This is an edited transcript of “The Ezra Klein Show.” You can listen to the episode wherever you get your podcasts.
Right now, A.I. is the driving force behind the gains in the U.S. stock market and U.S. economic growth.
It is sucking up so many construction resources, it now appears to be crowding out housing, manufacturing facilities and basically everything else you might need to build.
And the question, if only from a financial perspective, is: Is this all worth it?
Is all of this investment going to pay off? Or is A.I. a bubble, and one of these companies is going to miss an earnings target or get downgraded, and the whole thing is going to pop — taking all of our retirement portfolios with it?
So I wanted to have two different perspectives on this.
Paul Kedrosky is a partner at the venture capital firm SK Ventures. He’s a fellow at M.I.T.’s Initiative on the Digital Economy and has a popular newsletter at paulkedrosky.com. He thinks we are in an A.I. bubble and has been very vocal about it.
Azeem Azhar is a British technology entrepreneur, an author and an analyst. He is a creator of the Exponential View, which is both a great Substack I read often and a research platform where, among other things, they have been creating dashboards tracking whether there’s financial stress in the A.I. industry.
I think it’s fair to say that Azhar’s view is somewhat more optimistic than Kedrosky’s. So I wanted to put the two of them into conversation together to try to see why you might or might not think we are in an A.I. bubble.
Ezra Klein: Paul Kedrosky, Azeem Azhar — welcome to the show.
Paul Kedrosky: Hey, Ezra. Good to be here.
Azeem Azhar: It’s great to be here. Thanks.
So Paul, make the case to me that this is a bubble.
I mean, it’s a big claim. You’re saying that all these investors, all this money pouring in, all these sophisticated analysts at BlackRock and Goldman and wherever — along with the hyperscalers like Google and Meta — are all getting this wrong on some level, and you’re not.
So why do you think it’s a bubble, and they’re all idiots?
Kedrosky: Well, I wouldn’t go quite so far as they’re all idiots. [Chuckles.] That may be true, though I’ll step back on that one.
So what’s unusual about this — let’s just talk about, really quickly, what a bubble is. A bubble occurs whenever asset prices detach from any semblance of economic reality.
We have one of the largest competitive capital expenditure spending waves in history. So in Western economic history, there have been eight or nine of these moments when we build out some giant general-purpose technology — a G.P.T. Not the ChatGPT kind — but a general-purpose technology.
And we’re building it out like we were the railroads, like we were electrification, like we were the canals, like we were the interstate highways. And every one of those moments has led to overbuild.
There isn’t a major economic capex episode in Western economic history that didn’t end in a rolling series of crises, vastly inflated asset prices and a vast oversupply of capacity of whatever the thing was — railroads, fiber-optic cable, take your pick.
And so that’s the bubble. The bubble is — once again, we’re vastly oversupplying something because of the availability of capital and the eagerness of the suppliers of the capital to continue building out in anticipation of future demand.
Now, none of this is to say that A.I. is frivolous or it doesn’t work, or that there aren’t incredibly useful cases for it. That’s not the point any more than that was the point for railroads. Railroads were incredibly useful, and it still ended up where something like 50 percent of the track miles ever laid in the U.S. turned out to be useless.
So the bubble is this vast oversupply of something, largely fueled now by credit, which in turn will lead to massive business failures and write-downs of assets. And all of that credit has spread through the economy. It’s inside of insurers, inside of private credit, it’s inside of sovereign wealth funds.
We’re on the order of trillions of dollars now. And it’s probably going to hit $10 trillion before it’s done.
So all of this investment is at risk, and it’s across the entire economy and sitting inside of all kinds of — I mean, it’s even sitting inside of your individual 401(k)s. We’ve all seen the data that something like the Top 10 stocks in the S&P 500 are now 40 percent of the market capitalization of the index. So if you’re holding an S&P 500 index — my friend, you are very, very long on A.I.
Get out of my portfolio, Paul.
Kedrosky: In fact — yeah, get that [expletive] out of my portfolio.
So you’re very, very long on A.I., and this is the problem. The expression you hear all the time: It’s a one-factor world. Everything is leveraged to A.I. across every asset class, across all of retail investing.
And then the last stat I’ll give you, and the reason that matters, is we’re at record percentages in terms of the percentage of people’s net worth that is tied to equity. Half of people’s net worth is now in equity.
So follow that math through. If half of net worth is in equities, and that’s driven by the A.I. trade, people are even more overleveraged to this single trade than they ever were to dot-coms, and that creates this incredible fragility.
So all of this — this nexus of factors — is what creates this bubblish moment, which has nothing to do with whether A.I. is useful but has everything to do with why the other side of this will see a rolling series of crises that could potentially become really catastrophic.
I want to focus for a minute on the rationality of the core players — your Metas, your Alphabets, your OpenAIs and Anthropics, all of which believe that there’s only going to be a couple of A.I. winners.
Kedrosky: Right.
I mean, Mark Zuckerberg made this amazing statement where he said: “If we end up misspending a couple hundred billion dollars, that would be unfortunate. But the risk of not acting aggressively enough is considerably higher.”
And so there is this dynamic where I think they believe this is going to be a highly concentrated or oligopolistic market, where maybe it’s one model that rules them all or three models that rule them all. And so there is this race to be one of the three.
But that creates a pressure to invest way beyond what the market can collectively sustain.
Kedrosky: Right. That’s the classic competitive dynamic that has existed in every one of these capex build-outs.
It existed with the railroads. You could not afford to be left behind. You had to invest in it. And that’s why we ended up with duplicate routes between cities.
I mean, that’s the essence of the competitive dynamic crossed with a prisoner’s dilemma: I would like to slow down, but I can’t because no one else will.
And so we all continue on with this frantic competitive dynamic. And so that’s what you’re seeing incarnate there. And we saw that even a couple of weeks ago: Dario Amodei at Anthropic was suggesting that we need to — and this is a ridiculous expression — “pace the frontier.” Which sounds vaguely like we’re training for the Boston Marathon. But it’s this idea that we need to slow down.
Well, of course, everyone, in a classic prisoner’s dilemma, said: Go for it. You slow down. I’m not slowing down. And even worse, they said: Even if we all agree here, in the U.S., for frontier companies to slow down, China won’t. And then you get into this terrible discussion — the what-aboutery with respect to what China will do.
So this is the problem: There’s a race dynamic that has been created, that if you stop, there’s a fear that either my competitors will gain an advantage or my shareholders will punish me. And so as a result, the race dynamic continues, and that’s what guarantees the overbuild.
As long as you have a supply of capital and a race dynamic, you’re guaranteed overcapacity, much like whether it’s canals, railroads, electrification — take your pick.
Azeem, you and your team, over the past year to year-and-a-half, have been building out this dashboard, trying to track how much stress there is in the A.I. industry financially — trying to think through what it looks like for this to be on a healthy financial path and what it looks like for this to be a bubble.
So why don’t you walk me through your perspective — the indicators you watch most closely — and why you’re a little bit more optimistic right now than Paul is?
