lot of rather tedious debates that like neither of us, I imagine, has much patience for. Yeah, I appreciate that, right? And I was certainly paying attention to it like while I was putting it together. And there were certainly moments where I was like, okay, I need to address this or maybe I need to like work this in in some way. Or like, you know, obviously when the controversy around Karen Howe's book, Empire of AI, and her water figures kind of blew up, you know, as these kind of more pro data center people found something to pull out in her book in order to attack. It was like, okay, I need to go back and check my water figures and, you know, make sure that all looks okay.

But then also just seeing, like, how they were disingenuously using the water argument in order to try to, like, muddy the waters on the argument around data centers was also like, okay, this needs to be addressed, too. You know, where they try to argue that, oh, data centers don't actually use very much water, you know, data centers aren't a problem because golf courses or agriculture use so much more. And I was like, okay, this needs to be brought in there as well, because by the time it comes out, people will have heard more of this, you know, argument or framing around data centers, and it will need to try to unpack that and explain why, you know, water use is still an issue. even if when we look at like, you know, kind of the 10,000 foot level and try to abstract it, maybe it seems as though it's not as big of an issue. Though even then, you know, the figures I saw suggest that data centers are still a top 10 water user in the United States as an industry, right? And so, you know, there is still a problem there even if you look at that kind of large scale. But the core of the problem is a local one, not, you know, kind of an abstract national figure kind of one, right?

Well, it's kind of, well, it's certainly not just a national one, right? Like, as I love to joke with you, I am a hopeless provincial thinker, which is to say an American. There is a way that the American narrative takes up a ton of oxygen in this conversation. But as you point out, like a lot of the things that people are talking about and getting really rightfully upset about right now in the United States have been happening in other places already, right? The whole behind the meter thing was happening in Ireland well before it became a talking point of hyperscalers in the United States. It's actually just because you brought up sort of the scale thing makes me want to kind of jump to this question that I sent you about, about the problem of kind of scale in in telling the stories of data centers. Right. Because you have the kind of on the ground local conditions and then you have this sort of like planet scale project of turning the world into a computer.

And you have sort of those like mezzo level things right at the level of like a supply chain or a place like Taiwan that becomes a central kind of choke point. Companies will often, as you point out, use these exceptional cases to sort of defend the whole of what they're doing. Right. We have this like good hydropowered one or there's sort of manipulation of metrics to make things look like they're not as resource intensive as they are. And I'm curious about how you think about how those like how those tensions across scales have played out in the broader backlash. And I don't know, sort of for better and worse, how that's kind of affecting resistance to hyperscalers. Yeah, I think it's really interesting to think about that. Right. And I feel like there's an interesting continuity between this book and my first book as well, where, you know, the first book was also about the scale of the automotive system, right, and why that was fundamentally unsustainable. And so when we were looking at tech solutions to transportation, if you didn't address the automobile and the centrality of the automobile in the transportation system, like especially in places like North America, but not exclusively, then you were never going to clearly address the problem, right?

Because the lack of sustainability or, you know, the kind of issues with the system largely came out of that aspect of it, right? Just the scale and the demands that come with expecting everyone to have their own automobile and things like that. And so, you know, then looking at what we're seeing with computation today, like, obviously, we have seen the Internet expand to, you know, a massive degree. And, you know, almost everyone on the planet is online, though not everybody. the types of things that we do online have become more computationally demanding over time. You know, certainly there was this discussion around streaming and, you know, how that uses more bandwidth and more data like 10 years ago or so. But now we're at this moment with generative AI where, you know, there is this significant push to, you know, really make us use very computationally intensive tools, often for things that don't really require those tools at all, Right. And so there's a need for a significant amount of computation without necessarily having like the justification for why that is needed or why that is actually useful.

