Hyperscale: Paris Marx on the Ambition and Excess of Big Tech’s Data Empires

00:00:31 Andrew Keen: Hello, everybody. It's October 20, a Tuesday, 2026. As the year begins to get to its end, we're on the Financial Times best business books of the year. They put out their long list, and one of the books on it is a particularly intriguing one. It's out today. It's called Hyperscale: Ambition, Power, and the Relentless Rise of Data Empires. It's by my guest, Paris Marx, a writer, podcaster based in St. John's, Newfoundland, and he's joining us today. Congratulations, Paris, on the new book. He's already getting great buzz. I have to ask you. I, of course, went to Gemini to find out what hyperscale means, although I guess that might be a bit of a contradiction in terms since Gemini is owned by Google, the ultimate hyperscaler. What does the word mean?

00:01:26 Paris Marx: Yeah. So it— it's really about the scale, like, the size. Right? You know, especially when we're talking about data centers, it's data centers that are of a particular scale. You know, they— we— we're used to data centers. We've had them for decades. But when we're talking about hyperscale data centers, we're talking about these ones that, you know, because of the centralization of computation that we've had for the past couple of decades, particularly with the cloud providers, you know, these data centers have just become so large. And for me, the thing that stands out about hyperscale computing and hyperscale data centers is the demands that they make on the communities and, you know, the places where they are located. And so that is why, you know, we've been having these discussions for the past couple years about the amount of electricity, the amount of water, you know, just kind of the local impacts that can come, of these hyperscale data centers.

00:02:14 Andrew Keen: People often use the term Paris, big tech, but is hyperscale then a better word to describe, quote, unquote, big tech? And I— I'm guessing that the companies in particular are the big cloud companies like, Amazon, of course, with their web services, Google, who own Gemini, their cloud platform, and Microsoft Azure. So they're different from the classic AI companies that people seem to be obsessed with, the frontier companies like Anthropic and OpenAI?

00:02:47 Paris Marx: Absolutely. Like, there can certainly be overlap there, you know, because, obviously, Google and these other companies are working on AI models and AI products themselves. But when we're talking about, you know, hyperscalers, we're really talking about the data center side of things. Right? So if we're thinking about big tech, that can refer to a lot of different companies that, you know, do different things, and they are large in scale. You know, Apple would be part of that as well even though they wouldn't really be, you know, one of these massive kind of, companies building out a ton of hyperscale data centers in the way that we see with Amazon, Microsoft, and Google in particular, but Meta is also getting into that game, or has been in a significant way in the past few years. So, yeah, it really refers to just the kind of scale of computation that's being built out. And, certainly, we do see, you know, OpenAI and Anthropic building some of their own data centers, and those would be hyperscale data centers. But when you think of just the amount of data centers they have, the amount of computation that would be at their disposal that they kind of directly control, it would be, you know, significantly less than what we see with those dominant cloud companies, the Amazon, Microsoft, and Google, in particular. Though, again, Meta is kinda really moving in there too in the past.

00:03:55 Andrew Keen: Yeah. I forgot about our friend Mark Zuckerberg and Meta. And I'm guessing also you— you're not a big fan of Elon Musk. You've written a lot of stuff about him. He's not the most popular person, I think, in the world these days, particularly on shows like this. You read about Elon Musk being convinced that he's the future. I assume that SpaceX also is a weird kind of hybrid somewhere between the frontier companies and the classic hyperscalers.

00:04:23 Paris Marx: Absolutely. It's kind of fascinating how SpaceX over the course of, like, a year has been transformed from a space company to an AI company as xAI has been integrated. And, of course, when we saw when, SpaceX, went public, you know, the massive part of the kind of future, revenue projections were based around the AI potential of xAI and things like that. Right? And I think already, those predictions don't seem to be playing out, as I think many people could've expected. Right? But, obviously, certainly, Elon Musk has, you know, he has long been in this game as one of the early kind of cofounders and funders of OpenAI, of course, was, you know, kinda left that project or pushed out depending on who you talk to about that. But as generative AI has taken off, he has really sought to make himself part of that conversation and part of that discussion, which is why we have the founding of xAI a few years ago, and his push to build out some really large data centers, you know, in Mississippi and in Tennessee in particular, which, you know, as much of the reporting has shown has had some really significant consequences for the communities, around them, you know, in particular in terms of the local air quality and local impacts that can come with the burning, you know, of all the fossil fuels, from those generators powering them. Right?

00:05:39 Andrew Keen: Yeah. And you've written extensively, as I said, on Musk. You wrote that his Hyperloop was always a scam. I think most of us knew that. Although, one of the things you can say about Musk is whether or not the Hyperloop was a scam. He knows all about speed. And he may not have built Hyperloop, but he moves as fast as anyone to create SpaceX, as a hyperscaled company.

00:06:05 Paris Marx: Oh, absolutely. You can see that with his data centers too. Right? Like, just how quickly—

00:06:08 Andrew Keen: He's always on the exciting edge. I mean— and, of course

00:06:11 Paris Marx: Totally.

00:06:12 Andrew Keen: There's also the issue he's been promising, I guess, hyperscale, in the universe or in space, which is another abstraction which may or may not be real.