Azhar: Well, Ezra, when we’re looking at a technology boom — a technology cycle — we’ve got to keep two different things under consideration.
The first is: Are customers finding it useful? Are they paying for it? Are they increasingly paying for it? In other words: Is demand real? And is demand there?
And when you look back at bubbles that burst, it is almost always connected to a problem in the financing mechanism: the refinancing of shorter-term debt. It runs into a problem, and capital goes away, and you see all of the things that we associate with bubbles.
The question is: Where are we today? And where we are today is that demand is very strong from American and global enterprises.
There is not an overbuild yet. There is vast, vast undersupply of the compute needed for A.I. And the race seems to be: Can we get enough chips and power enough chips over the next two to three years to keep up with the demand?
On the financing side, we are certainly moving off the cash the hyperscalers have earned over the last 20 years or so. And we’re bringing debt on board.
But it’s not yet at the levels at which you would say: This is a historically high and complex position. So for me, today, it’s in that productive boom with the hints of maybe getting — forgive my French — a little bit stinky.
So I want to focus on the supply versus demand at this very simple level. What Paul is basically saying is that there is a strong incentive to build too much supply, and what you are saying is that there’s actually plenty of demand. So I want to decompose what we mean when we talk about “demand.”
So this is demand for A.I. usage, which is tokens — which, again, for people who are not as used to this, are: When you ask a question, the A.I. model is consuming tokens to answer it. When you ask the A.I. coder to code you a program, it is consuming tokens to code the program.
And so, functionally, what is happening with a company like OpenAI or Anthropic is people are buying token use from them. They don’t sell you Claude, they sell you tokens at the high levels. So that’s the reason tokens are an important concept here.
And so there’s first these questions of: How many tokens are people going to want, and how strong is the demand for tokens going to be? And then there is a secondary question of: How much will they pay for them?
Because the price per token is plummeting. So there’s this question of: Even if there is a lot of demand, is it going to be demand for the kinds of very expensive models that the American companies are investing in, or not?
So I’d like you to walk me through how that demand picture looks to you.
Azhar: Well, we have to start at a layer below the tokens. The asset we’re investing in is data centers, about 65 percent of the cost of which is the chip — the graphics processing unit or the A.I. accelerator.
So what I think is important to look at is: What is the demand for those chips to do the computing that, in some cases — most cases — produces tokens? And who are the customers of those GPUs?
And when we look at this, we see these hyperscalers — Amazon, Microsoft and Google — and then new labs called neoclouds. These people buy millions of these GPUs. They put them together in data centers that will now start to cost billions of dollars, and they rent those GPUs out.
And the thing that I spend a lot of time thinking about is: Is that capital expenditure on the actual productive asset turning into revenue with just enough return on investment to pay itself?
And right now, the reason we know we’re in a compute-constrained environment is that people are making extremely good returns in 18 to 24 months on an asset because they’re able to charge essentially Uber surge pricing for their GPUs. And about 50 percent of their revenues go to the two big A.I. companies, Anthropic and OpenAI, for their spending.
The other 50 percent is a mix of other A.I. companies and firms — companies directly running those open-source models that you alluded to or open-weight models on the hyperscalers’ infrastructure.
And then I think we get to your other question, which is: Well, if 50 percent of this is coming from Anthropic and OpenAI, what percentage of that is actually real customer spending?
And what I’d say is, if those ratios stayed the same for the next three or four years, we’d very much be in a world where this is someone effectively paying themselves around the back door.
But what has been happening is that ratio has been declining. Essentially, the hyperscalers are getting more and more business from real customers. And all of that, for me, is a sign of a market that is maturing week after week, quarter after quarter.
Paul, how do you see that differently?
Kedrosky: I hardly know where to start. So let me just start.
A lot of the “discussion” — and I use that word advisedly — with respect to how people talk about this stuff reminds me of that old joke: Mother comes downstairs in the morning, finds her daughter under the Christmas tree, and there’s a giant pile of horse manure. And the daughter is digging in the pile of horse manure, and the mother is like: What are you doing? What are you looking for? And the daughter says: Well, with this giant pile of horse manure, there must be a pony in here somewhere.
For me, a lot of this discussion about giant scale, with respect to tokens, is a lot like this: Digging in a pile of manure because there must be a pony in there somewhere. Because it’s this confusion of scale — a massive volume of tokens — with value.
So I’ll take Azeem’s point with respect to this prodigious use of tokens. Tokens alone don’t matter. It’s this pile of manure under the Christmas tree. What matters is tokens times price. And you mentioned this earlier: What we’re seeing over the last few years is roughly a 13-fold annual decrease in the price of tokens.
Now, that’s performance adjusted, so the answer of the frontier model companies will be: Hey, hey, hey, sure, the old models are falling like that, but not the new models. The new models are the ones that can break into banks. Those ones are still much more expensive. So those are holding on to their value.
But the problem is, of course, for most people, for most of what they do, they don’t need the much more expensive models. And as a matter of fact, most normal people get by perfectly adequately on relatively low-performance models.
A good example is this incredible uptake in Meta’s Muse product, which has dominated the app charts and everything else. It uses a relatively primitive model. It’s not using a leading-edge model, and nobody cares. It just works perfectly fine for what most people want to do.
Which lets me take it back to the beginning. One of the reasons people really mischaracterize this entire market — and this is an old venture capital kind of adage — is: Never confuse your early customers for what the real market will look like. An awful lot of what’s happened with frontier models is people confusing the early customers of frontier models for what you and I want.
So who were the early customers of A.I. models? Who really drove this? Coders. Something like 80 percent of the token usage and the revenues from the largest A.I. frontier model companies are coders or developers. And developers are really idiosyncratic for at least two reasons.
Why? One, they want the leading-edge models. They’re willing to pay up for them because they have really complicated problems. Two, they are what I like to think of as expansive in terms of their token usage, meaning that I enter a small prompt and say: Bro, build me a fitness app — and all of a sudden seven words turn into a million lines of code.
That’s not anything like what you and I do, for the most part. A lot of what normal people and normal businesses do is closer to a 1:1 relationship.
Or even compressive. Like I’ll say: I got 25 emails from Goldman Sachs babbling about something or another. Are any of those things worth reading?
And that’s compressive, right? Because I’m taking a whole lot of stuff and turning it into something small, and that feels efficient to me — or, similarly, some other email or a calendar, whatever. So the point being that normals — the people who these frontier models want to sell to now — are not really well represented by what people in the early days of the market were doing. Because coders are really unusual.
And you’re already seeing some of that: Meta’s cutting in half the number of developers in-house who are allowed to use external frontier tools. Microsoft, the same thing.
The days of token-maxxing — where you had charts internally that said who’s winning at using the most tokens — those are long over. Normal companies can’t tolerate that. People are blowing their entire budgets in months.
So my point is, we’re already moving away from the early days of profligate token use and everybody looking for ponies inside of giant piles of manure.
Azeem?
