We have a bunch of dynamics that look to me like some big pins going towards a bubble.
It looks to me that way too. Okay. So, and you have a very great concentration of everything being bet on it.
Mhm. In other words, its share of GDP, its share of wealth, its share of the stock market, what it means for the bond market, what it means for trade, what it means geopolitically for war, what it means for employment. There's a lot on that one thing.
Hi, I'm Nicholas Thompson, CEO of The Atlantic, and this is the most interesting thing in AI. A few weeks ago, I had Azim Azar on the show and we talked about one of the biggest questions in AI: Are we in a bubble? The technology is amazing, but that does not mean the economics work. And if you've listened to that episode, you know his answer was no. No, it's not a bubble. We worked through a whole series of arguments. I kind of made the case that it is. We batted each of them back. But of course, not everybody shares Azim's views. One notable skeptic is Ray Dalio. He is the founder of Bridgewater, the world's largest hedge fund. Ray famously predicted the 2008 financial crisis, which meant that as the housing bubble burst and unleashed economic chaos, Bridgewater's investments actually went up in value, as they did basically every year since he started the fund. He left the firm last year. Now, he shares his insights through his newsletter, Principled Perspectives, as well as through books like How Countries Go Broke. If you want smart insights on markets and investing and what happens when an asset gets overvalued, Ray's a good guy to know. For the past year, he's been sounding the alarm on AI. He said it's a bubble, but not for the reason you'd expect. So, I sat down with Ray to hear why and how we can avoid another cycle like 2008. And this podcast is from my team on the business side of The Atlantic, independent of The Atlantic's editorial staff. Here's my conversation with Ray Dalio. Ray Dalio, welcome to the most interesting thing in AI.
I'm excited. I'm delighted to have you. Most interesting thing in AI.
That's what we're going to try to get to. All right, let's just start your big theory. Five forces that drive history: debt, internal politics, the world order, acts of nature, technology. You describe world history with them ebbing and flowing. My first question: are we at a stage where technology, the fifth force, is becoming more powerful than the other four, or is it in the same balance it's been in historically?
Any way you cut it, they're intertwined.
And I just want to touch on those.
The first force is the money, debt, government spending, capital, very important in AI. Is this a bubble? Is this going to work or whatever? So that first force, where are we? Do we have a debt bubble? What's going to happen?
Okay, that's a big force.
By the end of this conversation, every listener is going to know the answer to that.
That's why I wrote the book, okay? Because it's mechanical. It still is mechanical. It doesn't change. When you owe, if you're holding a bond, you expect to get paid back. And if you're owing, you have to pay, right? So, that doesn't change with AI.
I thought you were reaching for my questions, Ray.
Wouldn't think of it. Okay. Number two, that force mechanical.
Okay. My main point is that everything's mechanical. Everything that happens has causes that make it happen. And if you understand the mechanics like that, then you understand how it works and that's important. Okay? AI doesn't change that. Okay? It'll influence it. But the second force is politics. Okay? Internal order. There's a political order, there's a social order. The more you get to left and right and you have a conflict, okay, there's a problem. So the first, if you have a debt problem, you have a problem. Okay, these end up being in big cycles.
Because there's a debt cycle, there's a political cycle. And that political cycle, as you come out of the conflict and you wipe out the debt and you have a war and you now start with a fresh slate like we did in around 1950 or after 1945, you have a fresh slate. You begin this cycle and then the debts rise relative to incomes and so on. And then the politics and it's connected to wealth gaps. So that second and fourth, wealth gaps and values gaps, causes as that cycle happens it causes more disorder because there's greater polarity, populism and so on. So that's second force very much related to the first force which is the debt money. And we're going to get to the other, the third force is geopolitical. In other words, the second is how we deal with each other within the country. The third force is how countries deal with each other. There's a world order and that has a big effect, right? And so we're going from what used to be a multilateral world order that the United States created after the 1945 war. It was a replica almost of democracy. But there's the United Nations and these multilateral institutions. The World Court, can you imagine that? The World Bank, the World Health Organization. That's over, right? And so we're now coming into more the way the world used to be, which is power matters. And so that produces greater conflict. So that's the third. The fourth, no particular order, the fourth is acts of nature. We have to be aware that they've killed more people in history. Nature matters and so acts of nature and certainly climate is an issue and probably as we come into a super El Nino in the year ahead or two we will see that matters. And then the fifth force is man's inventiveness of particularly new technologies, right? And so AI does, AI matters, but it's very much tied to these other forces so we can't look at it separately, right? How much money does it get? Money? Did they make money? What are the political and what are the geopolitical? I'm sure when we're talking about AI, we're going to talk about China. We're going to talk about what it means in terms of economics and war. Okay. So those five forces and they tend to operate in a cycle because the first three are very cyclical.