00:06:23 Paris Marx: I'm certainly skeptical of the data centers in space.

00:06:27 Andrew Keen: Maybe we can get to that later, Paris. So much to talk about. The subtitle of your book, is ambition, power, and the relentless rise of data empires. You're not the first person to write about the imperial quality of this AI— AI revolution. Karen Hao, who I'm sure— whose work you know, I'm sure you know her as well, was on the show a year or two ago. Her book on— I think it's called AI empires [as spoken: Empire of AI] began this kind of discussion. What's the imperial quality then of hyperscale and hyperscaling, Paris?

00:07:03 Paris Marx: Yeah. And, certainly, you know, I quote Karen in the book. Right? So her work has certainly, you know, inspired my work on that and my understanding of what we're seeing with, you know, these major tech companies. You know? She does fantastic work in her book, Empire of AI is great, and I would certainly recommend it to everyone. But, yeah, I think that the key thing here is that we really see these tech companies significantly expanding their power in this moment. Right? You know, essentially, we have not just the build out of these major data centers, but we can really see the way that the tech companies are flexing their muscle, you know, both in kind of transforming the way that we collectively live. But beyond that, we can also see, you know, how they are increasingly influencing the political system. Right? And certainly by pushing different technological tools into the way that we all live and the way that our societies function also has the effect of increasing the power that they have over, you know, virtually everything that happens in our societies. And so, you know, I do think that we need to be paying more attention to that. And certainly, as they're building out all of this computation, that comes with power too. Right? You know, they are getting governments to go along with what their vision of the future is going to be like, you know, subsidizing it, making sure it has the political and kind of regulatory space that it needs in order to expand. And, you know, in the process, you know, if we think about what the future of our society and kind of how we all live is going to be, often that is not a discussion that we are collectively having anymore. It is one that is being kind of decided and determined by these tech companies and then kinda given to the rest of us regardless of whether that serves us or not. And so I do think that there is kind of a major way that these companies and these executives, are kind of flexing an imperial muscle, with the control that they can carry out, not just, you know, within their companies, but increasingly on a global scale.

00:09:03 Andrew Keen: Yeah. And the thing— and we've talked about this a lot on the show, Paris, and I know you've given a lot of thought to it as well. It's the financing of these data centers that's so astonishing. The FT, which tends to be very good on these subjects, had a piece recently on the multiplying risks of financing data centers. These are mostly not government finance. They're financed privately, particularly by big tech companies, by the hyperscalers themselves who are selling bonds. Companies like Google. The FT are forecasting that $7 trillion will be poured into, data centers, by 2030. It's an astonishing number. To what extent in your book and in your thinking are the hyperscalers alternatives to government or working independently of government? And in many senses now, given the nature of this revolution and the economic power, technological power goes with it, are actually dwarfing government.

00:10:05 Paris Marx: Yeah. In many cases. Right? You know, we can just see the scale of these companies to see how they can compete. But then, also, so many governments are dependent on the services, you know, the computing power that they offer. Right? And that also gives them a lot of leverage, you know, which we saw kind of wielded last year when the United States sanctioned the International Criminal Court, and all of a sudden they lost access to, you know, their Microsoft email addresses and services and things like that. Right? Not— and other ones as well. You know, the dependence on these services has immense power, both for the US government because it can wield, you know, access to them, but also for the companies themselves and how, you know, they kind of deal with the US government but also other governments, around the world. And, you know, to pick up on the broader point there is a real way that these companies are flexing their muscle in this moment, and we need to be aware, of what is actually happening, of how that is playing out. And, you know, the financing piece is very important because it is what is enabling all this to happen. But, also, as we have these discussions of, you know, whether there is an AI bubble or whether the financing makes sense or whether there is gonna be a market for all this computation, you know, as more investors are dependent on or own some degree of exposure to the data centers or their debt, you know, that means that if things don't work out as planned or as the data center companies or the AI companies are, you know, leading us to believe, that means that the people who are going to be affected by the fallout are gonna be so much greater. You know? I like to compare it to the crypto industry. Right? You know, when the crypto moment or, you know, kind of all the investment that was into it in the kinda 2020, 2021 era started to collapse, it was— certainly people felt it, and you had people who had bought into it and, you know, kinda lost money, through various crypto scams and things like that. But in terms of investing in the companies, you know, it was pretty limited as to who was actually affected by that. You know, if the kinda AI bubble or if kinda the excitement about AI, doesn't play out in the way that it— that happens or if there's not a market for all these data centers, a lot more people are going to feel the effect of that in their pensions, in their 401(k)s, you know, in their investments, and what have you. Or even just more broadly, like, with the kind of economic contraction that would come of this not all working out potentially lose their jobs in that way, not their— not the AI taking their job, but just the economic damage that would come as part of it. And so, you know, I think that one, this is really worrying. I do think that we're seeing the kinda AI and data center market supported in a way— or kind of, like, lifted up and kept going in a way that isn't connected to the actual financials that we're seeing from these companies. You know, when we see the amount of losses that say OpenAI and Anthropic are reporting. But then the other piece of that is even some of the data center companies or the companies who have gotten really big on data centers in the past few years, you know, Oracle stands out as having really moved into this market. You know, there are real questions over their buy— viability and the bet that they have made on all this working out.

