AI’s Giant Money Loop Is Feeling Very 2008

0:02 Bloomberg Audio Studios. Podcasts. Radio.

0:06 News. This is kind of the thing that keeps you up at night when you start thinking about this stuff is there are all these different ways that if this goes wrong or if it doesn't live up to its promise, it will ripple out through the global economy. If it delivers, great. This is going to be good. There's going to be other dislocations and so on. But there's a broader promise there.

0:28 The bad scenario is is even worse than just the kind of stock market crash when you really dig into it.

0:40 I'm David Gurra, and this is The Big Take from Bloomberg News. Today on the show, Bloomberg's Sean Donnan and Bloomberg Economics' Tom Orlick on why the circular funding powering the global AI infrastructure boom is bringing back memories of the 2008 financial crisis. I think regular listeners are familiar with circular financing, where just as an example, you've got a chip maker, say, that invests in an AI lab.

1:09 And what that AI lab ends up doing is buying chips from the same chip maker that invested in it. And we go round and around and around. But if you pull back, there is also this circular flow of capital at a more macro level. And this is really what your story is about, Sean. Explain what we have going on here.

1:23 Yeah, look, we have this huge demand here in the United States for AI chips in particular. Most of those AI chips are coming from places like South Korea and Taiwan. The fact that these AI companies are buying those chips is sending a lot of dollars to those economies, which are building up these huge pools of savings. And those pools of savings are then flowing back in to U.S.

1:54 markets and U.S. equities markets and U.S. bond markets. in particular, and to many of the very same AI companies that are buying these chips in the first place. So macroeconomists talk about global imbalances, right? The imbalance between the savings and the spenders or the investors in the world. And we have global imbalances that are back up near their highest levels in 150 years. They're back up near where they were in the early 2000s before the 2008 crisis.

2:28 And people are worried about that because historically these global imbalances often lead to crises. And this time around, what's very different is who's sitting on the savings as well. And this kind of relationship between this boom that we have now that is feeding imbalances and also being fueled by imbalances.

2:53 Tom Orlick, how worried are you as you hear Sean talk about those global imbalances to turn to you as the macroeconomist here? You know, anytime anyone invokes kind of the run-up to the financial crisis or what happened in the run-up to the dot-com boom, you kind of perk up, take notice. Are these grounds for someone to be worried about what might be happening?

3:10 So certainly I would echo the points that Sean made. And I think I'd add a couple of things. So firstly, there's a positive here as well, right? There's a huge amount of excitement about AI. There's a huge amount of excitement about AI as a productivity booster. And this circular flow of capital, it's accelerating the AI revolution, right? These funds are accelerating the build-out of data centers, accelerating the build-out of models. And so if the tech optimists are correct and AI is going to make everybody better off, well, that's positive, right? We should keep that in mind. At the same time, there are some significant risks here. And I would highlight one difference between what's happening now and

3:55 what was going on in the early 2000s, right? So in the early 2000s, we had China, other Asian countries, the petro-states building up huge trade surpluses and then taking the money from those trade surpluses, giving them to their central banks and having their central banks invest in US treasuries, right?

4:16 So on one side of the transaction, you've got central banks who are kind of very cautious, very slow-moving, very risk-averse. And on the other side of the transaction, you've got the world's safest asset, US treasuries, right? This time around, the imbalances are back, but the actors have changed. On the surplus side, it's not very cautious, very risk-averse central banks moving the money. It's private investors who are much more nimble, much more focused on returns. And on the companies which are taking the investment, well, it's not the safe, secure US treasury market anymore. It's the AI hyperscalers. And if they deliver, well, everyone's going to be better off and the Asian investors are going to get paid. But

4:57 if there's a moment where the market says, you know what? We thought the profits from AI were going to be here and they're not. We're shifting from bulls to bears. Well, it's going to be a challenging moment.

5:08 Sean, you make a point in the piece here that when you look at the profits that these big shipmakers have made, be that Taiwan's TSMC or South Korea's Samsung Electronics. Those profits, those cash piles are rivaling foreign exchange reserves at central banks.

