Do you think that discussion of a bubble or the possibility of there being a bubble is the wrong discussion? Maybe there is something more important, more significant that you think people should be talking about instead.
I think it's the wrong discussion. I don't like the word bubble. Were automobiles a bubble in the 1920s? Again, a lot of the companies failed. They were not the best companies. Could you imagine companies, whether AI or not, today failing because of Middle Eastern war, war with Russia, mistakes from the Fed, half a dozen other reasons? Of course you can. So things can fail. It doesn't mean they were bubbles to begin with. A bubble to begin with is something like the South Sea bubble or the Dutch tulip bulbs that just made no sense. This is not that. I think the key question is just how much will society accept the changes coming from AI.
Today's number is 40. That's the percentage of Americans who think listening to audiobooks does not count as reading. Ed, true story. I recently had a book fall on my head, but you know, I have no one to blame but my shelf.
We're back. We're back, baby.
Tell me everything about your vacation.
I started...
You jumped to the punch line too quickly. You got to give it a second. Let me get going.
Go ahead. What did you do? I am sort of interested. What did you do? Where'd you go?
I was in Munich, or actually just outside of Munich in Ammergau, for a friend's wedding, which was incredible, in Bavaria. Then I rented a car with my girlfriend, drove down the autobahn into Kitzbühel in Austria. Spent a week in Kitzbühel hiking, chilling, going to the spa. It was probably, I think it was the best vacation of my life. I've been sort of pondering on it, but I think it's my number one.
And then I spent the weekend in London and I decided, you know what? I need to see the opening match of the Premier League. So I went to the Chelsea game and sat in the front row and watched Chelsea beat Fulham. So I'm feeling very good. I miss it. I mean, I love this podcast. I love doing this with you, but I could have maybe taken a few more weeks off, but I'm feeling very good. I feel ready. Ready for the year.
Oh, that's great. Yeah, you sound energized.
Yeah, August was good. I dropped my kid off at school. Nothing screams confident masculinity like crying in your car when your son walks out of Best Buy with his own mini fridge. But yeah, that was the big deal, was dropping my kid off at school. What are your takeaways having dropped your first son off at college? Well, the way you summarize it is your life gets smaller so theirs can get bigger, and it's just, you're a tide pool of emotions because everything just sort of hits you all at once. And there was all these things you were going to do that you didn't get a chance to do. I had all these, you know, we did a lot. I think I was more present than most dads, but you know, I had thought, oh, I got to take my son to Alaska. I want to buy a car and restore with him, and I want to take him to, you know, Latin America, and I wanted him to come on tour with me on one of my speaking gig things. And all you think about is all the things you were supposed to do that you didn't. But he makes it a little bit easier not to be sad because I think of myself as this spiritual, loving guy, and he's literally like, I got this, Dad, and shoving me out of the room after about seven minutes. And also, I mean, quite frankly, I didn't want to make it about me, so I thought I'd keep it together. Your 18-year-old does not need the emotional baggage of his father singing Cats in the Cradle over and over in his head. So yeah, I tried to hold it together. He'd be proud of me. I didn't cry, and we just fixed up his dorm room, and we threaded the needle between a comfortable, nice dorm room and one of these ridiculous Real Housewives of Palm Beach things where they turn it into the Ritz-Carlton, Charlottesville. So we went to Target. I'm like, I'm not going to buy anything nice. You can get anything you want, but it's coming from Target. And then the moment I lost it was actually after we dropped him, and I was back at the hotel, and I was sitting there, and I noticed this car stop at the dorm, or the business school, the law school's next to where we were staying, and this mom got out on the driver's side, and then a kid got out of the back seat, and then the dad got out of the passenger seat, and then they hugged, they embraced, and the dad immediately had to walk away, and he started crying. And I realized it was their own drop-off, and then I just lost it. So whoever that dad was losing it on the sidewalk kind of did my, like, express my emotions for me. Are you surprised by the significance of dropping the kid off at college? Like, I know people talk about it, and I know it's like, you know, empty nest, what is it, syndrome? Empty nest, whatever the feeling is called. I know it's a thing, but I wonder if it's more of a thing than you expected.