00:13:10 Andrew Keen: Yeah. It's fascinating. History, of course, never quite repeats itself, but it does rhyme. People are often thinking what kind of global crash might come. People often go back to the Great Crash of '29, or they go back to the, Great Recession of 2008 where the banks were supposedly too big to fail, so the government came in and bailed them out. I wonder— and, again, I'm thinking aloud, Paris. If there was a bursting of this— it's not a bubble, this irrational exuberance, which as you note, is enormously large and way more significant than crypto. Crypto is a flea on this l— in comparison to the elephant of, if that's the right metaphor of data centers and hyperscale. You would assume that the government, and Trump's already been articulating this in his own bizarre ways, that the hyperscalers, Google and Amazon and Microsoft, particularly in Oracle given Trump's relationship with Larry Ellison, would be seen as too big to fail, which could trigger the next global economic crisis.

00:14:30 Paris Marx: Yeah. It's— it would be fascinating to watch depending on how it all plays out. Right? Because I think on the one hand, you can see that the AI industry has been really influential with the Trump administration, and it has been getting a lot of things that it wants, you know, through the administration, the support for data centers, the support for, you know, their vision of AI that has come from Trump and the people around him. But we have also seen that, you know, what was it earlier this year or late last year, someone at OpenAI kind of floated the idea that the government might step in [ed.: the "backstop" remark came from OpenAI CFO Sarah Friar in November 2025] to help them out if things don't play out as planned. Sam, I think, was—

00:15:06 Andrew Keen: One of the Sam-isms.

00:15:09 Paris Marx: Yeah. And there was, like, an immediate backlash to the notion that that happened such that, you know, OpenAI had to come in and kinda— or had to come out and, you know, kinda backtrack on what they were saying. Right? And so, you know, one, I do think that there would be a pressure to do that. I think we would have to see how it actually played out in practice, you know, whether the government could really get away with, bailing out these companies in terms of the public mood, especially around AI and data center companies and the tech companies more widely at the moment. But the other question is, you know, how, like, a collapse or, you know, a kind of decline in the kind of viability of these investments would actually look. Right? You know, would it be kind of like a sudden contraction, or would it be kind of, you know, a kind of slow, drawdown of, you know, investment in data centers, you know, maybe some closures of data centers, that sort of a thing, really depends, I think, how that would actually play out in practice. You know, I think it's entirely possible that there's so much in this right now that if there was gonna be a contraction, it would kinda be like people would quickly pull their money out. But it is hard to say for sure. And exactly how that would play out would also, I think, in part determine the, you know, scale of the response.

00:16:27 Andrew Keen: Yeah. We did a show recently with a Financial Times, author. They'd— actually, the editor of Alphaville, Robin Wigglesworth, who has a history of the bond market [ed.: A Fabulous Debt]. It's a fascinating book. I mean, it's a little dry, a bit technical for some people. But he argues that, while stock market crashes aren't great, the real crisis will come with a bond crash. And, of course, what's so important about the hyperscale revolution that you're writing about is it's being funded by bonds, but not by government bonds, but by these big tech bonds. So I'm not entirely sure how that could all work out too.

00:17:07 Paris Marx: Yeah. Which is also quite unique, right, and kind of increases one, the exposure of the major tech companies, but also, you know, investors to this debt. Because, of course, you know, as people will likely know that these tech companies tend to have just a lot of cash. Right? You know, they make a lot of money. They don't spend all of that money. They typically have some big reserves that they rely on. But when we look at the major tech companies, they're spending so much right now on data centers, that they have been putting out these bonds, you know, basically to finance their data center construction. And in part that is, you know, if you listen to some peep— some analysts of how this is working in part a way to try to, you know, hide some of the scale of how much they're actually spending on the data centers at the moment. And so— yeah. Like, if this collapses or if the data centers don't work out as planned, there is one, more exposure for the tech companies, right, because they have more money on the line building out all this infrastructure than, you know, as they would in the past, more likely to be kinda creating digital products that, you know, don't require the same degree of physical kind of tangible investment. But, you know, also all of these investors, all these banks, all these, you know, pension funds, all this kind of stuff who have put money into this expecting a return. And, you know, if that return doesn't come, then they're all gonna be on the hook for that.

00:18:25 Andrew Keen: So, essentially, the future is riding on this. You've written quite critically. Again, you're not unusual in techno— in what you call the religion of techno-optimism of guys like Marc Andreessen, who's particularly obnoxious, I think, and thin-skinned. I mean, is there an argument, Paris, that these guys are betting on the future? That's the nature of capitalism. It requires massive risk. And clearly, for better or worse, the future is AI. So there's a reasonable chance— I mean, it's impossible to know, obviously. But there is a reasonable chance that this bet will pay off, which will make them even more powerful, even more successful. So I guess from the point of view of a critic like yourself, what's worse, the success of the hyper— the hyperscale revolution or its failure?