And so then when you like turn it to the infrastructural side of things, as you're saying, like it exists on many levels. And, you know, obviously, I kind of quote you talking about this in the book, if I remember correctly, where I do think that there is like a difficulty in parsing it out to a certain degree. I think that the thing that motivates a lot of people who are opposed to data centers is seeing how this scale plays out in their backyard or in their communities or in their region. Because, of course, you have people, you know, I remember talking to people in Michigan, for example, who were like, we're opposed to the data center near us, but we're also going to go help the people in other communities who are opposed to the data center there because we don't just want them to be moved down the road. and that, you know, our kind of supposedly neighbors or neighboring communities then had to deal with it when we kind of escaped it, right? And there is like this recognition that the demands are just too great, even on the local level in many cases, in order to justify the build-out that we see happening. But then like when we move to more of that global level, you know, as you're talking about,

when we think about kind of the broader project that these people are trying to bring into being, then even there you see that it is fundamentally unsustainable as well, right? So it's not just that the demands on local communities can be much too great, whether that is from the demands on the grid or the demands on the water system or just the demands that come with locating this near communities that then have to kind of live with what it means to have a neighbor that is a massive data center and what can come along with that. But increasingly as we see them build out more and more data centers, There's also the question of like, you know, the fossil fuels that are required in order to power this and how it is just kind of fundamentally disconnected from any notion of addressing climate change or, you know, having a planet that is livable for human beings into the future because of the amount of fossil fuels that they want to burn in order to do it. And so for me then, I guess the argument, if we think about the question of scale, is not like we need to eradicate all digital technology and completely get rid of it and go back to a pre-digital time and go back to pencil and paper and all that kind of stuff, though I love my pencils and my paper and I don't want that to go away.

But it's kind of like, what would it look like to have a computational system that exists at a scale that is sustainable and in line with kind of delivering technology that actually works for people, rather than going along with kind of the ambitions and the fantasies of tech billionaires that has led us down this path of computation that exists at a scale that is completely out of sync with the planet, but also with, you know, a healthy society, I guess. Right. I've been thinking a lot about how the kind of glimpses of that kind of relationship to computing to the extent it existed in sort of this initial era of personal computing was arguably an accident, you know, and you get you do a good job, I think, in the book of in your historically situating data centers and broader kind of computational technologies, acknowledging both the really like deep military foundations of this stuff, but also like business and

industrial ones. And also you like the Apple era sort of early pushing of this idea of, you know, the like bicycles for the mind thing, right? That was a marketing pitch, right? A hundred percent. And I was talking about this recently with Lane Nooney, who you should also have on your show. I'd actually have you had them on the show. Nope. But I recommend those sort of liberatory seeds that sort of like Ted Nelson, Homebrew, Computer club type stuff like that was not the mainstream experience of the personal computer and and it's kind of that's always been the exception not the standard by which this has occurred which isn't to say I think that it's impossible to make it the standard so much as I think it's like it's the thing that you're kind of it's it's like we've never really done it um and part of the reason is because to make any of this work at all, it's like there's this massive kind of financial project of having to do it, right?

The project of even making computation cheap is an incredibly kind of, like you have to have a lot of other kind of revenue sources than like people self-actualizing with their technology or whatever. I wanted to use that kind of as a way to get into talking a little bit about the financialization that comes up a lot in the book. Can I just pick up on the point that you were making there before we get into the financialization piece? Because I think it's really interesting, right? Because I think you're talking about kind of the early personal computer era. And I think in the book, I bring up a bit, like obviously you're talking about Apple and things, but there was also a lot of opposition around kind of the rollout of computation in workplaces in that time. You know, obviously, I've had Chris Carlson on the show who had the magazine Process World that, you know, was looking at workplace automation and, you know, how kind of the introduction of the computer was transforming things in like white collar work. Right. In ways that I feel like that aspect of it has been rather taken out of the narrative of computation.

Like we don't think about it or we don't talk about it nearly as much as we used to. But I even think a lot about like the early or not the early 1990s, but just the 1990s in general, where, you know, the Internet is being privatized. But we also have this narrative, especially today, that, oh, if we could just go back to the early days of the Web in the 90s, everything would be so great. Everything would be so perfect. Right. And certainly that was a moment where there was less kind of commercialization of what was going on online. You know, there's no question about that. And kind of the business models that take hold in the 2000s were not fully present by then or like the infrastructures to enable them had not been fully built out. But I do think we forget that there was still platformization in that time. You know, that was still very much a thing. If you think of like AOL and things like that, these were services that already existed and that people were using. And for many of them, this was part of their experience of the Web, even though there was this more kind of Wild West experience of it. But I feel like that also kind of negates the question of the materiality of the web in that moment as well, right?

