Transcription
Capital expenditure on artificial intelligence infrastructure currently makes up 1.2% of US GDP. Um, now, that's a proportion of expenditure, um, on one single technology, um, which is pretty much unheard of since the railroad boom in the late 19th century. Um, it's not normal, and to a large extent, it's being driven by just the four biggest tech firms, or the four biggest tech firms who are, who are investing in this. So, Amazon, Microsoft, Google, and Meta invested $und00 billion in capital in the past 3 months. So, just three months, that's triple the amount they spent, um, in 3 months, two years earlier. This is just a huge boom. Um, it's this spending which, in fact, is keeping the US economy growing despite Trump's tariffs. So, everyone has sort of been saying, how is it the case that Trump has this terrible economic policy, but the United States just keeps on growing? Well, one theory, um, which is borne out in, in, in the growth statistics, is that it's just because so much money is going into data centers, and that has a bit of a multiplier effect because people have to be employed to, to build those. Um, that all means it's an interesting time for the world's most prominent AI CEO to say this.
And so, late last week, Sam Alman, um, had a dinner with various journalists. Um, it was on the record, and The Verge published this quote. Um, so Sam Alman said, "When bubbles happen, smart people get over-excited about a kernel of truth. If you look at most of the bubbles in history, like the tech bubble, there was a real thing. Tech was really important. The internet was a really big deal. People got over-excited. Are we in a phase where investors, as a whole, are over-excited about AI? My opinion is yes. Is AI the most important thing to happen in a very long time? My opinion is also yes."
Um, so that was Sam Alman saying we are in a bubble, um, and investors are over-excited. He said elsewhere, um, that he thinks some people will probably lose a lot of money, and that warning from Sam Alman appears to have had an effect on the market. So, Fortune.com reported that US tech stocks had slid following Alman's warnings of a bubble. Um, and they also sort of highlighted a skeptical study from MIT. So, this is from, uh, Fortune, Nvidia. So, that's the company that makes, uh, chips, fresh off becoming the world's first $4 trillion company, sank 3.5%, while Palantir slid nearly 10%. The sell-off appeared to be sparked in part by an MIT report that claimed 95% of companies investing in generative AI are seeing no returns, and was potentially deepened by earlier comments from OpenAI's Sam Alman suggesting investors may be caught in an AI bubble. Um, late last week, Auburn drew a parallel between today's AI frenzy and the 1990s dot-com bubble, when internet company valuations spiked dramatically before crashing. And while the MIT study attributed failures to corporate learning gaps and flawed integration rather than actual AI model quality, the market reaction highlights growing concerns about AI's commercial viability.
Um, Helena, this is over, sort of, all the newspapers, all the financial newspapers, um, on, sort of, you know, Fox Business and everything this week. If you look on, on, on YouTube, um, people talking about the AI bubble finally popping. Do you think, uh, we're lined up for a crash?
>> I think so. I, I think that it's clear that there is a lot of venture capital going into an industry that has shown precious little real-world application in terms of actually being commercially viable so far. The issue, of course, is that in a world of Silicon Valley tech startups, everyone's looking for that new angel investment, right? Like, all of the kind of really rich people that we've seen right now aren't people who've been at the forefront of innovation themselves. They've been angel investors who've been able to make money off other people's work by being in the right place at the right time. You look at Elon Musk and you look at the PayPal Mafia, people like that, right? Like, there's a real issue that we have when it comes to people wanting to be the angel investor on the next big thing. And in the tech world, the next big thing is how you be the person who owns the new thing that becomes private infrastructure, as in part of the economy that everybody has to engage with. Meaning that there are so many future returns because the whole society will depend on it. For example, you look at people like the Gary Vaynerchuk of this world who became angel investors in Facebook. Now, that's become part of the private infrastructure. Obviously, loads of people in society use it and engage with it, for example. That might go away, right? Other private companies have to rely on things like Facebook to be able to engage in the market at all, thereby making it a ubiquitous part of the tech landscape, and that's why the valuation of companies like Meta is so high. You are now part of the private infrastructure. Everybody has to engage with it. Look at, for example, Uber Eats, Just Eat. That's now part of the private infrastructure because you don't call up a takeaway anymore to be able to order what you want. You have to engage with this app that has all of the stuff on it. So that means valuation goes up because it's been replacing an old infrastructure with a new one.