00:19:20 Paris Marx: Oh, I would probably say its success, honestly. I feel like we probably need its failure so that we can, try to rethink what has been happening here. Right? Like, if the scale of a crash of even, like, you know, the dot-com boom and bust happens, like, does that give us an opportunity?

00:19:36 Andrew Keen: But that would be minor because the dot-com boom and bust, of course, again, was of a— I mean, not of a handful of companies, but it didn't create a real crisis in the economy as this would.

00:19:49 Paris Marx: Yeah. And we can already see that, you know, the scale of investment tied up on— in this is, like, far greater than, you know, the dot-com crash. Right? But, you know, I guess to get to the core of your question, I think I would take the crash over the success because the crash gives us an opportunity to, you know, rethink and reassess in a way that the success wouldn't. Right? And that is not to say that I think that, you know, AI technologies or generative AI specifically is, like, completely going away even if there is a crash. The question is, I think there's a few questions. Right? You know, you said the tech industry is betting on the future. I would say that they're betting on a specific vision of what the future is and should be. Right? And it's a vision that gives them a lot more power and authority over our societies, the way that we live, you know, all these sorts of things.

00:20:38 Andrew Keen: Yeah. But let me just jump in here. I mean, I'm not at any— I'm far from being a defender of these people. But I think their goal is not control of the society. Their goal is profit. I mean, these are private enterprises that determined by profitability rather than control.

00:20:56 Paris Marx: I would say— I would argue it depends on who you talk about. So I definitely think that there are some people who are much more interested in, you know, the project of how much money am I making and how do I make more money, and that is the key to what they're doing. I think you have other people who do have broader ambitions and more ideological goals for what they are pursuing. You know, you can look at someone like Peter Thiel, for example, who, you know, certainly likes to make a lot of money, but I also think he has a particular vision of the type of society that he wants us all to be living in. And in part, his investments are to try to realize that and certainly his actions and that sort of a thing, which is not to say that he doesn't just invest for profit as well. He certainly does. Right? But then I would say, you know, if you look further on that and you look at the writings of people like Sam Altman and some of these other people, they also have a very specific vision for what the society that we should all live in the future should be. You know? And that is a society where, you know, they build the computer that can think for itself, and that eventually we merge our brains with the computer. And, you know, that we kind of create simulations of people within these massive, you know, data centers or, you know, computers of the future, whatever those are gonna be. And I would say it's a very science fictional vision, but that science fictional vision also kind of shapes the type of technology that they try to develop because they are trying to realize that particular future that is oriented toward what they're interested in, what their goals are, but might not be aligned with what actually most people in society think our future should look like. Right? You know, I would argue that the reason that we have generative AI in this moment rather than another form of AI or another take on AI, is because they are trying to create these models that can, you know, respond to any query that can do anything that has all this kind of knowledge in it because they see that as the stepping stone to the creation of, again, you know, the replication of the human mind within the computer. Not necessarily that that is the best way or the most efficient way or the most socially conscious way to pursue AI development, but because it is the one that's aligned with, again, these kind of science fictional priorities.

00:22:59 Andrew Keen: Yeah. And you've written that, or you've argued that hyperscale data centers actually pose a greater threat than, quote, unquote, AI science fiction. And, of course, you didn't mention him, but there's our friend, Elon Musk. You've written about it, and he's been in the news recently spending huge amounts of money investing in the global extreme right. So I think I'm not sure on Thiel. I mean, Thiel is the classic Bond villain. Nobody likes him. I'm not sure how much power. I mean, he's a very, very wealthy man, but he doesn't have quite the power of Musk. And he hasn't been quite as explicit investing in the global extreme right or coming out so explicitly in support of one kind of racist group or another. So I think Musk is really particularly chilling in this sense.

00:23:49 Paris Marx: I think it's, like, I think it acts in different ways. Right? Like, Peter Thiel was there very early on in the Trump project in helping it move forward. And we have certainly seen him, you know, basically create a network of people, that are more aligned with his right-wing view of politics. And then those people kind of act to move those things forward. Right? So it doesn't need to be Thiel who is kind of directly out there, in a very kind of open way trying to move these things forward. Whereas Musk, I think, is someone who just kinda craves attention and the limelight and for people to be paying attention to him.

00:24:23 Andrew Keen: Like you know who. Right?

00:24:25 Paris Marx: Exactly. So he acts in a way that is much more like, you know, trying to be out there. He wants people to see him doing things. He wants people to be writing about him and commenting about him even if it is in the negative, because he is— you know, that— that's just the type of person he is, which I think is much different than Thiel in, you know, many ways. Right? So which is not to say that they don't kind of impact this in their own means. You know, obviously, Thiel has been much more engaged in pushing these right-wing ideas within the tech industry for much longer than Elon Musk has, I would argue. But, you know, Elon Musk also has a much deeper, you know, kind of pool of resources to draw on in order to, you know, basically help to realize these things. Right? And so we've seen that over the past couple of years, him really kind of explicitly back right-wing movements, use his social media platform in order to, you know, give them extra energy. But also, of course, because of the nature of the US political system put a lot of money, into funding, these groups and, you know, basically, the Republican Party in general to help it succeed.