Azhar: Yeah. In his giant pile of manure, Paul also had some sleight of hand.
What Paul talked about was the frontier models and tokens. And what I talked about was the capital expenditure and the capacity we’ve been talking about, which is the GPUs. I think that they are two different things that merit two different discussions.
I mean, what happens to the frontier model labs — OpenAI and Anthropic — with this business model of ever more powerful A.I. systems that have a smaller and smaller set of users, is quite different to the question of what’s going to happen with the trillion-plus dollars or so of capital that will go into building data centers upon which American and global businesses will start to run increasing amounts of A.I.
And I think we need to separate those two out. One is really about the fundamental viability, longer term, of Anthropic or OpenAI and what they might do in the face of the pressures that Paul has articulated.
And I think the other is whether the GPUs and this new digital infrastructure that’s being built out will find uses among, predominantly, enterprises. But of course, as we rightly heard, consumers, too — through things like Muse and other agents. I think those are different questions.
I’m confused by your separating these questions. And I recognize that maybe this is an important point of confusion to untangle, because people can mean different things when they talk about whether there’s a bubble and what it would mean for the bubble to pop or to deflate.
As far as I can tell, if it turns out that OpenAI’s or Anthropic’s or, for that matter, Alphabet’s or whatever’s business model doesn’t work and that, in fact, people are perfectly happy to use DeepSeek, from China, for a lot less money, and Anthropic collapses in on itself — that is going to pop the economy.
That looks to me like it would become a contagion in the stock market. I mean, I don’t necessarily think it will throw us into a depression. But I think that would have a very, very, very big effect on valuations, on what people are willing to lend in the future.
Do you not think that?
Azhar: Well, I think what you have to do is zoom back out and ask: What is the fundamental risk that emerges with an infrastructure build-out? Which is that you don’t have enough customers who are demanding a product and paying enough for that underlying asset.
If enterprises can move to open-weight models that are being run directly by a hyperscaler, that is just the demand moving. It may hurt Anthropic’s margins, but it still monetizes the underlying capital expenditure.
And then one of the things, the second thing ——
Which are these chips — basically, the data centers.
Azhar: Which are these chips ——
Anthropic might collapse, but Nvidia will be fine.
Azhar: But even so, Anthropic has many, many degrees of freedom. I mean, in the last 24 or 48 hours — before we recorded this — it cut prices and released even cheaper, more efficient lightweight models. These are still models around which they have a pretty decent margin.
They will be doing the things that software companies have done for decades, which is trying to construct reasons for you to use their software beyond the core capability. That’s why they have Claude Cowork, and they have Claude for lawyers.
And that’s why I think this question of: If Anthropic disappeared tomorrow, overnight — well, I can’t quite imagine the circumstances under which that would happen. But what I am saying is that you’re seeing the end customer demand, and that’s the thing without which this will always end horribly.
I want to ask you both this question: What, for you, counts as the bubble popping?
Because there’s a world here where you have, say, a giant stock market correction, but the data centers are still being used. It sounds to me like you’re saying that’s not a bubble popping — that’s just a correction or something?
Azhar: Well, historically, when bubbles have popped they have resulted in two things. One is a huge decline in the level of investment in that particular asset. And they have been accompanied by a significant crash in a relevant asset market — whether it’s the stock market or something else. And that is the test that I would use.
I think we may end up in a world where there are grindingly low and embarrassing returns for credit funds and for hyperscalers because of the investments they’ve made. But that degree of pain is not a bubble bursting. That is just: Somebody made some bad investments, and they bought high when they should have bought at a lower price.
I mean, I think I would use a very, very classic historical definition, which is that we need to see the level of investment drop by 50 percent, and we need to see a really, really substantial — and typically it’s been 50 percent or more — correction in whatever the relevant market is.
Paul, same question to you.
Kedrosky: OK. So I disagree on almost every point that Azeem is making.
These are arbitrary definitions, and I think — back to my original point: Something like 50 percent of people’s current net worth in the United States is tied into equities. Most equities are currently tech. Most of tech is currently A.I. That’s going to only get worse as these two monster initial public offerings come to market over the next, I don’t know, two months or whatever.
OpenAI and Anthropic?
Kedrosky: OpenAI and Anthropic. And just to put that in context, by the way: OpenAI, Anthropic and SpaceX, if you want to include it — combined, those three I.P.O.s are larger than all post-World War II tech I.P.O.s combined. This is a really unusual moment.
That’s a crazy stat.
Kedrosky: It’s a crazy stat. And not only that, the combined value of those companies is larger than the value of the Nasdaq market capitalization at its peak. Like the bubble is sitting right in front of people, and they want to say: Well, I expect you to come with blond hair, not brown hair.
Dude, just because it doesn’t look the way you want it to look doesn’t make it any less fragile.
So you could imagine, let’s say, just for the sake of argument, the Anthropic I.P.O. gets pushed out. It doesn’t happen in November — it happens next year. There will be a sharp revaluation in their ability to raise funds, as well as in the companies supplying them. Because everything is levered off the valuation of these companies that are currently sitting in private markets.
So I like to say: The failure here is overdetermined. There are so many ways it can all break, and break really catastrophically.
I’ll give you another example: What’s called the Shiller cyclically adjusted price-to-earnings ratio, which is sort of a longer-term measure of where U.S. equities sit: We’re sitting at levels we sat at just before the crash of 1929. We’re sitting at absolutely, historically unprecedented levels in terms of market-wide valuation.
And this is the problem: People lack imagination and simply refuse to engage with the broader data about what’s actually happening in the world because they say: Well, this doesn’t look like the last bubble, so we must be fine.
Well, let me ——
Azhar: But I think ——
OK, go for it ——
Azhar: I’m just going to come in directly, Ezra. I think that it’s very easy to use very colorful language. I’m always mindful of ——
Kedrosky: No, this is data. It’s not colorful language. It’s actually numbers. These are numbers.
What’s your response to: We’re sitting at historically unprecedented numbers in terms of the cyclical price-adjusted earnings? Is that good in your book?
Azhar: One of the responses about the Shiller CAPE ratio is that the Shiller CAPE doesn’t say anything about timing. What it does do ——
Kedrosky: No, because it’s an adjusted long period of horizon. It’s not just a history ——
Azhar: It’s a historical, 10-year lookback that can sometimes understate underlying changes in the regime of the economy and the market ——
Kedrosky: OK, sure.
Azhar: But there’s something else that’s going on, which is, I think — a lot of what you’re describing, Paul, is already priced by equity investors, and it’s being priced by the credit markets.
We’ve seen it just in the last few weeks. We’ve seen an I.P.O. for a data center company in Australia get pulled ——
Kedrosky: That’s right. Yes ——
Azhar: Because investors weren’t happy with what they were being promised ——
Kedrosky: That’s not why it was pulled.
Azhar: Well, why was it pulled then?
Kedrosky: It was pulled because of this unprecedented moment where $2 trillion I.P.O.s are in orbit around us waiting to come to market and are sucking all the risk capital out of the market as people say: Yeah, this looks really cute, this I.P.O. we have, but we don’t have any capacity to do that right now.