And I think that this has been a continual part of our narratives around technology, whether it's looking at the Apple keynote and just seeing the beautiful products on screen and never seeing what exists on the other side of that, is that even to enable what was happening in the 90s, you still had all this chip and hardware production that would have been largely occurring in Asia by that time, but not exclusively. considerably that was having impacts on communities, right? That very much goes against kind of the narrative, I guess, that we have of like the early internet in that moment and kind of the empowerment and stuff that came along with it, or at least it complicates it to a certain degree that I feel like is often not taken into account. You know, especially when you hear about the more, you know, kind of libertarian or anarchist narratives of technology in the web, it feels like the production of the technology is so often left out of that. And the assumption is just that, there are people somewhere who are going to have to provide this for us and they don't get kind of brought into the project and the vision of what that is. And then, of course, those types of people would say,

if we're hosting servers, we're just going to have our own servers and we're going to host them locally and all this kind of stuff. Those servers do still have to come from somewhere. But even in the 90s, there is still all of this kind of hardware that does exist. There are still data centers that are being set up. Like, you know, none of this stuff is so, so different from what we experience today in aspects, in ways it is. But I think that there's also a narrative that we miss out on where it's easy to look back with nostalgia and say, if we can just go back to that moment, things would be great. because so often that kind of nostalgic picture leaves out a lot of what was actually happening and focuses on the parts of it that we want to, you know, kind of hold up or pay attention to or valorize to not deal with the complications of what existed then as well. And so that's why I think it's actually more, you know, if we're thinking about what something else looks like, it's actually more useful to actually think about what a project would be that would, you know, create something that is very different from what we have today and grapple with the kind of difficulties and nuances of the reality of digital computation

or digital technology rather than assuming that it was so much better at a certain time. Yeah. I think it's in like the big score, Mike Malone's history on Silicon Valley. He just very casually is like, yeah, like Apple computers were being built in with like piecework, you know, soldering being done by Filipino women in Redwood City houses. And it's not even the main point in the book, but it's this massive double take moment. That was the Santa Clara Center for Occupational Safety and Health did a lot of work in that period being like, hey, don't solder in your house. That's bad. That's unsafe. Please don't do that. It's probably bad. I remember the first time I encountered that story was in Malcolm Harris's Palo Alto. And I was like, wait, this was a thing? What? Yeah, man. It's, uh... Yeah, I feel like the... As someone who, like Malcolm, is from Silicon Valley, I have a lot of, uh...

I have a lot of frustrations at the way that that is not treated as a geography and is sort of treated as an ideological condition Because when you really look at the history there some really incredible stuff Speaking of history I want to get back to talking about your book Yeah I think you do a good job of situating some of the roles of financialization in both the historical kind of build out of a lot of this stuff. I think you really, I really appreciated your discussion of ZERP as a part of the rise of cloud computing. I think that in some ways, the rise of something like an AWS is often treated as this sort of fait accompli. Like, well, obviously it serves this need, right? And it's, look, it is pretty good engineering. I'm not going to, like, you don't, you can't kind of pretend that it isn't. And the ability for there to be a market for AWS services initially really had to do with there being kind of a glut of startups.

What conditions even made it possible for there to be a bunch of companies that wanted to use AWS's services. Absolutely, right? You know, there's just so much money sloshing around, and it's like it needs to go somewhere. And so you have all these tech companies that are, like, lining up for the venture capital money. You know, they're lining up at the trough. You know, Mark Andreessen and all the boys are throwing out money at them. Like, the money is just unlimited, essentially. And, of course, you know, it's a very unique period, you know, kind of post-recession in particular, where the interest rates are just so low that it doesn't make sense to put your money into traditional investments because you're not going to get anything back on it. And instead, you should be throwing it at riskier bets and hoping that some of them pay off. And because of the way that the tech industry works and because of the way that some of these companies can really just take off on a global scale and start reaping massive profits because they can corner a market and then either through advertising or subscriptions or whatever kind of mode that they choose,