First off, I think it's difficult and a big moment for a lot of people. And it's just so different now. My drop-off was, my mom gave me an iron, a used iron, and said, 'Here, you're going to need this.' That was my big drop-off, was she left an iron on the kitchen table for me to take to college. That was the emotional drop-off. There was my drop-off in 19, you know, '92, '82. This is a much bigger deal. You know, drop-off is a much bigger deal now. For me, I think I... Well, it just marks a different point in your life where your kids are no longer at home. And the thing I'm wrestling with a little bit on an existential level is I've always described my purpose as raising, you know, preparing my sons for others. And it's been a nice source of purpose for me. And now you kind of worry like, okay, am I going to have a purpose moving forward? So it marks time. It's like basically now really all I have to look forward to is the ass cancer.
Don't lie. You've been looking forward to that for years.
Yeah, I've been talking about it a lot. Look, it's impossible to explain until it actually happens to you. But I tried to hold it together such that, again, I didn't want to do what I always do, and that's turn it to me. And he seems happy. I speak to my boys every day, which makes things easier for me. So I talk to him about his classes and what's going on.
Oh, that's good. And also, everything that I'd hoped for in terms of UVA is out of central casting. You know, this campus is great. The kids seem really smart and nice, and you know, he's having a great time, and so that makes it easier. So we'll see. We'll see.
Sounds like you're not sure what you make of it yet.
I'm an emotional tidepool. I don't know. I don't know how to process it all is the bottom line.
I'm like, what are your takeaways?
Yeah. No, I don't, I don't, I don't have...
Well, we'll check in on it for sure. I appreciate that. Yeah, it's an interesting time. Well, now that we got the life advice or life reflections out of the way, we have a very, very interesting discussion with someone who we've wanted to have on the show for a while now. So don't miss it. We've been away on break for the past two weeks, and in that time there have been a ton of developments in the world of AI. One of the biggest was the hacking of Hugging Face by rogue agents from OpenAI. But beyond that breaking news, there are still some lingering questions about AI that are keeping the market on edge. For example, are we in a bubble? How will AI reshape the job market? And is the massive spending that we're seeing from these companies actually sustainable? So to help us make sense of it all, we are turning to someone who a lot of the other experts we've had on this show often listen to. He has consistently been named one of the most influential economists and thinkers in the world. And he has been described as, quote, 'the man who wants to know everything.' So here is our conversation with Tyler Cowen, economist, author, podcaster, and chair of the Mercatus Center at George Mason University. Tyler, great to be joined by you today. Very excited to get into this conversation. I'd love to start with some questions about the AI bubble, whether or not we are in one. And I would especially love to start with a clip from an interview that went viral recently. This was from the Diary of a CEO's interview with Ed Zitron, who has become famously one of AI's biggest critics. I'm going to play you what he said about AI, and I want to get your reactions to start this off.
I think generative AI is at its heart a con. I don't think it is sold as honest software. I think that they overstate both what it can do, what it will do, and the underlying financials to the point that they are misleading the entire world. And they're actively exploiting the weaknesses in journalism, in our economies, and indeed within the responsible parties with sell-side analysts, governments, and all over the shop.
The word 'con' is a strong word.
Yeah. I mean, what do you call something where from the very beginning they've sold it in the terms of magic as this thing that will replace all jobs, that will cure cancer, as all of these things, and when you look at it, it's boring cloud software that's extremely expensive and unprofitable and also unreliable at its core.
You are a world-renowned economist. You've written a lot about AI. You also serve on Anthropic's economic advisory council. So you think about this stuff a lot. What do you make of those comments? Do you agree with them?