OK. But I think that, in a way, actually goes to Azeem’s point.
When you talk about these massive I.P.O.s that are circling — we’re talking about OpenAI and Anthropic. And OpenAI has gotten pushed back a little bit for when they seem like they’re going to launch, actually, into the public markets.
The thing that is interesting about that, from the perspective of it’s not all a bubble, is that sophisticated investors in mass numbers who really, really do have skin in the game here are looking at these and saying they make sense.
So when you hear this from other people, Paul — and you’re an investor, and you think about these things — what mistake do you think all these people are making?
When you say it’s just math, it’s clearly not just math, because everybody else can do math.
Kedrosky: Oh, of course ——
It is your interpretation of the math. So where does your interpretation of the math differ from the market?
Kedrosky: Sure. So you have to think about why these kinds of I.P.O.s are happening and why people are piling into them. It’s partly in anticipation of what they think could be a fantastic business with a long-term franchise — da, da, da. But it’s also because they think that when it comes time, and when the lockup expires, they can sell the stock to someone else who thinks that.
This is the essence of — going back to the very beginning: Why do I.P.O.s exist? I.P.O.s used to exist to allow you to raise money for a company that needed to grow. These are not that. These companies have done their growing.
They’re probably actually going to be, I don’t know, $100 billion in run rate revenues by the time they come out, or $70 billion or something like this.
So these are exit events. These are people inside the company saying: I don’t need my stock anymore. You can have it.
So they’re selling their stock to people on the outside because, at this valuation, they’d rather have cash than Anthropic stock, or cash than OpenAI stock.
At that valuation, I’d like to have cash, too.
Kedrosky: Yeah. Sure, absolutely. Hand it over.
But the point being that this is not a gesture where they’re saying: You know, I really like our prospects with the company.
There’s going to be selling shareholders across the board. Because this is an exit event, not a financing event.
Sure, but it is a gesture where other people are saying: I do like your prospects as a company, and I want to own this.
Kedrosky: Sure. The same thing was said in anticipation of the SpaceX I.P.O. And we have seen what happened subsequent to that — all of this fervor before it turned into a light roller coaster decline.
So my expectation is you’ll see the exact same thing because these are, in market terms, exit events. You’re going to have a wall of selling shareholders showing up post-lockup, selling it to anyone who will show up and want to take the shares from them.
And that’s why this needs to be understood for what it is: an exit event. People are buying in anticipation of selling to you.
If you want to, it’s like the old line about the poker table: If you don’t know who the greater fool at the table is, it’s you. This is that, and I don’t understand why people don’t see it that way. Because that’s what makes this so unprecedented.
Azhar: I’ll just come in on that. I mean, I think what Paul describes is really what many I.P.O.s have been like for many, many years. And when we take a look at Anthropic and whether it can be a sustainable business, we probably need to look at what it will show in its SEC Form S-1.
The S-1 is a document, essentially, for what this company is when you’re going to buy it at I.P.O. — to look at what’s happening with its revenue trajectory and its margin trajectory and the breadth of its customer base and whether customers are sticking with it, whether those accounts are growing.
Beyond what has been, frankly, a lot of patchy reporting and leaks — some of which are accurate and some of which are inaccurate, but not much of which gives us a cleaner picture of the company.
And I think that draws out one of the really big issues we’ve all had to deal with over the last two or three years, which is: These companies have gotten very big, but they are still private companies, so they don’t really need to disclose much, if anything at all. And what comes out is a smattering of rumor, scuttlebutt and the occasional well-sourced leak.
They have essentially been structural to the A.I. trade so far, and they have been, as a consequence, the thing that has spun the valuations of the wider market, whatever it is now — 38 percent tech. But we have not been able to inspect, the way we normally can with public companies, what’s really going on.
And I think what happens with the I.P.O. is that we won’t get every answer, but we get some of the answers that we need about these businesses. And this ——
Kedrosky: We have a partial answer now, though, just on that point — which I think is amazing because it came from — what’s that thing with the horror movies where the killer is inside the house? This is a little bit of one of those, where Anthropic accidentally leaked who one of their largest customers was.
They said: We needed to have — what was it called? — a mega account executive or something like this, where they were trying to hire one.
Then they quickly realized that by saying who their largest customer was, you could triangulate and say this was for Meta. So this was their account executive to handle sales to a single company. It worked out that Meta accounts for something like 10 percent of their revenues. And so this goes back to the ——
Meta is 10 percent of Anthropic’s revenues?
Kedrosky: That’s right. So ——
Azhar: But that was in 2025, right?
Kedrosky: Right. So ——
Azhar: And let’s just be clear ——
Kedrosky: And just to be clear about why that matters, though ——
Azhar: No, but Anthropic’s revenues — but why it matters is that Anthropic’s revenue in the first quarter of this year was bigger than the entirety of last calendar year.
Kedrosky: Sure.
Azhar: So what we’re doing is we’re looking at a rearview mirror and saying: Oh, we hit a squirrel 700 yards ago.
What we need to do is look out of the front window — the windscreen.
Kedrosky: This is why these companies are so unusual. We have never seen companies of this size go public with such a small number of customers. The concentration of their customer list — it might be eight companies that make up 80 percent of their revenues. That doesn’t happen.
I was an equity analyst years ago, and this would be a colossal source of risk for me. I’m looking at it, and I’m saying the dependency on such a small number of customers who are mostly using it for the same thing is historically unprecedented, to have such customer concentration.
So what they accidentally leaked was the fragility of the business in terms of its exposure — not just to a single customer but to a single customer who was doing that same old thing of writing code.
And this is the point I want to make. That’s what makes this so unusual.
Azhar: The beauty of what Paul has just said, which is: Eight customers, 80 percent of the business — that’s 10 percent per customer. That would have to be disclosed in the S-1.
So we have, actually, a very, very falsifiable prediction that we will be able to check when the S-1 comes out, if Anthropic does successfully go public. And then we’ll know: Was it really eight companies representing 80 percent of the business?
And that’s really helpful, because if Paul’s right, that’s a bit exotic. It’s a bit spicy.
I think Apple is a really interesting company here because, unlike Alphabet, Google, unlike Meta — obviously OpenAI and Anthropic are A.I. companies. Microsoft was a major OpenAI investor and is now also trying to make its own frontier A.I. Oracle is heavily invested here and is now just right above junk-bond status.
Apple just decided not to compete in making a frontier A.I. While their competitors are now spending $100 to $200 to $300 billion in capex every year, Apple is spending, what, $15-ish billion in capex per year? Their stock this year has performed better than any of these other companies’ stocks.
Does it imply anything that Apple, in being the one major player here to not try to compete in the frontier model Olympics, seems to be winning?
Kedrosky: That’s the one ——
Azhar: Apple doesn’t ——
Let me start with Azeem here.