start making money hand over fist, that all these investors are jumping on that. And of course, even if they don't make money, the investors can still make a significant profit when the company goes public because for some reason these companies can be worth crazy amounts of money on the stock market even if they don't make any actual money. And of course, we're probably going to see that happen very soon with Anthropik or OpenAI. unless the bottom falls out of the AI industry before they get to go public, right? It's just, it makes no sense. Yeah, and there were so many windows, there are so many windows to recoup even when something isn't successful, right? Like Sam Altman's first startup, Looped, was acquired by another company that his venture, the venture capital firm that had invested in it, had also invested in, and it really, like, when you kind of look at the details of it, like there was really it was really just an effort to kind of tidy up a mess right like this company wasn't gonna actually succeed and they needed to kind of recoup the costs right and like sam walks

away with a tidy profit goes to my combinator the rest is hagiography um totally it also makes me think of like what was it confinity and and the initial x right and they both also had if i remember correctly the same investor who was just eventually like you're both doing online payments you need to merge and figure it out and make some money that way, right? Of course, creating PayPal and Elon Musk being very mad that it wasn't called X. And that's a whole history we don't need to get into. But yeah, I think it's a really interesting moment because, of course, Amazon is also like a pioneer of this model, right? Of don't go for the short-term profit. Don't try to turn a profit really quickly. Significantly expand your market share and what you're doing, and then you can make the money later and the investors will get paid off then. But let's go for domination first, and then the money will follow. And so Amazon moves into e-commerce, but also so many of these other areas. And part of that is, of course, cloud computing, because it needs that for itself,

but it also finds that there's a big opportunity to sell cloud computation, to basically centralize computation in one place, start to build out its own data centers. Initially, it's kind of leasing data center capacity from other companies, but then the Amazons, Microsoft, Googles, they all move into building their own data centers over time, though they still do lease a lot of capacity even to this day. And the whole cell is like, listen, you startups, you have something else that you're trying to do. You don't want to have to deal with spinning up servers and whatnot. Buy the servers from us, buy the capacity from us, then you don't need to worry about it. And you can just run your business and you can have it take off. And, you know, you can compete because if you are wasting time basically spinning off servers, then that's time that your competitors are improving their products and, you know, kind of beating you in the same market, right? And then I feel like on the kind of more institutional side, you know, the big companies that are already there, there is a slightly different cell where it's like this is not your core competency, you know, taking care of this computation, taking care of these servers.

so why should you be bothered with it when you can just pay us to do it for you? And plus, you know, this also offloads some responsibility and some risk that you are taking on yourself because we can do more kind of cybersecurity stuff and whatnot that you can't do by having this centralized. And, of course, you major companies have already been under this wave of outsourcing for a while now as you have been pushing call centers offshore and trying to offload parts of your business that are not core to subcontractors and things like that. Even going back to Apple, how they were really a pioneer of moving their manufacturing over to Asia for their computers and things like that. Companies are already doing this and Amazon or later Microsoft and Google can come in and say, hey, we are doing this for computation. Just give this to us, pay us to do it. Now you don't need to be worried about it anymore. and there are a load of other benefits that come of that. And, of course, the pitch initially is that you're going to save money. I think that pitch becomes more elusive over time.

But that's not because computation is so expensive. I would argue that is in part because, you know, you get on the cloud, but then there are all these other services that the cloud providers start to sell you, and that is how your bill really starts to creep up, right, because you need all these AI tools to, you know, understand your data and what have you, right? So, yeah, I do think it's an interesting moment. And the thing that really stands out to me, and part of the reason why I feel that that's an interesting thing to bring into the book, right, talking about the history of the cloud and the history of computation, is because I would argue, and I think that this is an accurate argument, that there would be no generative AI today had that kind of centralization of computation not started two decades ago. because without that amount of computation being held by a small number of companies, it would have been harder to train these models and make these AI products available in the way that the companies have done now. Yeah. You also wouldn't have had the same network or kind of world of companies, arguably.