AI is like magic in many ways. There are many, many tasks where it does better than humans. It doesn't mean it will take away your job because your job involves the physical world as well and dealing with people. But for intellectual tasks, it's a truly remarkable achievement, one of the top in the history of mankind. There's a tweet circulating on Twitter about all of Ed Zitron's wrong predictions so far. OpenAI just hit a billion dollars of revenue from ads. The revenue to date for Anthropic and OpenAI has been incredible. There's also Grok, Meta, other companies, the Chinese waiting in the wings. I'm not saying every company will make it. There were plenty of car companies in the 1920s. They're not mostly around today, but cars are a big, big thing, and AI is too.
When you think about this idea of the circularity of the AI ecosystem, this idea, and this is something that Zitron has talked about, that you have a lot of these companies, these big tech companies that are investing in Anthropic and OpenAI, and then OpenAI and Anthropic, and then turning around and spending those dollars on compute that is sold to them by basically their investors. And the fact that, you know, OpenAI and Anthropic have made up such a significant portion of these big tech companies' AI revenue, for example, Microsoft, 70% of their AI revenue last year came from OpenAI. Does that element of it concern you, not the technology, but the way that the technology is being sold right now?
New things bootstrap themselves all the time. You can think of Nvidia as a kind of lender or buyer of last resort for the sector. You could say the same about Microsoft, arguably Google, maybe Meta as well. There's a lot of capital in the sector that increases its long-run chances of making it. Again, the point is not that every single company will prosper, but the stuff works. There's no reason to be skeptical, per se, about the economic future of AI.
If one of those companies were to fail, because you're saying, you know, maybe these companies won't all make it out alive, is that not quite a significant statement if it's not a certainty that some of these AI labs wouldn't make it out of this, or that they wouldn't prosper or succeed at the valuations that we're seeing? Looking at how systemic they have become to our economy, is that not something that worries you?
It's a super competitive sector. You could say in a way we should hope they don't all make it, that some of them have much better products than the others. Are there economic costs to companies going out of business? Of course. The companies themselves are not very heavily financed by debt. You could say there's other parts of the supply chain, data centers and the like, or people investing in those that are financed by debt that could have some bad macro consequences. But again, every new technology we've had, including the internet, including cars, including railroads, there are ups and downs to the cycle. Not every company makes it. The really good stuff basically works and it sticks around with us forever, like the railroads do. Do I worry about day-to-day volatility? Of course, but keep it in perspective, right? Every sector has had this. You have to ask yourself, does the product work? And the answer here is a very clear yes.
We saw the dot-com implosion again. Going with your analogy, like, you know, there's value here and there will be ups and downs. Do you think that that is a scenario that is probable, possible? Do you think that that is something that should be talked about?
It's possible, but there are big differences. So revenue growth for the current AI companies looks much better than what we saw before the dot-com bubble burst, right? The degree of capitalization in the sector is much better and much stronger. Just our confidence in the technology. So a lot of things are possible. Again, volatility would not in the slightest surprise me. It's been a historical regularity. Pets.com went under, but you can buy pet food online all day long if you want, and that's the world we're headed for with AI.
Do you think that discussion of a bubble or the possibility of there being a bubble is the wrong discussion? Do you think that maybe there is something more important, more significant that you think people should be talking about instead?
I think it's the wrong discussion. I don't like the word bubble. You know, were automobiles a bubble in the 1920s? Again, a lot of the companies failed. They were not the best companies. Could you imagine companies, whether AI or not, today failing because of Middle Eastern war, war with Russia, mistakes from the Fed, half a dozen other reasons? Of course you can. So things can fail. It doesn't mean they were bubbles to begin with. A bubble to begin with is something like the South Sea bubble or the Dutch tulip bulbs that just made no sense. This is not that. I think the key question is just how much will society accept the changes coming from AI. That is indeed a very open question. We might decide to stop or halt or slow down those changes. And there I think we should be very agnostic.