Azhar: Apple has a business model. That means they don’t need to run a frontier model. We buy their phones, we buy their laptops, and then we might subscribe to the odd movie or something like that. So they make money regardless of the A.I.
But I also think there’s a really important cultural dynamic. Apple cares a lot about detail.
I remember this story of Steve Jobs, when he was rolling out the iPhone many, many years ago. He phones the then-senior vice president at Google who is responsible for the partnership and blasts him out because they’d used the wrong color yellow on the then-Google logo.
Apple really cares about detail, and they care about control. And large language models are powerful in many ways, but they’re not about control. They can get very, very hapless.
And I think that one cultural reason Apple would have held back, even though it’s so well-positioned to be the personal agent — it’s got the device, it’s got the privacy — is that it won’t want the mess that a large language model will ultimately deliver if you’ve got a billion people using it. Which is what Apple’s estate looks like.
Paul?
Kedrosky: So yeah, I don’t think Apple was very smart about this. I think they mostly screwed it up and got lucky.
So they tried various times to make forays into A.I. I mean, they had teams, they disbanded teams, they were going to integrate it with Siri, they weren’t going to integrate it with Siri — and they didn’t ever get anywhere.
What they benefited from, though, was that by their inaction, they were a classic free rider. They got to free ride on everyone else’s gigantic expenditures. And they come out the other side smelling like roses because they didn’t actually have to make the expenditures. They can use what’s being built by others and sit on top of it.
But I think, writ large, we’re going to see a lot of that. Because people are burning money on a bonfire — which can benefit lots of other companies. So Apple is sitting on top of it, not having to build models.
Look at companies like Cloudflare, for example. I mean, here’s a company that was largely forgotten — it was basic internet infrastructure. Its stock has tripled in the last, what, 18 months because suddenly people are like: Wait a minute, wait a minute, wait a minute. If we have all these bots ranging around the internet, that’s like a new form of attack that we need to make sure that we have protection from. Who’s out there that can deal with all of this new traffic?
And so a lot of this investment that we’ve seen out there will turn out to be beneficial to a whole host of companies whose names are not OpenAI, Anthropic, Grok or whatever else, and I think that’s a really important point.
And that goes back to the days of electrification in the 1920s. Who were the primary beneficiaries once the utilities all began collapsing in the crash of 1929? Well, of course, it was the people building things who required electricity — air conditioner manufacturers and others.
And so, yeah, for sure we’ll see a lot of this accidental piggybacking and free riding on top of this.
Azhar: But it’s not accidental piggybacking and free riding ——
Kedrosky: Apple didn’t plan this ——
Azhar: It’s called complementary services and complementary businesses. That’s what economists, when they’re being very moderate with their language, describe in the process, which is: A general-purpose technology shows up, it continuously improves, and then it starts to create the need for complementary services — whether it’s maintenance or whether it is repair or whether it’s something you can now do with the technology.
This is the normal evolution of a general-purpose technology as it finds its way in, and in some cases, it’s a bit of free riding. In other cases, what has happened with Cloudflare is that we’ve been able to, across the economy, automate so many bots and A.I. agents that businesses on the internet are turning to Cloudflare to try to protect them if they can.
Let me ask something about these big companies that has come up a lot in the bubble debates I have followed.
The major companies we are talking about here — your Alphabets, your Metas, etc. — they just have a lot of money. And so they’re doing this from a very healthy balance sheet, which is not the case sometimes for these railroad companies, etc.
When Zuckerberg says it would be unfortunate if we misspent hundreds of billions of dollars, what he is also saying is: We can misspend hundreds of billions of dollars. It’ll be fine.
And one argument I’ve seen is that one reason you don’t really need to worry about this is that it is being driven by companies that can absorb losses on their own balance sheet without contagion in the rest of the economy. You’re not dealing with something like derivatives in the 2008 financial crisis.
Paul, you’ve thought a lot about the private credit dimensions. I mean, there has been a movement toward debt in some of the way this is being financed now.
How do you think about the question of: Are we — the broader society — protected from this? Because these companies have ungodly amounts of money to lose if they so choose.
Kedrosky: No, we’re not. So this is a funny thing.
A year ago, people told me to stop babbling about this stuff because these are smart companies, they have prodigious cash flows. Who are you to tell them what to do?
And I was like: Oh, shucks, you’ve got me there, so I’ll stop saying this.
And then, of course, we’re now a year later. What does the data show? Spending is moving from cash flow into debt.
So you put up your hand, and you say: Hey, wait a minute, wait a minute. Remember that thing you said last year? It’s no longer true. So should I worry now?
And it’s like: No, you don’t need to worry now because they have pristine balance sheets. These are great credits.
And I’m like: Ugh, [expletive] — my timing is wrong twice on this.
And then, of course, to be serious, what happens whenever you have a pristine balance sheet? What Wall Street is very, very good at is making it less pristine because they’ll continue to raise money for you, and raise and raise and raise for you. Because the fees are tremendous. And as they do so, your credit rating declines.
And we’re seeing that with Oracle. We’re seeing that we’re on the cusp of that with Microsoft. We’re on the cusp of that with some of the other hyperscalers, because that’s what Wall Street does.
They see a pristine balance sheet, an “unlevered company” in the term of art in the industry. And they’ll continue to burden it with credit.
And so the way that these companies try to get around it — Meta is probably the best example — is they use somewhat more exotic structures so the debt doesn’t actually sit on their balance sheet. They use things called special-purpose vehicles. They use things that are off-balance-sheet financing.
They’ll say: Hey, hey, hey, that’s not my debt. I know I control it, but it’s over there with that funnily named thing that sounds like a New Orleans bagel. That’s somebody else’s debt.
And so this is all just a shell game. And the reality is that rolled up, we’re now at a point where tech, whose main advantage historically was fast growth, no debt — tech is now the largest issuer of investment grade bonds in the United States.
This is an industry whose main claim to fame, other than growth, was that they were relatively clean. They’re now the exact opposite.
These are hugely debt-encumbered companies, increasingly, that are dominating the issuance marketplace to the point that it’s even having an impact on sovereigns. So one of the reasons we had the bond market freakout over the last two or three weeks was because of the role of hyperscalers in issuing credit at the same duration as the tenure in the United States. Which, at the margin, was actually outcompeting the U.S., which caused yields to have to go up on the tenure.
So that’s how crazy things have become. So yeah, no, it has completely transitioned. Which is going to have huge consequences down the road in terms of this debt overhang that starts to look increasingly like Japan circa 1990 — all these corporations with giant balance sheets loaded with debt that are trying to work it off.
Well, the other side of that, Azeem, is that it’s not just debt that is being crowded out where these hyperscaler companies are competing with everybody else for money. It’s also construction, it’s also real materials.
I’m sure you’ve seen these charts going around that imply there’s real crowdout happening. That yes, we are building, we are pumping money into A.I. capex, but home building, in a country that does not have enough homes, is going down. Spending on manufacturing plants and factories is going down, or at least the share.
And so there does appear to be a movement from investing in other parts of the economy, at a construction and capex level, into investing in A.I.