00:25:34 Andrew Keen: The one company that we haven't talked about yet, I mean, I'm curious as to your take is the— is the most valuable company currently in the world. I think it's valued at about $5.5 trillion. Jensen Huang's, Nvidia, he's been in the news recently. He clearly has the eye and ear of Donald Trump. How does Nvidia play in the high— Hyperscale economy? Are they a hyperscaler, or are they the, if you like, the Levi's of this particular gold rush, they're simply supplying the hyperscalers with their chips.

00:26:10 Paris Marx: Yeah. The— definitely more of the latter. Right? You know, there would be no AI moment as we have it today without Nvidia's chips and its graphics cards and that sort of a thing. You know? And, obviously, we understand that with how much that company has just exploded and taken off, since the generative AI moment began. Right? Because, its graphics cards are, you know, are central to it. And not just in the past few years, like, you know, it has developed not just the graphic cards, but also kind of the software infrastructure, for a lot of modern AI development, which, you know, I think a lot of people didn't realize until a few years ago. And so now we have these data centers just stuffed full, of these Nvidia chips. It has really had to scale up the production that it does. That has made, you know, Nvidia an even more powerful company, but it has also made Jensen Huang very powerful and very wealthy at the same time. I feel like for them, I feel like they have— they are riding a wave, you know, that they just so happen to be—

00:27:11 Andrew Keen: Like a [unclear]— particular boom.

00:27:15 Paris Marx: Yeah. And— which also, I think, also puts them in a very risky position. Right? You know, they have made a lot of money. They have seen their value really take off in the past few years. But if things don't work out, they don't have a lot, you know, of other kind of revenue streams or whatnot in the way that a Google or a Meta or an Amazon would to fall back on. You know, it's part of the reason that, you know, when Michael Burry, you know, the man— the investor who made a lot of money on the '08, '09 crash, when he kind of bay— basically betted against the financial system and betted correctly, you know, when he was looking for a company to short and a company that would be, you know, kind of the most exposed to this generative AI moment bubble, whatever you wanna call it, that he chose to basically identify Nvidia because it does not have those other revenue streams. So much of the revenue that is— has made the past few years is dependent on generative AI being, you know, the next transformational thing that these tech companies say it's going to be. And if that doesn't play out or if there is kind of a reduced demand for computation and for its graphics cards as a result, then that means that, you know, a ton of its market share is just gonna dry up, and then it's, you know, gonna come down in value very significantly.

00:28:29 Andrew Keen: Just like Cisco. Although Huang was in the news recently, gave an interesting interview to Ezra Klein, New York Times, in which, at least in the way I listened to the interview, he was throwing Anthropic and OpenAI under the bus saying, look. If your products are that dangerous, you simply shouldn't be shipping them in the context of the Hugging— the Hugging Face case, associated with, you know, AI supposedly being able to somehow— Bots being able to, if not think for themselves, act for themselves. I wonder, stepping back slightly from that, Paris, whether— the media these days is obsessed with the coming IPOs of OpenAI and Anthropic and Sam Altman and Anthropic's Dario Amodei. There was a recent interesting podcast on Anthropic's quest to give AI morals. Who knows really what that means? A lot of Anthropic and OpenAI people are quitting, giving all sorts of interesting interviews, confessions to the media. But I wonder, stepping back from all this, whether it's in the interest of the bigger players like Nvidia, Google, Amazon, even Microsoft, to kind of throw Anthropic and OpenAI under the bus. Because the hyperscale revolution isn't riding on Anthropic or OpenAI. They're just the frontier models. And maybe even, it seems clear that Nvidia benefits as much from the Chinese, models. So would it be fair to say that people who aren't really following this as carefully as somebody like yourself misunderstands the nature of this revolution. It isn't being driven by the OpenAIs and Anthropics of the world.

00:30:21 Paris Marx: Yeah. I would largely agree with that. You know, I always remember that in the months after ChatGPT was released, you know, end of November 2022, there was a real discussion that, you know, now Google was in jeopardy and its search dominance was in jeopardy.

00:30:36 Andrew Keen: Yeah.