Like an Uber doesn't happen without an AWS, right? Like, as you point out, AWS kind of does this, like, long game, like, lose a lot of money and then become the incumbent. And Uber has the same playbook. But, like, one of the reasons they were able to do that is because they could scale up quickly on platforms like AWS. Same with the other kind of rideshare companies. Same with Airbnb. Same with Palantir, man. And like there are, yeah, all of these, like the ability to make that leap is very much predicated on the availability of cheap compute that someone else's problem. While, you know, simultaneously, you know, lots of companies are getting money to go out there and fail fast from venture capital. Right. Like it's like a lot of other incentives aligned in this very nice way that that made AWS make a lot more sense. Like when it started in 2006, I mean, I have this weirdly distinct memory of hearing like an NPR story that's like Amazon's going to sell a computer server or something now, you know, and thinking like, oh, that's odd.

I'm 20. I don't know. But yeah, by 2008, 9, 10, like it makes a ton of sense. Right. Another kind of financializing and or role of kind of marketizing that you talk about that I wanted to make sure we talked a little bit about because it bothered me so much was was the use. I was the sort of like use of it in like greenwashing. Right. The like problem solving environmental impacts of of industry through inventing a market through which to kind of buy and sell like culpability. please share with your readers what is a water replenishment market? I wish we didn't even have to know you know I know it's so frustrating right because and you know people who have paid attention to you know climate fights and things like that for a while will not be unfamiliar with these things and how you know when we face a problem there are this

group of people that step in and say you know what is actually the most efficient way to address this problem. It's not to just change how we act, but it is to introduce a new way for us to buy and sell financial products in order to address the problem, and then the market will respond, right? Then the market will do it. That has worked out so well because, as we all know, emissions have dropped precipitously since carbon markets have come into place, which is me being very sarcastic because that is not how things are playing out. And it's like even just as a quick side note, like in Canada we're in the process of totally dismantling like the climate legislation, even shitty climate legislation that we had built up over the past decade. And the argument is that those mechanisms were too harsh and what we need to do is focus more on pricing because then the companies will act to reduce their emissions that way. And it's like, dude, we have played this game. We have done this. This is not working.

But anyway, to get to the point, if you look at these data centers, people will be familiar now about the fact that in some cases, they're using a ton of water. In other cases, they're using a load of fossil fuels. And increasingly, the data centers are using more and more fossil fuels. But of course, they don't want to appear as though they are bad guys, right? Or at least they haven't wanted to up until recently. I feel like they care less and less about being held to account for at least the amount of fossil fuels that they use. Certainly, they're trying to muddy the water on how much water that they use. But this is where these market mechanisms come in, right? And so a data center can use a load of fossil fuels, a load of gas power or coal power or something like that, or they can use just a ton of water in order to cool down the chips within the data center itself. And then if you look at the sustainability reports that the companies put out, it will look like, you know, they haven't used nearly as much fossil fuels or they haven't used nearly as much water because even though they're using the water and they're using the fossil

fuels, they're then going into these markets and buying credits to be able to claim that actually they replenished a bunch of water somewhere else. And so that negates the water that they actually use in one place. or they bought a load of carbon credits that like protects a forest or help maintain a renewable energy project somewhere. And so that means that even though they burned all those fossil fuels to power their data center, that in practice, you know, they offset those fossil fuels. And so they didn't actually do it, which is just like there's so many problems with that, right? Because on the one hand, the water is still used, the fossil fuels are still burned. And then there's the question of like, you know, in many cases, especially when we look at carbon markets, okay, you bought these offsets or you bought these credits or what have you. Would those actions not have been taken anyway, regardless of whether the offset was purchased or something like that? Or is the degree of benefit from purchasing that credit actually going to be maintained over time

and actually be reflected in the way that you report these things in your sustainability reports or whatnot? Because what we see time again in carbon markets is that they like the claim that they are significantly offsetting and reducing emissions with projects. For example, if you look at the tree-based products, the ones where they're supposedly protecting forests, that actually the benefit takes a long time to be realized because the tree doesn't just immediately pull the amount of carbon that you burned out of the air. It takes time for that to happen. And in many cases, the trees actually get either burned down by wildfires or get cut down within the time that it would have been needed to actually see that kind of benefit to the to the climate. So, yeah, anyway, I'm going on a bit too long about this, but it's all just like a complicated way of them trying to appear like they're not being nearly as bad as they actually are. And I think it comes back to that scale thing we were talking about earlier, because saying you've addressed, you know, all of the water use you did in Arizona because you did something that was good for like a watershed in, you know, Delaware.