Yep. All right. So it was a good answer, Ray, because every single sentence I wanted to respond to with another question, but let's do this. I want you to go back in time a little bit and then we'll get to where we are in the debt cycle, where we're in the political cycle, how AI is going to affect it. So, you ran Bridgewater, most successful hedge fund in America, and you ran it for 40 years.
Um, about years, I controlled it for 40, for 50 years, and I probably ran it, I stepped out 48 years.
But what's interesting and relevant to this conversation, you were wildly successful, you were using a form of AI and obviously forms of high technology but very much a rule-based system as I understand it from going back and reading how you set it up, saying these are the principles, these are what I believe, these are the rules. Explain how Bridgewater used AI and let's talk about some of the differences between that and the way AI works now.
The term AI began and so the concept in 1956, of course. And it's had different forms along the lines. So what I did in building Bridgewater was I would think, one of my criteria for making decisions and I would specify those. I would go back and back test those.
And then I would computerize those.
So give me an example of a rule.
If growth, measured by the initial claims for unemployment, was rising and then I combine it with commodity prices as a daily measure of inflation. If I put those two things together, when growth is high and inflation is rising, if I want to then be short bonds rather than long bonds, is that a good decision?
Right. Okay. So, you figure out these rules, you figure out these systems, you build it, keep going.
Okay. So I learned then about the powers of being able to do it that way. What the main powers are, first of all, if you have those rules you can apply them everywhere. Like if I was thinking what am I going to do to what country do I want to invest in, what do I want to do, when I have those criteria I apply it everywhere. And then also complexity can be handled in a much better way, speed is in a much better way, less emotions entering into the decisions so I can make much better decisions. And so it's more like doing the research on the criteria, programming them up, and it's like building a factory that doesn't have workers. In other words, the old way, it seems to me the mistaken way of everybody was to weigh things in their heads where it's the working of weighing those things in your heads, but then also then thinking what are they, program them, check it. So it was like building a computer chess game. In other words, it's my criteria. How would I play chess? And the system would be dealing with the world. And then I'm dealing with the world. And then I reconcile those two things. What move would it make? What move would I make? And if they were different, I would say, why would that move be different? And I would say probably 65, 70% of the time it's because I forgot or couldn't program that complexity. And maybe a third of the time it's something else that then I say, 'Oh, I better systemize that.' And over a period of time they become one and the same.
Because you've learned something new about commodity prices or the bond market.
That's right. In other words, the criteria. There's still information. There's still criteria. There's still cause-effect relationships. So, and the causes happen before the effects. So if you know those cause-effect relationships, whether it's the most advanced AI or whatever thinking you have, to understand those cause-effect relationships.
Right, so you build these rules, you very famously you end up with a 9% return during the stock market crash of 2008, like you understand where things are going in ways that no one else or very few other people do. How many rules do you systematize into the Bridgewater system? Was it a hundred rules? Was it 100,000 rules?
I would say there's a certain number of markets and then there's a certain number of rules per market that adds up to that. So I would say it's hundreds, but I don't want to make it abstract. I want to bring it to, let's say, each person.
If you went around and you simply said I'm going to pause, I've made a decision.
You made a decision to have me on this show or whatever it is.
And you start and you pause and you think, why did I make that decision? What are my criteria? And you just spit them into your iPhone or whatever and you start to think about that, you will acquire those criteria that you're using. And then if you do that, you will be able to systemize your decision-making process, which is certainly where we are now. And if you do a good job and you know how to do it, you would also back test those rules to see how they would have worked in the past to help provide the better principles. I call these principles. What are these principles that if I do this, I'm going to get a better outcome than if I do that. That's where we're going in AI.
So let's go back to the Bridgewater story. So Bridgewater succeeds, you're able to anticipate when bubbles are going to burst, you're able to anticipate what will prick the bubble, right? Knowing there's a bubble is one thing. Knowing what will prick the bubble or anticipating or guessing or estimating, 2026, where are we now in the cycle, particularly when it comes to the question of a tech bubble? And then very relatedly, if we're in one, which I've heard you say, what could prick it?