And so, when it looks at AI, you see this new way of working, or all the jobs it's going to take. For example, that's the line that comes from the, uh, the AI progenitors. It's going to replace jobs that we don't need anymore. It's going to really increase productivity. Whichever AI tech company, or whatever it is, is the one that replaces that level of infrastructure by being the person who has the jobs being replaced, or that replaces the jobs that's important. Suddenly then, every company that relies on that kind of labor is going to be starting using AI, and you're going to be the one with the intellectual property that's going to be part of the private infrastructure. But of course, if that use case doesn't emerge, it's all out of venture capital that's massively spiking the value of these tech startup companies, or people who are looking to become the next big thing. And if they don't get there, it's completely worthless, and all that money has been sped up the wall. And on top of this, you have all of the baseline level which would provide the value for AI, such as Nvidia, who provide the chips for all of these AI center data centers and things like that. They are also going to lose out because they are the ones who are supposed to be there to cash in on the need for the, the public and private infrastructure that would be there to facilitate the private infrastructure for the AI companies. So there's a really big kind of cyclical bubble going on here, not just in terms of the companies themselves, but the companies that rely on those companies being successful too. So there really is a problem if we don't actually find the golden egg within the kind of morass of these AI companies. And you might get a crash even if there is ultimately a golden egg.
I think that's what's interesting about sort of having read about this a lot over the past week, which is the idea of there being a bubble doesn't necessarily mean the technology is, is all hype. Um, it might be, but it doesn't, that's not what the bubble necessarily implies. Let's look at what happened in the dot-com bubble, um, which was in the year 2000. So, this is the NASDAQ composite index, and so it shows the stock valuation of the main tech stocks, um, and between 1996 and its peak in March, or its then peak in March 2000, you can see the valuation of stocks in the index increased by five times, and this is when everyone was really hyped about the internet. Um, and then it fell by, or fell to a fifth of the size of its peak by October 2002, so giving up all its gains during the bubble. So, this is the, the dot-com bubble in 2000. People super excited about the internet. The internet is the future. They invest in all of these different new internet companies which they think might be the next big thing. Um, and then everyone realizes it's overvalued, and the whole thing just, just crashes. Um, in that crash, um, many online shopping companies, including pets.com, Webvan, and Boo.com, went bust, although I'm sure someone owns the pets.com URL now, doesn't it? Um, tech giants like Amazon and Cisco Systems also lost large proportions of their market valuations. Um, ultimately though, um, companies like Amazon, of course, recovered and went on to be some of the most profitable companies of all time. Um, and the dot-com bust didn't mean the internet wouldn't transform society and the economy.
And then, even on the sort of very, let's say, limited perspective of the NASDAQ index, um, look at how it's changed up till today. And so, this is up until 2025. And you can see that with that huge increase in 2025, um, it looks a bit like a blip, what happened in 2000. Now, in fairness, it's a, it's a, it's a long blip because it took 15 years for the NASDAQ to reach its pre-crash level. Um, but since 2015, and especially since COVID, um, tech stocks have soared, and they're now worth four times what they were at the peak of the dot-com bubble. Um, of course, the speed and extent of that increase over the past four years does look a bit like a crash could be coming. I think it has a lot to do with, um, cheap credit in the COVID period.
Um, will the AI build-out though still change the world even if there is a crash? And this seems to be what Sam Alman is suggesting. If there's a crash, don't let that make you think that AI, it's a dud. Um, and he is still suggesting in the same interview, in fact, where he said that this might all be a bubble. Um, he said, uh, this about data centers. So, he said, "You should expect OpenAI to spend trillions of dollars on data center construction in the not very distant future." Um, so he's saying that this might be a bubble. It might pop, but also we're going to spend even more on data centers in the near future.
And on a recent podcast with Fio Von Orman, had this interesting exchange about data centers.
>> Starts to make our planet look like, um, a software board, like a.