people still live in Arizona and still need to drink water, right? Like, there is this, and it feels also weirdly of a piece of that sort of, like, utilitarian long-termist thinking where it's like, well, if only a billion people die, it's fine because, like, we're going to, you know, have tens of billions of people living in space or something, right? Like, it's really a misunderstanding of, like, what timescale you're trying to live on and how you interface with and relate to other human beings. Absolutely. Absolutely. And it's just really frustrating to see the way that they get away with it, right? And, you know, you're talking about water, but like on the carbon side of things, you know, I feel like a lot of people don't realize that they are spending a lot of money and expending a lot of lobbying power to actually shape the way that carbon accounting rules are being written so that they can do these deceptive practices and not be held to account for the actual amount of emissions

that they are emitting into the atmosphere, right? Especially as they have really thrown their climate targets to the side in order to pursue generative AI and the build-out of these massive data centers. I just think it's like a fundamental problem that they need to be held to account for, and they're getting away with it because, you know, again, we live under capitalism, and this is the way that things are supposed to work under capitalism we not allowed to really question what the capitalists say is the way that we supposed to address these problems even when they very clearly are not working And we passed 1 degrees now and emissions in the United States are going up again and plenty of countries are now saying that their climate targets are in jeopardy because there are so many gas-powered data centers being built out. I really just don't understand how we have allowed things to get to this point, all because, you know, AI is generating a lot of money and investment right now and all of our governments, like, need to get their piece of it or something.

To me, it almost reads as though it has to generate a lot of money, right? Like, you write a little bit about sort of the circular financing stuff and the sort of so much has been put into making it succeed that it's not even that this is how it's supposed to go. It's like, if it doesn't go this way, the capitalists don't know, don't have another move, right? Or they do have a move. It's the state bailing them out. It's also an interesting moment because also you write a little bit about sort of these efforts in the EU and other places towards trying to kind of like create a more like sovereign stack and like realizing like maybe, maybe depending on the superpower countries was a mistake. But that comes with its own sort of like messiness. It's not like a tidy sort of like, well, we'll solve the problems of undemocratic capitalist computer by doing democratic state computer. Right. Or also just our own version of undemocratic, you know, like problematic computer that exists everywhere else.

Like one of the really depressing things has just been to see how these conversations around like digital sovereignty and whatnot have been very effectively co-opted by industry, right, in many different parts of the world. and certainly in the EU, where it has become, we don't necessarily want to use the American technology and certainly not the Chinese technology. We're still going to have all the American companies still build out a load of data centers for some reason that doesn't really fit with the narrative, but we're not going to question that. But at the same time, you know, us becoming more sovereign and independent just means that we need to replicate everything that the American companies are doing on our own. So we need to build out a load of massive data centers. We need our own generative AI models. We need our own Palantir so that we're not using their Palantir. Is this really the move? Like, why can't we think on a much more fundamental level about the technologies we use and rely on and whether they're actually benefiting us and, you know,

start to try to chart a different way? And I would argue that the reason that that doesn't happen is because the money is all tied up in this way of doing technology and digital technology. And until that is like fundamentally broken or, you know, that our governments are given permission once again to not just pursue the thing that means more money all the time, but can actually take a bit of a longer view and recognize sometimes, you know, you make investments in something that don't provide an immediate return, but provide benefits in other ways that, you know, provide returns down the road in other means that, you know, kind of rethinking technology is just not something that is allowed to be on the agenda until that happens or something. I remember when you first kind of like, you know, did your, some personal news share that you were doing another book. The working title of it was Too Much Compute, which at the time I definitely read as Too Much Computer, which also like, that's a great. Which I love. Yeah. 100% too much computer. It can be, or I've found that it can be really hard to get people to like, actually get into a degrowth-y conversation or really think about like a moral economy of compute that like actually ask the question like, how much do you actually need? And what do we actually need it to do? And I think some of what makes people itchy about it is because there's a sense of like, constraint. I don't like constraint, right?