I guess if A.I. works out — great, that’s fine. But if it doesn’t, or even if you just think these other things are underrated, as say, maybe, for homes — homes are a more stable investment. They don’t give you returns like investing in Nvidia does, but they’re pretty important, and I have a cleaner way of explaining why more homes would make more people’s lives better.
How do you think about the crowdout question here?
Azhar: A year ago, back in the summer of 2025, I wrote about how I assessed whether we’re in a boom or a bubble. And one of my observations was that at the moment, the hyperscalers are spending out of their own cash flows and their own balance sheets.
As soon as that starts to change, and they go to external finance, you introduce lots more risk. And my threshold for more risk was about 30 percent, which is about the level that we are approaching now at the end of 2026. These things are not set in stone.
The crowding-out problem is a real problem, and it’s really a complicated and thorny issue. I think one place to think about is what happened in the Netherlands in the 1970s, when they discovered a lot of offshore natural gas in the North Sea, and the entire economy oriented around gas extraction. And you got this thing called Dutch disease, where effectively that part of the economy was sucking up all of the activities.
And there could be a risk that will happen in the U.S., if it isn’t already happening right now. And I think then the question is: What do you do about it? And how do you actually adapt to that?
And I’m afraid my only answer is that it requires something of a policy stimulus to make sure that those sectors get the activity that they need. Because I’m not persuaded that market forces on their own would do that.
That seems like a tough prescription at a time when, I’d say, we have both very erratic leadership, but also our borrowing costs are going pretty high amid incredible levels of debt.
Azhar: Well, look, your politics are your politics, and you get to vote every four years and make the choices that you make.
[Laughs.] And then live with them.
Azhar: Right, and then live with them. [Chuckles.] But it doesn’t have to involve subsidies. It can involve other types of policy interventions.
It’s a really difficult problem. And it will continue to happen as long as the returns on the other investments are higher and as long as that is creating a supply shock for electricians and construction workers and HVAC technicians and all the other things that are needed in data centers.
There is a “but” to all of this, which is that, of course, there’s a lot of upskilling going on. There are many, many electricians being trained. There are many HVAC engineers being trained. There are many construction workers being trained. There are construction firms that are growing and that will start to have spare capacity.
And I don’t know how that ends up. But in a year or two or three, that capacity and those skills will be in the economy and able to be used.
So I want to ask this not in terms of stock markets and financial flows: What does this economy look like in which companies and people — and governments, for that matter — are using enough A.I. to make these investments, writ large, pay off? Even if some companies may not end up being in the winner’s circle here and may fail?
So Azeem, you spend a lot of time thinking about this. If you were to try to paint a picture of what things look like at normal companies — your GEs, your Coca-Colas — in this world where the A.I. bet pays off, what does it look like?
What is different? What is not being spent on if so much money is being spent on A.I. tokens? Like, what is happening here?
Azhar: When I talk to bosses, and I talk to them across Europe and the U.S. — dozens every year — I hear this same sort of thing, which is: We’re making progress with A.I. It’s harder than we thought. We’re going to keep doing it, and we’re starting to get results.
And the more confident ones say: We’re getting more and more results.
But a large part of it is coming now — I think will come, unless businesses grow very, very quickly — from just a slow reduction in the replacement of workers. Companies that are getting more productive with A.I. will be delivering more dollars per employee. And if they’re not growing faster than that productivity growth, then their headcounts will start to fall.
But we should just add that there are very few A.I. native companies. When we think about electricity, the electricity-native companies were companies like the Ford Motor Company, with its end-to-end production line. But most of America didn’t get fully electrified for another 15 or 20 years after that.
And I don’t think I can really point to a Main Street firm that is outside of Silicon Valley that you could say: This company has become A.I. native. What you can see are bits of the business that look like they’re using a lot of A.I. ——
OK, but if the A.I. native company that you’re talking about is a company where you’ve managed to replace people with A.I.s — now you seem to be really quickly tipping into the much-worried-about world in which A.I. is net negative for jobs across the economy.
Unless you believe there are all these new companies starting that are somehow very human-labor intensive, even though, in theory, the new companies starting are going to be more A.I. native, so less human-labor intensive — so is that actually a good world?
And then if that’s the world, who’s spending more money in this world where people are losing jobs?
Azhar: Well, I think there are a number of different things going on. One is that there are still things that the A.I.s won’t be able to do. And those will be the layers of humans doing verification, doing the last sign-off, doing the things that the models cannot do well. What would typically happen is that those people will end up being paid more and more, and better and better, as a consequence of sitting at that choke point.
I think that the second thing that you will start to see is a whole class of roles that we can’t imagine within companies — whether it’s things like the webmaster, which didn’t exist in 1993.
And then I think the third thing is really the cultural dimension of a general-purpose technology. This is where Paul was describing the G.P.T., is that they tend to construct entire new classes of work and behavior that we hadn’t been able to put our fingers on.
And if you just think about Henry Ford — well, because of Henry Ford, we ended up with motorcars. Because we have motorcars, we ended up with people having suburbs and suburbs to vacations and vacations to all the services that sit around them.
Now, I’m not blasé about that. I think the challenge we may find is that it takes longer to create those new roles than anything else that happens.
But the other observation I would make is that my time horizon for this runs in five, seven, 10 years — not one or two, not two or three. Because I do think it just takes time for management teams to absorb what it is they have to do with this technology.
And it’s just not the case that you get the best model from OpenAI and say: Fix my company — and it fixes it. These things take time. And that’s what I hear when I talk to bosses. However fast they want to move, it just takes time.
Paul, how do you see this?
Kedrosky: We’re basically repricing labor, right? I mean, this is the fastest adopting technology in history, and its fundamental function is to reprice cognitive labor.
And I mean, this is the deep struggle. Because the article of faith — the West Coast ideology on this stuff — is technology always creates more jobs than it destroys. And so it becomes just a thing that people repeat — like a mantra that they hope will become true.
And the reality is we’ve never seen a general-purpose technology that had such broad applicability, was adopted so quickly and was repricing labor at the same time. And when I say repricing labor, I mean it’s making it vastly cheaper to run things like call centers, for example.
And as a result, this change is going to be destabilizing long before we figure out how to turn them all into — I don’t know what the equivalent of webmasters is.
And so this is the problem: It’s happening so much faster than any prior G.P.T. This arm-waving about creating more jobs than it destroys belies the lived experiences of people who are affected by this stuff.
And so it can quickly become just some strange mantra like: Don’t worry your pretty little head — everything will be fine.
So no, I think the disruption will vastly outpace our ability to create new positions.
This has also struck me as another form of economic risk for not just workers but actually A.I. and for some form of this bubble popping or deflating, if it is a bubble.
Because you can imagine a world where the problem is not whether companies are willing to adopt A.I. That actually — particularly under more economic strain — interest rates keep going higher, maybe there’s a recession of some sort, and companies begin moving very aggressively to replace human labor — which needs health care and unionizes and has opinions — with A.I.