00:30:36 Paris Marx: And, you know, basically, this new company was gonna come and kind of, like, eat its lunch. Right? And what we've really saw was, I would say, basically, permission for Google to release its AI models. You know, it had been working on this stuff. It just didn't feel that it could release it in the way that OpenAI released ChatGPT because, you know, it had regulators watching what it was doing. But once another company had released this, had put it out into the world, it could, you know, kinda jump on the bandwagon, do the same thing. And, of course, it also owns all the data center capacity and has been building out so much more the past few years. And so I often see, you know, the kind of actions or the role of these more dedicated, as you're saying, you know, frontier kind of AI companies, the OpenAIs and Anthropics in particular, as giving permission and, you know, kind of, allowing the major tech companies to move along and to do this as well. Now we do know that Microsoft is very reliant on OpenAI's technology in particular. So, you know, maybe it wouldn't be super beneficial to them if everything else kinda collapsed. The other piece, you know, on the Chinese models, it's interesting because they are often less computationally intensive. But, of course, you know, if they're less computationally intensive, you'll probably have a greater use as well, which would make up for the, you know, kind of relative decline in, you know, kind of per model computationally use if you wanna put it that way. Right? So overall, you'd have more because more people would be using them. And, you know, that can be the Chinese ones. There are also other kind of open-weight models that people could use there as well. But largely, I think that, you know, even if we had OpenAI and Anthropic implode, that, you know, companies like Google in particular would benefit immensely from that. And especially as you see greater scrutiny on those companies, like, you can see it in multiple ways. On one part, the scrutiny certainly gives those companies a lot of attention. You know, that can be seen as negative attention, but it also, I would argue, makes it appear as though the products being offered by Anthropic and OpenAI are, you know, incredibly powerful, you know, more powerful than they are in reality. And, you know, that kind of negative narrative, the fear narrative works in that way as well, especially as they are kinda pre-IPO, and trying to make it look like they are building this incredibly powerful technology that maybe even thinks for itself and, you know, we can continue on with that narrative. I think it's less in the interest of, say, a Google to make it look like that is what they're doing, which is also not to suggest that, you know, they're kind of like the ethical, generative AI alternative. I think largely what we need in this space is, a lot more regulation and even just beyond the regulation, but, like, assessment of what these, chatbots and image generation technologies are really being used for, the impacts that they're having on society, so that they can be properly understood and, again, you know, potentially regulated and reined in if that, is deemed necessary in those cases. But until we actually, like, understand it properly, it's hard to, I would say, kind of, take apart the narratives of the AI companies and the larger tech companies specifically, to say, listen. Maybe they're distracting us or kinda they're putting us on to, you know, kind of some fantastical vision of what this technology can be when we see more and more stories about the real impact of the generative AI technology now that we probably should be addressing rather than talking about something that might happen in ten years that, you know, personally, I would see as, you know, very unlikely to actually occur.

00:34:11 Andrew Keen: Yeah. Dario and, Sam would probably say we don't understand it. So— and, of course, the— I mean, the Google story is really interesting because a lot of the power of its stock and the fact that these your hyperscalers are driving the Wall Street boom at the moment is built on the fact that in their quarterly reports, they're always claiming profitability from their investments in the frontier company. So if they crash, then it wouldn't necessarily impact that well on Google stock. So, Paris, you're a man, and, you know, correct me if I'm wrong, but you're a man of the left. The book comes with,

00:34:53 Paris Marx: That's accurate.

00:34:54 Andrew Keen: It comes with blurbs from people like Naomi Klein, other major figures on the left. What do you make of this, way in which the populists, both on the right and the left— I don't know if you would define yourself as a populist. I don't even know if you're comfortable with that word. But what do you make in particular of the Bernie Sanders, Steve Bannon alliance. Maybe it's an unholy one on data centers and the hyperscalers. I'm not sure either of them quite understand what's happening, certainly not with the sophistication that you do. But are you a little nervous of the way in which the radical right in particular seems to have embraced some of the arguments you're making in hyperscale?

00:35:39 Paris Marx: Yeah. Absolutely. Like, there's no— well, it— I would say it depends. Right? Because we have seen, arguments that are critical of the tech industry and these technologies from the radical right for quite some time. Right? You know, Bannon was uncomfortable with these major tech companies, you know, going back many years. You know, I haven't listened to his show, but I know I read Naomi Klein's book, Doppelganger, where she describes, you know, having watched it quite a bit. And, you know, apparently, they are very, you know, motivated by in a negative way, you know, these visions of, like, transhumanism and things like that as well. Right? You know, that we hear from some of the major tech people like Sam Altman, for example. But, you know, that to the side, I guess, or considering that, I'm really uncomfortable seeing Bernie Sanders kind of appearing alongside Steve Bannon at a conference or, you know, even seeing that there can be some kind of allyship with leaders of, the radical right there. And I'm even uncomfortable with some of the positions that Bernie Sanders has taken on these issues. You know, I very much welcome his advocacy of, a national moratorium on data centers in the United States. Not to say that we don't need any data centers or we shouldn't build any data centers, but I do think that we need, like, a proper regulatory and planning framework and environmental framework in place, before further development, you know, occurs. Right? And, also, just that, you know, and this is not necessarily what Bernie is proposing, but I think it's about time that we actually have a conversation about, you know, what resources we're dedicating to these technologies and to this computation and whether that is justified for the products that we're getting. You know, I would say that we're sacrificing a lot of resources. We're burning a lot of fossil fuels for these data centers, ostensibly to power generative AI. But is generative AI even delivering on the promises that the tech companies have made around, you know, efficiency, productivity, all this kind of stuff? I would say no. But we're getting a lot of negative effects on the other side.

00:37:36 Andrew Keen: But it's very— I mean, there have been so many reports, so much, research on this, and it's still not entirely clear. It's not clear if it's either yes or no. And it— the reality, for better or worse, Paris probably is a mix of both, isn't it?