The way I want to convey it and try to convey it in the book is what are the cause-effect relationships? How does the machine work?
There are basics such as if the government has a lot of debt, how is it going to pay that debt? And then if companies are taking a lot of debt and selling capital, they have to raise capital in order to do these things. There's mechanics. There's a certain supply and a certain demand that enters into that, right? Okay. So now you look back and you would say an answer to your question for me, which is a bubble question, not an AI question because the AI can do it as well, it just has to replicate, is that throughout history, and I think it is true today, that there are miraculous inventions that come along that dramatically change the way the world works. So if you look back at those and so on, there's a mechanics to that which means that people want to invest in those things that they believe, that their experience, like in the 20s, okay, first time there's electricity in houses and you have a refrigerator and light, first time movies, first time radio, cars and airplanes and so on, everybody looks around and they say that is obviously a thing to bet on. And then everyone wants to bet on it and then it becomes a thing. Okay, but everything has a price and is that price too high and then there's a need for cash. So the way they work is people will borrow money and it goes through a cycle. So there's what companies do and there's what people do, investors do. What the companies do is they say I need to compete for capital. I don't know whether I'm overinvesting or underinvesting. There's two choices because it's very unknown. There's two choices. First, I better put as much money as I possibly can because I'll certainly lose if I'm not going pedal to the metal. And then there's those who might overinvest and nobody gets it right.
Nobody. And it's so hard right now where you could be a thousand times larger with your annual run rate at the end of the year. It's at the beginning. It's very hard within that range.
That's right. It's hard. And therefore you have to distinguish a good technology from a good investment.
And people classically fail to do that.
This is good technology. We don't know if it's...
Right, and then they say to themselves, I can borrow money or the companies can borrow money. Okay. There are limits to how much money you can borrow. And what happens is you also have another dynamic that wealth increases relative to money. Okay. So let me put this in an example. You can today or any time you can make an offering. I'll raise $50 million at a billion-dollar valuation. Okay. You only put up $50 million, but all of a sudden you're worth a billion dollars. That company is worth a billion dollars. It may not be worth a billion dollars. But now you have a lot of wealth. Okay? And so there's a lot of wealth around. Wealth is not money. Okay? You can't spend wealth in order to spend. You either have to sell wealth to get money to spend or you have to borrow, which still means you're going to have to come up with cash at some point. Wealth is not money. And so when you're in that situation, you see very classically that that particular wealth grows relative to money. Okay? And then there's a pop of the bubble. The pop of the bubble. So very different from the bubble. The bubble increases. Okay. And the question is even what do we define as a bubble? A lot of people think, well, will it pay off in the end? It's not that. The end never happens. 25 years later is the end. And that's not the verdict. The question is do you have a crash or does something go down in the intervening period? And that happens when they want to sell the wealth. They have to sell the wealth. Let's say for example if you had wealth taxes,
Right? California is proposing a billionaire tax. Everybody has to pay it.
They don't have the cash, they got to pay it,
So that's the mechanics, or you've got to come up with the cash and that is what pops the bubbles. Okay, so that's the dynamic and there are a couple other things. Okay, is it in strong hands or weak hands? In other words, the holder of that, are they able to take losses and they'll go through it or will they be in a position where they get scared, they sell out or they need the cash because all of a sudden they thought that they were wealthy but all of a sudden they're half as wealthy and what do they do? Do they buy more? Do they sell? And there's that dynamic that takes place and that's how bubbles work.
Okay, fantastic. Thank you so much. It seems like we are quite close to a number of moments that could lead to a burst in the bubble. We have potential wealth taxes that are being discussed. Anthropic is about to IPO. All that wealth is going to turn into money. People will have to buy the stock.
Well, no, it won't turn into money from the get-go,
Right? There will be a process.
There will be somebody's going to want to turn that into money because they want to buy that house or that apartment.
That's right. Right. So the Anthropic employee will sell the wealth for the money so they can buy the house or the yacht or give it to the cause.
And someone will have to get the money. In order to get the money to buy the wealth they may have to sell off their other wealth, right?
We don't know what's going to happen. So you have all of those triggers. You do have, we saw with the Situational Awareness Fund this past July, you have very eager investors who are extremely leveraged who at that moment, you know, Leopold Aschenbrenner had to acquire a lot of money to cover a bunch of bets. So, we have a bunch of dynamics that look to me like some big pins going towards a bubble.