And then there’s a huge backlash. There’s already a huge A.I. backlash. People are organizing against building data centers. They don’t want them where they live.
But if you know anything about the history of occupational licensing or building homes or anything else, it’s actually not that difficult for, say, teachers to organize and say: No A.I. in the classroom. Actually, education is off-limits for A.I.
Doctors all of a sudden start to organize. Maybe there are a couple of bad events where people are getting bad diagnoses off A.I., and so it’s locked out of hospitals.
Anthropic, in its risks to their I.P.O., says existential risk to humanity is possibly one of the risks of this I.P.O.
If there’s some major control incident where, instead of hacking Hugging Face or OpenAI, it hacks Wells Fargo, it hacks something in China, and all of a sudden regulation comes in not a little bit but all at once.
All of those things seem like real risks. The possibility of discontinuous regulation seems higher to me than elsewhere. It also seems — I don’t want to say inevitable, to me — but quite likely, to me.
So what do you think about that?
Kedrosky: This has been my complaint from the very beginning of the process by which these companies have launched models.
They’ve essentially been conducting unlicensed drug trials on the public. In the sense that these are hugely consequential technologies that are launched in a way that, were they a drug company, you would immediately have the F.D.A. on your back.
And yet, because of their technology, companies can do these mass trials: Oh, by the way, we’re going to transform all of education. We’re going to launch these things that can penetrate government websites and other things.
So this idea that you can do this with impunity in technology, because we live in the timeline where, so far, these things haven’t killed us, is just ridiculous and indefensible.
And I think we’re inevitably headed toward a world where there’s going to be — after some kind of terrible event — some kind of post-hoc regulation imposed. But it will only be after consequential damage.
And I’ll give you a specific example. This is the way I often try to characterize it for people. Because they’ll say: Well, what’s going to cause that? As if you can just come up with this stuff.
If you think about it, there are probably five, 10, 15 different sorts of scenarios we can come up with that might lead to regulation, none of them probably having a probability of, let’s say, maybe more than 5 percent.
So let’s say I have six events, each of which has a 5 percent chance of happening. What’s the likelihood that any one of them happens? It’s about 20 percent.
If I have a dozen events or possible scenarios that are all bad, but each of which has no more than a 5 percent chance of happening, what’s the likelihood that one of those happens? It’s more than 50 percent.
We have way more than a dozen ways that these things could cause trouble and end up requiring regulation. So looking at it that way, it’s almost inevitable. It’s like a 50 or 60 percent chance that something happens that drives regulation. And yet people continue to sit around and say: Well, you know, this isn’t that likely. Or: That isn’t that likely.
So no — regulation here is inevitable. It will take the form of something like the F.D.A. Maybe like some kind of Atomic Energy Commission — because it’s just as consequential and because there are so many different ways for them to have broader consequences in society.
And it’s amazing to me that people sort of glibly pass that off.
Azeem, how do you see this political risk dimension?
Azhar: Well, I think, when you look at industries — industries grow quite well when there’s a little bit of regulation. The pharmaceutical industry is a great example.
And when we think about who’s going to pay for this, the A.I. companies — from the model labs through to the hyperscalers — are going to be making a lot of money. So it’s fairly clear that there is money in the system, and there are firms that can cover some part of the bill.
And I think one of the things that’s really exciting about the data center protests that you’re seeing, particularly in the U.S., is that actually it is people expressing their sense of unease about this technology, the way it’s manifesting itself.
And the way to look at that is that is a democratic expression. That is an expression of will. That is an expression of uncertainty. And what you would hope to come out of something like that would be a better social license that people are more and more comfortable with.
Now, I suspect that within the U.S., because you have state-level governance power, that some of this may be done on a statewide basis. What’s happening in the U.K. is that there is an attempt to get people developing data centers to reinvest annuities into the local community.
Because one of the things that we don’t really talk about much is that a huge data center might employ very, very few people, indeed. I mean, it may be billions of dollars, but it’s a handful of jobs, and that’s not really ——
Which is crazy — two to six people for a multibillion-dollar facility.
Kedrosky: Well, and if you do the math in terms of the cost per jobs created, it’s often — I was looking at one in New Mexico recently. It’s $4 million in terms of subsidies per job created, leaving aside the cost of the center. If you look at the tax abatements and the required improvements to local utilities, the numbers are insane.
Azhar: So none of that works long term. But the useful thing is that there’s money in the system. Because there are profits being made up and down the stack. And one would feel that because of that, there’s an obvious place for that bill to land.
And that bill can land through — I don’t think self-regulation works. And I would agree that we will need to find — whether it’s local, national or international — real rules about how these systems work and the harms and the damages that they might cause.
But the thing to me that’s interesting in that — of course, there’s going to be regulation. That’s clear to me.
The range of how that could fall — all the way from pharmaceutical regulation, where what we want is for pharmaceutical companies to produce safe drugs, although people may argue we’ve made that too slow, also. All the way to, say, nuclear power regulation, where, fundamentally, what we’ve ended up wanting is no new nuclear power plants.
And depending on what happens with A.I. and given how negative people are on it already, one of the possibilities here is that it could swing such that the answer on a bunch of things is simply: No, we don’t want to be replaced by robots. Which, actually, when you say it aloud like that seems like a fairly straightforward thing and probably the kind of thing that people are going to express in a democracy.
Or: No, we don’t want robot agents crawling all over the economy. No, we don’t want to be called endlessly, as a restaurant, by your Muse agent trying to get reservations everywhere. Because now you’ve just made it functionally impossible for a restaurant to ever answer the phone again with a human being.
And that the collective expansion of all this — I mean, I know people who are managers at different firms, and they’re just saying: I’m just drowning in A.I.-produced reports all of a sudden. My email is just full of obviously A.I.-written inquiries, which is making my responsiveness all across the board go down.
I’m not just saying: Will there or won’t there be regulation? But I think one risk is if people decide they don’t want this in fairly large swaths of the economy or used in some of the ways people are expecting.
There’s a very good piece by John Herman in New York magazine about the beginning agony of companies that all of a sudden are absorbing tremendous amounts of Muse agent inquiry.
And we’re not prepared for everybody to have an agent calling you as a robot, trying to negotiate down the price or get a reservation or sell a concert ticket or just whatever it might be. And people may actually not like it.
Kedrosky: Yeah.
Azhar: They might not like it. And the trouble is there appears to be some revealed preference here, and that picture is not even.
So only about 2.5 percent of Americans pay for an A.I. subscription, which — considering they’ve been available for three years — is a very slow takeoff.
I love this stat in one of your newsletters recently, that a larger number of Americans pay for a G.L.P.-1 than for A.I.
Azhar: Well, yeah, because one is really quite useful, and you feel the impact within a matter of a few weeks.
But hundreds of millions of people in America and around the world do use A.I. through the free version, and when companies like Meta start to put these tools in front of you, I’m sure people will end up using them.
Even if they vote with their complaints and with their social media posts saying they don’t want any A.I., either it will become inescapable or they will still have it on their phone, and they’ll be using it. And I think we have to just contend with that possibility.