00:37:51 Paris Marx: I think it's always a mix. Right? But we can still kinda try to gauge out the net. Right? You know, is it a net positive? Is it a net negative? Because I feel like we did have a lot of promises from the tech companies, you know, not all of which have panned out. But then there have also been a lot of drawbacks that have come with these technologies that they didn't tell us would occur, but that are causing more and more, I would say, social harm. And certainly, if you look at the data centers, environmental harm as well. But just to pick up on what you were saying about Bernie Sanders, I feel like part of the problem is there as well is that, I think that Bernie Sanders listens a bit too much to what the tech executives say the technology is and the harms it might have rather than looking at the actual harms that we're seeing. Right? So he's focused on the impacts on workers and, you know, kind of these broader threats about what AI might mean for humanity. But I would say, you know, what about the deepfakes? You know, what about the reports that we're seeing on cognition and, you know, people's ability to kind of, you know, think critically and engage with the, information that they're consuming? Or, you know, the— where we're seeing the cases of people having mental health spirals after developing dependencies, on these chatbots. What about all the fake information that is being generated and further polluting the information environment in a negative way? You know, and we can go beyond that to how they're being rolled out in public services and all this kind of stuff. Like, I think that there are a lot of issues here that have come with these technologies that we should be having regulators talk about, that we should be having politicians talk about. But I don't think that Bernie Sanders, unfortunately, is having the right conversations and pushing that in the right direction. You know, and, again, you say I'm on the left, but, and I would also call myself on the left to be clear, but that doesn't mean I'm fully aligned with, you know, and I'm not just gonna kind of uncritically back, you know, what these people are doing. You know, I think Bernie Sanders has done a lot of great things, but I think he's really offside with where he should be on the, AI discussion, unfortunately.

00:39:47 Andrew Keen: You're also the coauthor of a very different kind of publication from Hyperscale, a white paper on reclaiming digital sovereignty. There was a recent report, and New York became the first state to pause hyperscale data center permits. This seems to be the eye of the storm, the data center issues. What are you suggesting then, in the book when it comes to data centers? Of course, there is the argument made by many people, and it makes some sense that why would a local community, a poor community in Kentucky or Tennessee or elsewhere, other states which aren't particularly wealthy, why would they turn down the opportunity to build data centers which would create wealth, jobs, prosperity, certainly better than coal mining? So what— what's your position on data centers? And are they the eye of the storm? I mean, data centers become this term now that people throw around. I think half the people who— or more than half the people who use it don't even know what they are and don't quite understand how they connect to the bigger issue of hyper— of the hyperscale revolution that you write about in the book. But are they the eye of the storm, Paris?

00:41:04 Paris Marx: Yeah. Definitely. You know, I think if you talk to some of these communities, they might say at least the coal mine created jobs. Right? You know, the data centers, unfortunately, don't create a lot of jobs, and a lot of these data centers get a lot of tax benefits and tax incentives that make it so that they won't actually be contributing much to the community for a very long time. You know, we're starting to see some of that overturn now and some of that change, but for, you know, for quite some time, how this worked, was that the data centers would enter and would expect a big tax package, in order to set up the data center, in that community. Right? But I really see— I think there's a lot of hope in the opposition that we're seeing to data centers. Right? In that people are getting motivated around this specific issue. You have people, you know, again, as you were saying earlier, from the left, but also from the right, and, you know, from right across the political spectrum coming together on this. This is not just like, you know, a bunch of lefties who are getting incensed or even a bunch of people who are on the right wing getting mad. You know, there's a lot of issues bringing people together over this, you know, specific problem. And I do think that that creates an opportunity to have more conversations, not just about what a data center means for your community, but about the kind of tech industry and set of technologies that we have become reliant on for the past number of decades. You know? I think that the tech industry clearly sold us a promise for what digital technology and the digital revolution was going to be, and I think that they delivered on part of that promise for quite some time. Right? There were very clear benefits and still are, of these digital technologies. But I also think that those benefits have eroded over time, especially over the past decade, as there has been this greater push to extract as much profit as possible, from these various services and platforms. And that has made it so that, you know, people are not feeling the benefits or people are not enjoying using them in the way that they were before. You know, the streaming services are more expensive and are often limiting the amount of content. The social media companies, you know, have a lot of issues with them such that people are really demanding regulation from them now, and that's just to touch on a few things. And so I think that we're at this moment where people are looking for something different. People are looking for a reassessment. And the question is, what is that really going to look like? And, you know, part of what we discussed in that— the report that you mentioned on digital sovereignty, which is obviously a bit more focused outside the United States and kinda getting off of some of this US technology that we've become dependent on, but it's really to say, what technologies are benefiting the public? You know, if the current technologies that we have are not doing so, what would that look like, and what structures would we need to set up to develop technology differently? And especially now that we look at the data center and generative AI moment, I think that there's a real question of, like, you know, we're being told that we need a certain amount of computation for the technologies that the tech industry wants to build, the generative AI tools in particular. But if we took a step back and said, you know, what technology would actually serve the public? What would benefit people more broadly? And does that require the amount of computation that the tech industry is telling us they need right now? I think that we could find that we could create technology that is socially beneficial, while, you know, kinda dedicating or needing to sacrifice far fewer resources than what the tech industry wants right now for its current vision of what technology and what digital technology and what AI should be. And I think that it's time for us to have that conversation, especially when people are not comfortable with the amount of power that the tech industry has and how it is necessarily wielding that power. And, you know, then if you look at the opposition to data centers, I think that that is a real, opportunity to start having those broader conversations about what technology we use, how we develop technology, and how we maybe do that, you know, thinking a bit more about the public benefit rather than just what shareholders want or what, you know, major tech CEOs think is the way forward.