Nobody wants to be told that they can do less stuff. But also, like, to the extent people like that sort of libertarian mythic, like, mind bicycle computer dream exists, there's an assumption like, well, that means that I can do anything. So, like, why would you say that I can do fewer things? Right. I don't know if there's a real answer to this question. It is a hard problem, but I think in trying to make the case for a more democratic approach to how computation and computational resources are allocated on a societal level, how do you think through, I guess, the case for that beyond maybe also a sort of shaming thing? Because I feel like that also just has demonstrably not been particularly effective with most people. Totally. I think we definitely need to think of like a carrot rather than a stick approach, you know, to trying to get people to come along with what an alternative is going to look like or what it could be. And I do think that, you know, it's not that there isn't an answer to it, but I think that the answer to why it can be difficult to get people to reconsider their technology use is a very complex and nuanced one. Right.

And I think it depends in many cases on, you know, different motivations that people have. You know, I think part of it is that we have been sold this idea for the past few decades that this is more convenient, that this is improving our lives. And even when it starts to turn against doing that, you know, even when, say, the Facebook platform turns from, you know, a platform where you're seeing what your friends are doing and you're sharing stuff with them to one where, you know, you're being fed a lot of ads and you're being fed a lot of posts just to keep you engaged, that it is hard to pull back. Because on the one hand, you have a social network that is still kind of tied up in this platform that can be hard to get rid of, especially if your own social connections have frayed. But also that there is kind of an appealing part of using the platform. Like there is kind of, I would argue, an addictive part of it as well, right? Where there is something to seeing all of this engaging content. There's a reason that people watch it. you know, it's not just because they're like dumb or something like that, or, you know, they're,

they're being forced to buy a platform. It's because there, it is kind of responding to things I would argue in our minds that, that keeps us engaged, that keeps us coming back to it, that can make it difficult to fully get rid of and drop. Right. I wrote about this earlier this year. Like I used a ride hailing platform for the first time, which, you know, I, I had never done that before. It wasn't Uber, it was Bolt, actually. And I was kind of struck by just how quickly I became fearful of taxis, you know, against my own will, you know, because now I couldn't see exactly where it was at all times. And I could see how there was, how the user interface worked in order to convince me that this was the way I had to do things. And that, you know, I shouldn't be doing it in the way that I had been used to doing it and didn't have a problem, you know, kind of getting a taxi or getting a ride for decades, you know, up to this point. You know, there are benefits that people got from using technology in this way. Even as those benefits have eroded, it's hard to kind of pull back and change what you're

doing, that they have been designed in such a way to keep people part of it. And there is also a network effect where if everyone else is doing it, why are you going to kind of drop out that that's difficult to do, but also to explain. but then I also just think that people are in general really busy. They have a lot of things on their plate and, you know, expecting them to do a lot of work around how they use technology is I think a big expectation for a lot of people when they have jobs to work, they have bills to pay, they have kids to take care of. You're asking them to do like a difficult thing and dedicate their time to this when they don't have a lot of time. And so I think that's part of the reason why It needs to be something that people can actually be brought along with, right? That there is like a kind of positive vision of what this can be. You know, I do think that we do see elements of that already today. And I do see that tech companies, I think, are even recognizing this. You know, if you look at like even a dating app like Tinder, they are promoting more in-person things because people are getting tired of just being online and just swiping on the app.

And they want to be, you know, in places with one another. and so yeah I think that there's a response there but it's it's again still like a captured response to to what is occurring and so I don't have like the answer to how we get people off I do think that that is like an ongoing project that's going to take a lot of time but I also think that there's there's hope in this moment especially when we do see that people are really pissed off about AI I think people are also quite pissed off about social media and you know just kind of like how digital technology works more broadly. You know, I think that these movements against data centers and, you know, the fights against data centers have been really helpful, actually, in a moment where there's not a lot to be hopeful about on the political level in many cases. And that's not just in the United States, but in many other countries as well. And that there is the inkling of something there. The question is, how does that anger get used, right? And can it be kind of redirected into a positive vision of technology that works differently, that is more around a