It looks to me that way, too. Okay. So and you have a very great concentration of everything being bet on it.
In other words, its share of GDP, its share of wealth, its share of the stock market, what it means for the bond market, what it means for trade, what it means geopolitically for war, what it means for employment. There's a lot on that one thing.
Right. Going back to your forces, just I'm going to... American political system not in the best shape. Internal politics don't seem great. Geopolitics with China not great. Potential shock to the system with a move against Taiwan. Xi's in his 70s, wants to do it in his lifetime. Who knows what's going to happen. Maybe it's better to do it before Trump leaves office. We don't know. Right. Climate, another wild card. You could have disasters. Right. So there are all kinds of forces that make this a perilous moment with bond prices going up right like...
Bond yields going up, all kinds of problems at the moment.
Right. And a basic thing about... well, there are two questions I think. Is this a bubble? And I'm just saying I created a bubble indicator that has all of these things. And this is a bubble measured analogous to the 1929 bubble in that order of magnitude in terms of that mix. But let's say I'm wrong.
But is it less or more of a bubble than 2008?
2008 was less of a bubble.
It had a big debt than 2000 in terms of that double dynamic of a bubble on your bubble.
There was a housing... Okay. There was, 2000 was a bigger bubble than 2008, which was a big bubble because of the debt.
The subject was different. The subject in 2008 was houses.
Okay. It was real estate and it was structured in that way. And so that was a bubble that was going to break. But if you're taking the enthusiasm, the whole stock market thing and everything in that, the bigger bubble was also in 1929. But anyway, the point I'm trying to say is those conditions exist. But I think the question, when we know that and we also know that everything is one thing, everything is AI. Okay. Then what an individual should do is don't make my everything AI, right, whether you want to bet for it or whether you want to bet against it, don't make one thing that is highly volatile and is leveraged and so on your one thing bet.
Let me ask you a couple of sort of questions that push against this hypothesis. So one would be, Ray, we've had this situation for a bubble for a while and we've had a bunch of exogenous shocks, right? We had tariffs. We've had the Iran war. We had the situation with Situational Awareness. None of this has even seemed to have come close to popping the bubble. Like we're just going to get through the Anthropic... We had the SpaceX IPO, right? We had a lot of wealth and money. We're going to get through the Anthropic IPO, the OpenAI IPO. Why can't the dynamics that have sustained the economics for the last year sustain it for the next three?
I would ask each person who asks that question to first of all understand the mechanics and see the charts. So let me give, let's say what I would say is the debt bubble.
Okay, let me start off with the bond market.
Okay, that's the opposite. There's an asset and liability. The asset is the bond. The liability is the debt.
Okay. And so what I would say is, okay, the mechanics I want to tell you, let's say let's start with the US government but we can go beyond that. And what we would say is the way it works is that it works the same for the government as it works for you and I with debt, meaning you take it on, you've got to pay it back in terms, and the only difference is the government can print money. And so what happens is the way these come to an end is very simply two things happen. The credit system is like the circulatory system. And what happens is if money and credit is lent and it earns more than its cost to service that debt then you have a healthy system. So providing debt and leverage is good if it produces the incomes that are necessary to service that and the world goes on to a better place.
And you could see that in the numbers by seeing how much debt service is creeping up on spending. When debt service increases relative to spending, like you or for the government, that means that your ability to spend is less because you have this debt service payment and it's like plaque in the circulatory system building up and then that's what causes an economic heart attack. Okay, that can be watched. I'm going to go through those numbers in a second for you quickly.
Okay, the second thing is there's a supply and a demand. In other words, every year you have to, if you're running a deficit, that means you have to sell your debt. And that means there's a certain quantity of it and there's a certain buyers of that debt. And do they want to buy it? And so either of those two things can cause a problem. And you can monitor that just like your cardiologist could monitor that plaque building up in your... Okay, so here are the numbers. The United States government spends about $7 trillion a year and takes in about $5 trillion a year. So its deficit of $2 trillion, it means it's overspending by about 40%.
And it's done these deficits for a long time. So the amount of debt that it has is six times the amount of money that's coming in,
And not only do you have... So your interest on that debt has now become half of that deficit. It's about a trillion dollars. It's $2 trillion you got to sell. About a trillion of that is to pay the interest rate.