5:22 Yeah, which is astonishing, right? The big savers here aren't, as Tom says, the country's central banks. They're these chip makers and associated companies and so on. And look, we talked to a number of eminent economists for this, and one of the people who we spoke with was the former IMF chief economist, Pierre-Olivier Gorinchas, who has been studying this stuff for decades. And he says, look, you know, in one way, this is the global economy working as it should. You know, there is a potential boost to productivity that is coming, as Tom says, and it is not being funded domestically. It's drawing in money from people who have it. And that makes a lot of sense. But, you know, he and others also say, Even wonderful, sensible

6:12 episodes do not mean or do not have to be or often aren't a blank check, right? And there's this kind of duration problem, right, as well here. And you hear this when people talk about the circular financing more closely when it comes to, you know, the data centers and the chips in them and the relationship between the companies involved. You know, how long... Is that going to take to deliver the profits that you need? Well, it's the same thing at the macro level, where how long is it going to take for this productivity payoff to pay off? And if it doesn't pay off quickly enough, those chip makers are going to be hit both in terms of potentially declining demand for their product, but also potentially on the investment side as well, where they're investing

7:03 back in.

7:04 There.

7:04 So there is a potential to amplify a potential crisis. Tom, what does this mean for U.S. treasuries?

7:11 So you have this kind of voracious appetite to get into the AI boom. At the same time, you've got globally a lot of concern about the state of the U.S. fiscal house, $ 40 trillion debt and the like. To borrow a phrase here shamelessly, how does the U.S. make treasuries great again in light of all of this?

7:30 I think the world has changed, David. If we go back 20 years, 25 years, it was a world of ample savings. China saving a huge amount from its current account surplus. The petro states saving a huge amount from their oil sales, channeling those savings into the US treasury market. And it was also a world where other investment opportunities, well, they were there, but there was certainly nothing as exciting as the AI revolution. And a world where US treasury borrowing was under control, right? We were coming out of the Clinton era where reducing deficits, reducing the debt had been a priority. And now, here we are in 2026, China's not putting its saving into the US treasury market anymore. Petro-states,

8:20 are they still going to be buying treasuries? Don't know, but it seems like a question mark. And on the other side of the ledger, well, because of the global financial crisis, because of COVID, Because of years of wide fiscal deficits, the US debt is blowing up and AI has arrived as this huge sucking sound, right? Which is just absorbing capital from around the world. So how to make treasuries great again? To come back to your question, it's a big challenge. US 10-year treasury rate right now is above 5%. That means the US interest costs each year, more than a trillion dollars That's money which could be used for education, for defense, for infrastructure. With treasury rates where they are, that's not where it's going.

9:10 And there is this thing that here is very, very different from the early 2000s, right? In the early 2000s, those global savings that were pouring into US treasuries were having the effect of lowering interest rates, right? Lowering the yields because there was higher demand. Now we have this kind of competition from AI assets, the bonds, the equities.

9:31 That means that that is taking away demand from the U.S. Treasury's market and having the effect of actually leading to higher yields and higher interest rates here in the United States. So that is also fundamentally different.

9:51 More of my conversation with Bloomberg's Sean Donnan and Bloomberg Economics' Tom Orlick after the break. I want to get into the backdrop to all of this, of course, which is what happens if AI doesn't deliver on the promises it's making to all of us right now. So, Tom, can you just play out at a macro level sort of What would happen if all of this goes bust? If the AI trade were to unwind, who would get hurt?

10:22 So huge amount of optimism about AI, huge amount of potential there, right? So there is a path here where AI delivers on its promise, but we don't have to look too far back in history to find moments where there was an extremely promising technology and investors piled in and there was what Alan Greenspan would refer to as irrational exuberance, a moment of realization, the profits aren't coming as quickly as we expected, and a market crash. The dot-com bubble, clear example of that, but also many other examples in history. The railway boom of the 19th century, for example. So if that happens, who gets hit? Well, of course, all the AI champions, they get hit, their stock value goes down, Of course, all of the people

11:18 who've invested in the AI promise, right? People here in the United States and investors around the world. Who gets hit twice? Well, it's those companies in Asia and those economies in Asia, Korea, Taiwan, China, Japan, that have in one way or another been benefiting from exporting into the AI boom. They've been selling the chips. They've been selling other stuff which goes into data centers, and they've been using the profits from that to invest in the AI boom.