00:44:55 Andrew Keen: Yes. Canadians and Europeans are always talking about conversations. Paris, I'm not convinced that— I mean, maybe you're right, but I'm not sure it's very practical. I'm talking to you from San Francisco, the heart, especially now, really, the heart of Silicon Valley to certainly the heart of the AI revolution. You're an expert on transportation. San Francisco streets now are increasingly dominated by self-driving cars like Google's Waymo. Your previous book was Road to Nowhere: What Silicon Valley Gets Wrong About the Future of Transportation. You've written about how, you've written about the high cost of Uber's small profit. I'm guessing you would say the same about Waymo. What can we learn about transportation? I mean, given that this is another area of expertise in terms of addressing the hyperscale question. What— I mean, you could argue, well, people want Uber. I mean, it provides jobs. People want self-driving cars. It does, certainly in San Francisco. It makes going out on the street certainly less dangerous. So are there things that we can learn in a positive way, even if your book was called Road to Nowhere? It was, I mean, it wasn't just nowhere. We're going somewhere in our Waymo and Uber cars. What can we learn positively about transportation in terms of guiding this bigger AI revolution in a way that we want as society?

00:46:26 Paris Marx: Absolutely. I think it's an important question. Right? And what I would come back to is that, you know, one of my big takeaways from that book and that I think has really carried through is that it's not just new technology that is going to solve transportation problems. Right? You know, these tech companies entered our cities and really said that if you just give us some time to develop this new technology, we're going to solve traffic, and we're going to solve the cost of transportation, and we're going to address equity problems and, you know, on and on and on. And, you know, the— they didn't really do that. And a lot of the technologies they promised were not even realized or certainly not in the way that they claim they would. And so for me, it's a real lesson that if we want to improve streets, if we want to improve society, it's not just saying, you know, how are we going to fit this new technology into this goal, but rather what is the goal, how do we achieve it, and can technology serve that? You know, I would say if we're trying to address a lot of those problems with transportation, you know, road deaths, the amount of time people are stuck in traffic, all that kind of stuff, you'd wanna invest more in public transportation. Right? Not just technology that is gonna try to make cars more efficient. Because I feel like one of my takeaways from that first book was, again, it was kind of the scale of the automotive system, just the kind of amount of cars that are on the street that are fundamentally, you know, the problem there. Right? There's just not enough space for all that cars in densely packed cities, and so you need more efficient ways to move people around. And just using different technologies is not going to deliver that. You know? There needs to be kind of a physical form. And so then if we move forward to what I'm talking about in this new book, again, I think that scale is part of the problem here that we're building out so much computation that it just becomes, you know, fundamentally unsustainable when we look at the amount of fossil fuels that these companies are now building out in order to power these hyperscale data centers. You know, it makes it so that addressing climate change is very tough, if not impossible, and we're all going to feel the effects of that in years to come. And then the question is not just how much computation is needed for it, but is that amount of computation justified for the technologies that we're getting out of it? And that is where I'm really not convinced. You know, I don't think that generative AI justifies the sacrifices that we are all collectively making, for the chatbots and image generators that the tech companies are selling us. And that is, I think, you know, the conversation and the discussions that we need to have, but also to challenge kind of the vision of the future that these tech companies are putting out there and to start to take back a bit of that power and that— even in a sense, like, imaginary power for ourselves to say, we shouldn't just be aligned with what they think the future should be. We should have our own vision for what that future should be and think about how we bring it into being and how we take back some of the power in order to do that.

00:49:06 Andrew Keen: In other words, we gotta be critical of not just hyperscale or the hyperscalers, but the idea of scale itself. Maybe there's a lesson. Paris Marx is based in Canada. Of course, it was the Canadians who pushed back on Google in terms of scale and cities. They rejected their Sidewalk initiative in Toronto [ed.: Sidewalk Labs' Quayside project]. Congratulations. I'm not sure if that's the right word, Paris, on the book. It's certainly an important book, and it's going to have a significant influence. It's out today. I just have to ask you about your name. It's a wonderful name. Were your parents, did they— were they— before naming you, were they reading Brave New World?

00:49:46 Paris Marx: No. I don't believe so.

00:49:49 Andrew Keen: What— what's the story on the name? Paris Marx, memorable name.

00:49:53 Paris Marx: It— it's definitely memorable. I certainly get a lot of comments on it. You know, it's a bit more, I think, the choices of a younger person that, brought that name into being. Beats Paris, Texas.

00:50:04 Andrew Keen: Yeah. And, it's, as I said, it could come out of Brave New World. And I guess this book, Hyperscale, is the next chapter in our Brave New World of AI and hyperscalers, and very, very important, new contribution, critical contribution, of course. Paris, great to have you on the show. I'd love to get you back. So much more to talk about. Thank you so much, and congratulations on the book.

00:50:29 Paris Marx: Thanks so much. I really appreciate it.