social project to improve people's lives rather than just how do we keep people looking at screens? How do we keep them looking at ads? How do we sell them on the system that clearly isn't working for them, but that is making these companies a lot of money? And if I can say it, I feel like that is kind of what I want my next book to look into. And so when you talked about too much computer, I was like, maybe that's what I pitched my next book as. And then it gets like retitled as something else. But yeah, I don't know. I feel like we're in a moment and I feel like, again, in the way that I felt a couple of years ago that data centers were going to take off, I do feel like there is a move toward pushing back on this digital technological infrastructure that's been created for us and that's not working for us. And it's just the question is, what does that look like? How do you bring people along and get people involved in that and make it, again, a more democratic thing that people can be more engaged with and can actually be part of co-creating rather than just being kind of sold something else in the way that they were sold the system as well. And so, yeah, I think, I don't know, I think that in that sense, we're in kind of an interesting

time. And I definitely agree with you that the rapid push to build out more data centers has really forced this conversation in, I wish it didn't have to have happened, but maybe it's like, it's helpful that it's happening this way. I do, I mean, and you talk a little bit about this in the book, that, like, crypto in, like, the 2018-ish window did also start to really push this forward because it's a lot easier to persuade people that, like, crypto is useless because it is just, yo, you're turning gasoline into computer money. Like, this helps you. Hey, don't forget the ape pictures, you know? God, oh, I know. Well, I mean, the apes are gone. But I think that AI, generative AI and large language models, because of the both the novelty, the market desperation to make it the next big thing.

And certainly the like, yeah, the like wow factor of it has made it a lot harder to say like, oh, this compute is also maybe too much compute and doesn't really need to be happening the way that it's happening. Yeah, it's it's I have been joking to people for the last year, you know, like a decade ago, I was writing explainers that are like, this is what a data center is. right and like this is where they are this is like you should care about them and now you know 10 years on it's like everyone knows everyone hates them and that that's kind of rules i don't know i'm actually pretty into that totally i wonder if there's been traffic back to your old pieces people are like what are data centers and they're like yeah this person angry burrington they had to figure it out well i mean most of it's paywall now so i don't think that's the The Atlantic, yeah. Sure. Yeah. But yeah, I'm really excited to see how the book further kind of like pushes a lot of this like public discussion further.

And I think it does a really good job, especially kind of as you're wrapping it up and framing this as like being not just about the data centers, But obviously the data centers are kind of where where some of the more abstract aspects of these political debates become really, really tangible. And yeah, congratulations again. Thank you. I appreciate it. And, you know, obviously, thank you so much for like coming on the podcast to chat with me about it, because when I was thinking about people to talk to me like you were the person who so obviously like I wanted to talk to about the book. Because, again, I think that we're very, like, simpatico on how we think about these issues. And, you know, certainly I'm interested in your perspective on this as well. And, of course, you know, we should say that your book, you know, was one of the books that I look to in understanding these infrastructures has been reissued. And I know there's not many copies left, but people should try and go and grab one.

Yeah, I think I may have to do a second printing. Perfect. Good. Because, yeah, it turns out people still want to know how the Internet works, even though they already know what a data center is. So that's nice. How dare they? How dare they? Yeah. But we can put the link to that in the show notes as well so people can go get a copy. Ingrid, just want to thank you again for coming on Tech Won't Save Us. And this certainly won't be the only time that Tech Won't Save Us listeners hear from you. We'll have to have you back on the show next year. Thanks so much. I really appreciate it. Well, yeah, thanks. This was a real treat. Thanks for joining us on this special episode of Tech Won't Save Us. Ingrid Burrington, who interviewed me in this episode, is the author of Networks of New York and host of the Rip Corp podcast. Tech Won't Save Us is made in partnership with The Nation magazine and is hosted by me, Paris Marks. Production is by Kylie Hewson. Tech Won't Save Us relies on the support of listeners like you to keep providing critical perspectives on the tech industry. You can join hundreds of other supporters by going to patreon.com slash tech won't save us and making a pledge of your own.

You can also pick up a copy of Hyperscale by going to hyperscalebook.com to get more information. Thanks for listening. Make sure to come back next week.