>> It does. You know, when you see it from the air, I was really struck by that. I was like, this looks like the motherboard of a computer.
>> Yeah, it looks like the motherboard of a computer. Start to see like how the planets in, like a lot of these like, uh, sci-fi movies, a lot of them look have that R2-D2 look on the outside of them because they've been.
>> Covered in data centers.
>> Yeah. Which is kind of wild. Do you know where we're going and you're not telling? Are you going?
>> I don't. I don't.
>> You promise, dude?
>> I don't know. I mean, I have all my guesses. Like, I do guess that a lot of the world gets covered in data centers over time, but I don't know because maybe we put them in space.
>> Like maybe we build a big Dyson sphere on the solar system and say, "Hey, actually makes no sense to put these on Earth."
>> Yeah. Uh, Helena, um, does that make, I suppose you could do the pre, because obviously you, you want your data center, you know, if you're using AI, you want there to be not that much latency between your phone and the data center. So, I don't know how, sort of, it being this big donut around the Earth, also kind of a big donut around the Earth that might also block the sun. I don't know. I don't really like the sound of that. What did you make of that idea of of data centers potentially covering most of the earth, or if not being in a kind of, was it, was he describing a kind of donut around the the earth, or was I misunderstanding something there?
>> Yeah, it's the Dyson sphere. I actually did a project at my A-level physics class on a Dyson sphere, which was very interesting. Um, but I mean, it's just, it just seemed to be this again, um, hype marketing tool going, "Oh, look at, look at the potential what AI could bring if it becomes this ubiquitous technology, we're going to need to be able to build it on the moon." I'm like, I mean, the Adam Smith Institute said we should privatize the moon basically as a joke so they could get column inches. So, let's just temper, let's just temper our expectations here. But I mean, as again, you, even if there isn't a bubble, as you rightly point out, it could be that there is still plenty of utility that would grow over time rather than be part of this kind of rush to be the, the next big thing that might require a lot more investment. But do I think it's going to be going as far as necessitating a Dyson sphere or necessitating colonizing the moon to be able to place data centers on there? I, I think that's very far-fetched.
What I do want to make a point about though, what I do want to make a point about is I talk about private infrastructure a lot in the first half of what we're talking about there. There's also like social infrastructure, ensuring that you have government subsidy for these things, right? Which is why, obviously, I tech companies still need to ensure that despite there being a potential bubble, they have people in the right places who want there to be investment in AI as being the technology of the future. In terms of government, like we've seen in terms of how the Tony Blair Institute have really kind of sold on this idea of AI is the next big thing to ensure that they're going to revolutionize how the state is run, and this really is the goose that laid the golden egg because you can guarantee you have some level of part of the infrastructure of tomorrow if you have government ministers on side. And socialists and capitalists both hate this. Socialists hate it because they believe that it's undermining the role of the state in maximizing social utility rather than doing what the donors want them to do. And capitalists think that it's undermining the primacy of the market by having the state choose winners. Everybody hates this stuff. Everybody hates this. And this is why we've seen like this investment coming for AI. It's why we saw, uh, the Science and Technology Minister Peter Kyle say to the Alan Turing Institute, "Actually, you need to change towards focusing on projects around things such as defense and stop what you're doing right now." It just so happens that the things that they were looking into was, is AI going to be safe? Is AI actually going to have human applicability? Right. What are the problems going to be with the adoption of AI technology? Weird. Weird how a government that seems to have loads of links to, uh, think tanks who are very pro-AI have ministers who want, or who at least ask at some point, the, uh, think tanks that are funded by the government to not look into what they were doing that might hurt the profits of AI companies, but move towards something that the government wants to see for political reasons in terms of defense spending. So, there is a real problem that we have in terms of the level to which that there is at least some level of corporate capture in terms of governments wanting to push this AI stuff so that they ensure that the golden egg does get laid. And of course, they want to open up loads of new planning permission for all these data centers in the UK. Might turn us into Coruscant, and the kind of voters that Labour needs to rely on to ensure that they stay in government, they really don't want the green and pleasant land turning into Coruscant, I can tell you.