I think we also have to realize — and this is where we think about technology as being this gigantic, unpermissioned experiment —which it has been for 40 or 50 years. That what you’re describing about robocalls from A.I. agents into hotels and restaurants — or as my orthopedic surgeon was telling me, he just gets long, 10,000-word emails from patients who have clearly given ChatGPT their medical notes.
This is unfortunately going to happen, and that cost has now been pushed into the system, and we’re going to have to figure out how to deal with it. I now pay — it’s not a lot — about a buck a month to my phone operator to screen out spoof phone calls that are essentially robocalls.
That’s what’s going to happen. And we’re going to call that gross domestic product, those additional costs.
Kedrosky: I think, though — your comment, Ezra, really reminded me of this: We have this idea in Western capitalism that it’s always about efficiency.
In a weird way, societies require friction. And when you take away frictions, even frictions that you didn’t realize existed, all of a sudden, it has consequences. There have been examples of people using large language models to launch claims in small claims court, for example.
Do we want that to have less friction? Do we want more of that? Do we want your example of the restaurants?
And we have all of these frictions that exist in society, and blithely removing all of them is creating, in tech parlance, almost a denial-of-service attack on modernity, where all of a sudden, we’re having stuff launching at us from every angle, almost like Russian bots attacking a website. Except it’s coming from all angles because of the increasing number of bots and other things.
And I think that’s what people are responding to. It’s this out-of-control feeling as all of the frictions drop away. And suddenly, I can’t get access to things, or nothing works the way I’d expect.
So simply dealing with the reduction in frictions has huge consequences across society, and I think people are reacting to that.
And what’s going to happen is it’s going to be these two A.I.s bouncing things back and forth. And it’s just going to be a more and more and more unpleasant economy.
Which is why I do think there are going to be large areas where people just say: Maybe not.
Let me end on this question, which is first, for you, Azeem. What would you need to see to start thinking: Oh no, this is getting pretty bubbly?
And then to you, Paul, the reverse. What would you need to see to think: Oh, maybe I do have this wrong, and these investors have it right, and there’s a lot of money to be made here?
Azhar: Well, there are so many risks that we have talked about, and there’s all the complexity and fragility in the financing. The thing that keeps this on the other side of healthy is whether real customers are spending real money and getting real results.
So the thing that would make me uneasy would be if I saw a couple of quarters where the amount that American enterprises were spending on A.I., that the growth rate was halving, say, two quarters in a row. That would suggest that they’ve filled their appetite, and there’s not going to be the revenue to keep this on the right side of healthy.
Paul?
Kedrosky: There are so many. I mean, I’m convinced I’m wrong about most things, so I’m always happy to find more things I’m wrong about.
So I just look at economic data. If we began to see, I don’t know, productivity growth go well beyond trend, that would be striking.
I mean, so far we’re sitting at around trend data with respect to U.S. productivity growth. There’s no sign of A.I. improving productivity. If we saw participation rates sharply increasing, that would be something that would help to tell me: OK, A.I. is an inclusive technology in bringing people back into the economy. That would really help.
If we saw G.D.P. growth going back above trend — we’re still sitting around the levels of the post-World War II trend, at the low end of it — any of those things would tell me that something big was going on, like a true productivity shock in the economy that warranted the kind of expenditures.
So if any of those, and probably, the three of them at once happened, that would tell me I was wrong. If those productivity numbers come through, then fine, everything changes. But if it doesn’t, you’re looking at something that looks more like the 1930s than it does like the 2000s.
All right. I think that’s a good place to end.
So always our final question: What are three books you’d recommend to the audience? Paul, why don’t we begin with you?
Kedrosky: Sure. “Middlemarch” by George Eliot is my go-to on thinking about normal people being swept along by the winds of change as a major technology revolution is taking place. It’s set at a time when the Industrial Revolution was taking hold, and I think it’s a tremendous book about that.
So the other one that I like to recommend is “Liberty’s Dawn” by Emma Griffin, which is a collection of the average workers’ biographies and notes from during the Industrial Revolution. It’s just a wonderful, wonderful look at what it was like in the middle of that wrenching transition. And as people would talk about economic data and everything else, what it was like to live through it.
What did it feel like? What was the pain I felt? And what are my hopes for the future? It’s just a wonderful book about that.
And the last one that I always point to is about Samuel Insull, a largely forgotten figure. He was probably one of the most important entrepreneurs of the 1920s. His empire in electricity shows what happens — building an empire on top of a G.P.T. and even doing that successfully can still end up in collapse and calamity. But it’s a wonderful story. So “Insull: The Rise and Fall of a Billionaire Utility Tycoon,” by Forrest McDonald.
Azeem?
Azhar: So my three books — the first is “The Day the Universe Changed.” It was written by BBC science journalist James Burke back in the 1980s. I read it as a teenager, and I’ve just reread it, and it is all about how, as we develop new types of knowledge, we start to rethink our role in the world, what matters to us and how unsettling that is. And he walks us through 1,000 years-plus of that through Western and Asian history.
The second book is “Fathers and Sons” by Ivan Turgenev, which is a book written in the 19th century. It’s at that moment where the ideas of science and modernity are making their way into Russia, and you have the generational conflict between the more traditional generation and the younger generation, and what stays as a bond between them — what even the most recent breakthroughs in science can’t tell you about what it is to be human. I think it’s very relevant.
And then the final one, which I’m just rereading again, is “Candide” by Voltaire.
Azeem Azhar, Paul Kedrosky, thank you very much.
Kedrosky: Thanks, Ezra.
Azhar: Thank you.
You can listen to this conversation by following “The Ezra Klein Show” on the NYTimes app, Apple, Spotify, Amazon Music, YouTube, iHeartRadio or wherever you get your podcasts. View a list of book recommendations from our guests here.
This episode of “The Ezra Klein Show” was produced by Rollin Hu. Fact-checking by Kate Sinclair, Mary Marge Locker and Julie Beer. Our senior engineer is Jeff Geld, with additional mixing by Aman Sahota and Isaac Jones. Our recording engineer is Aman Sahota. Cinematography by Kyle I. Kelley. Video editing by Steph Khoury and Dani Dillon. Our executive producer is Claire Gordon. The show’s production team also includes Marie Cascione, Annie Galvin, Kristin Lin, Emma Kehlbeck, Jack McCordick, Marina King and Jan Kobal. Original music by Pat McCusker. Audience strategy by Shannon Busta. The director of New York Times Opinion Shows is Annie-Rose Strasser. Transcript editing by Kate Wilkinson and Marlaine Glicksman.
The Times is committed to publishing a diversity of letters to the editor. We’d like to hear what you think about this or any of our articles. Here are some tips. And here’s our email: [email protected].
Follow the New York Times Opinion section on Facebook, Instagram, TikTok, Bluesky, WhatsApp and Threads.
The post Is There an A.I. Bubble? A Debate. appeared first on New York Times.