I'll put forward a thesis and you tell me where you think we have a right or wrong. You mentioned the automobile market. It feels to me that an apt analogy would be that the frontier models are German automobile manufacturers: higher price, prestige, better technology, larger margins, and that the open-weight Chinese models are more like the Korean or the Japanese auto market that end up putting more cars, or AI, in the hands of users around the world. Do you feel that's an accurate description? Do you feel that the market is kind of bifurcating?
Open-source models make frontier models easier to use. You use frontier models for the frontier tasks and you use open-source models for simpler tasks. So it doesn't have to be either-or. Right now the best proprietary models are better for tough tasks. That makes a big difference using GPT-5. You can do many things that you cannot do with the Chinese open-source models. I think China also is more likely to get scared of its own AI than America is. So there's no guarantee progress on the Chinese front will continue. So again, there's a lot of competition, but the idea that American business is just going to wholesale rush to Chinese models and abandon the best American models, I don't expect that.
So looking at the market dynamics, I've read that $2.5 trillion in capex so far, about revenues of $150 billion. And in all the technologies we've talked about — railroads, the internet, the grid, highways — the technology oftentimes survives valuations. This just feels eerily reminiscent of all these other innovations where you get above a certain level of GDP spend on capex and you have a pretty serious correction regardless of the viability or the importance of the technology. I mean, you're an economist, you look at economic cycles. This feels eerily reminiscent of those different innovations where the technology survived the valuations.
Again, a correction wouldn't surprise me, but keep in mind this technology has advanced much more quickly than the others. We're in a much wealthier world. Debt plays a much less significant role in this sector. There's massive capitalization in the sector due to the major tech companies who would love to be able to buy up a great proprietary model if it came to that. So in all those ways, this is not like, for instance, the railroads. But again, I absolutely expect big ups and downs. No one should be surprised by that. As you point out, that's the norm in history.
We'll be right back after the break. And if you're enjoying the show so far, send it to a friend and please follow us on YouTube, Spotify, or wherever you get your podcasts. Support for the show comes from VCX, the public ticker for private tech. For generations, American companies have moved the world forward through their ingenuity and determination. And for generations, everyday Americans could be part of that journey through perhaps the greatest innovation of all, the US stock market. It didn't matter whether you were a factory worker in Detroit or a farmer in Omaha. Anyone could own a piece of the great American companies. But now that's changed. Today, our most innovative companies are staying private rather than going public. The result is that everyday Americans are excluded from investing and getting left further behind while a select few reap all the benefits. Until now. Introducing VCX, the public ticker for private tech. VCX by Fundrise gives everyone the opportunity to invest in the next generation of innovation, including the companies leading the AI revolution, space exploration, defense tech, and more. Visit getvcx.com for more info. That's getvcx.com. Carefully consider the investment material before investing, including objectives, risk charges, and expenses. This and other information can be found in the fund's prospectus at getvcx.com. This is a paid sponsorship.
We're back with Prof Galloway. Let's talk a little bit about the labor markets, and that is the kind of idiocracy scenario where these tools result in such an incredible increase in productivity that we can afford to just pay everyone to stay home and their jobs go away. When I look at the labor market, what I see is a market where you wouldn't know AI existed if you didn't know it existed. I challenge anyone to discern where this chaos in the labor market is supposedly manifesting. Curious to get your predictions or sense for the intersection between AI and the labor market, and where the market appears to be getting it right or wrong about this somewhat of a catastrophizing around the labor markets.
I agree with what you just said. I think in programming it's quite noticeable, but there's some very recent papers, one by John Hartley, and you just don't see it in the numbers. Some modest amount of jobs might be going away because of AI. Some modest amount of jobs might be created because of AI. Those two numbers appear to roughly balance. AI will spread over time a lot more slowly than AI advocates often seem to think, and I think labor markets will be pretty stable.
The only thing you said that seemed unusual to me, or I would maybe... I wonder, everyone's talking about kind of a supply crisis, or that we don't have the infrastructure or the supply chain to support the increase in demand. And it looks to me as if on the front end, several players that were anticipating more demand, whether it's Meta or X, have immediately pivoted their infrastructure where they're now leasing out their supply because they haven't been able to create the demand they had initially anticipated. Do you think there's a chance that we might go from what is anticipated being a supply crisis to a front-end demand crisis where there are just a small number of players creating front-end demand?
There's some data that have come out from the two main companies on how many tokens people are using per month, and that number has just been going crazy through the roof. So I don't think there'll be a demand crisis. At some point, there may be an energy sector crisis. Right now, the US has a huge share of all the world's compute. There's no guarantee we keep that position. So to build out compute, secure the supply of chips largely through Taiwan, make sure that over time we increase our own energy and number of data centers against strong political opposition — those are big challenges. I think we'll squeak by on all those fronts, but again, you could definitely see things going wrong.
All you would need is one decision from the Chinese Communist Party to invade Taiwan, and we'd be living in a very different world, right?
On the demand for tokens, which is true, I mean, the demand is going up significantly. One of the criticisms, or at least the concerns though, is that that token usage or that token demand is currently being subsidized by the AI labs. I mean, Anthropic, OpenAI, they are losing money on every prompt that a user puts into the system. And so I think there is a concern that if we were to not subsidize this, if Silicon Valley, if VCs, if big tech were not to basically support the business models of OpenAI and Anthropic, then maybe it would be different. Maybe the cost of token usage would be higher and people would be less inclined to use as many tokens as they're using today. What do you make of those concerns? Do you think that they're valid?
In the short run, the price may well need to go up for the reasons you mentioned, and we already see what is possibly some rationing of compute. But that said, if you look at data, how much has the price of tokens fallen since, say, GPT-4? I forget the number, but it's more than 100x. So we're doing very well on making the systems more efficient. And we'll see, you know, which curve outraces which other curve, but it's not that there's no progress in making the systems more efficient. There's been incredible progress. But in the short run, should the price go up? Probably.
But aren't costs going higher for these companies? I mean, my understanding is that it's hard to know because they're not public companies. We can't look at their audited financials, but my understanding is that Anthropic and OpenAI are spending more each year.
A better model costs more. Yeah.
Right. Which would assume, I mean, to me that means that we're not necessarily going in the right direction in terms of making this a profitable business that is necessarily sustainable. To me, it seems like we're technically going in the wrong direction if we're talking about profitability, which is we're losing more money.
Well, I'm not advising anyone to buy any particular asset. But the worst-case scenario is that capital gives a massive subsidy to consumers and we end up still having the model and someone else buys it. That's a great outcome for an egalitarian, say. I don't think that will happen. I think the companies will do fine. There's a reason why people talk about Anthropic at two trillion. I'm not sure what the current number is for OpenAI, but I think it's over a trillion. Those are guesses, but those are smart people with real money on the line, and they're saying the companies are worth trillions. On average, we should believe them.
Why should we believe them in your view?
Because it's their own money on the line. Once things go public, you'll have the chance to short them, right? You're free to do that. I'm free to do that. I have no plan to do that whatsoever. The way you really make money in markets is to go long. I believe in America. I believe in tech. I think artificial intelligence works. Again, I have no favorite horse in terms of which particular company, and I definitely think they're not all going to make it. But again, some of the big ones really will.
I guess just from the economics perspective, you know, regardless of whether someone wants to go long or short, because of how dependent the US economy seems to have become on AI, or at least the AI buildout, the amount that S&P earnings growth is dependent on AI, the amount that the entire stock market is increasingly dependent on AI, and the extent to which the ecosystem itself does depend on those two companies, does that not make it bigger than, you know, some will win, some will lose? I mean, aren't we all implicated, whether we're long or short or not?