11:53 So if all of this turns south, they face a double blow.

11:57 Sean, pick up on that.

11:57 I mean, you're referring the piece to these places as techno states, singling out South Korea, Taiwan. How precarious a position are they in right now with the investments that they've made?

12:07 Yeah, look. I think one of the, you know, most interesting conversations I had in reporting out this piece was with Beatrice Vetter de Morrow, who is the head of the Center for Economic Policy Research, which is based in Paris, and was actually the person who was called up by the French presidency ahead of the G7 earlier this year and asked to come up with a report on global imbalances.

12:30 And she makes the point in the piece that, you know, we are not in a safe space for because she's worried about these imbalances and so on. And she had originally kind of when the call came, she's like, oh, global imbalances, it's kind of boring subject. But over time, as she looked at this, came to see it as this is just a perfect way to look at some of the risks that are out there in the global economy right now. But she also at one point said, you know, these techno states, they look a lot like resource dependent economies. And That means that they could suffer, potentially, from Dutch disease, right? And Dutch disease is this idea that we often associate with countries that depend on oil or mining, kind of extractive industries, and the economy becomes

13:22 dominated so that when there's a downturn there, those economies suffer much more than they would otherwise if they had more diversified economies. So the real concern in places like South Korea and Taiwan is that they now have economies that are increasingly concentrated in this one area that has all these questions about it. And so there's a potential effect from that. And I think that is an additional risk to what Tom has talked about in terms of the markets and the risk to investors and so on.

13:51 The other big thing people talk about is, look, a significant amount of U.S. growth right now is coming from this AI investment boom. U.S. GDP is growing at 2% to 3%, and that growth You know, there are different estimates out there, but as much as a full percentage point of that growth, some say, could be coming from the AI investment boom. And if that goes away, that means you have a significant slowdown in the U.S. economy.

14:20 And that would take that demand out of the global economy, and that would resonate through the global economy in different ways.

14:27 Sorry, just a point of clarification. When Sean refers to techno-states, he's talking about countries uh economies like korea and taiwan where a significant share of the economy is based on technology um not to a frame of mind that follows listening to 90s british dance act the prodigy.

14:49 You couldn't resist and i i could tell you were sitting on that for the talk of trying to wait so long to say that um Tom, how close are we to get into a verdict on whether or not this is a bubble or not? And I sat down with Howard Marks, big investor from Oak Tree a few weeks back, and he reminded me, of course, you can't really call a bubble until it's popped. You know, you look at the warning signs that we've talked about. You look at the spike that we've seen in yields. Do you feel like we're reaching a precarious point here?

15:19 So it's notoriously hard to call bubbles, right? And clearly there is real potential. with this technology and if you look at the revenue run rate for anthropic for example well it's already a technology which is delivering on that promise right that said valuations do look very very elevated a couple of things which caught my eye recently first planned ipo of an australian data center company was pulled in the last few days um apparently because investors thought the valuation was too rich. OpenAI, also generating quite a lot of revenue, but the latest news suggesting the run rate for 2026 isn't quite as high as investors thought it would be. So is it a bubble? Is the bubble about to burst? Don't know the answer, but the

16:20 more there is that trickle of bad news, the more chance there is that a moment of reckoning arrives.

16:30 We should also be clear that, you know, just as we have seen big capital investment episodes in US history that have turned out to be incredibly good for the economy in the longer term, they often have been associated with a downturn. So the 1890s, the railway investment boom, is, goes gangbusters, and then it turns and the economy goes into a recession afterwards. But, you know, we would all argue that railways were good for the U.S. economy in the longer run. The dot-com boom in the 2000s, you know, it came after a huge build-out in terms of fiber optic networks and so on that, again, have been good for the long run for the U.S. economy, but some of the valuations got ahead of things. You had a

17:15 recession that came along there. There was an amazing paper that came out of Brookings and the National Bureau of Economic Research just in the last few weeks out of a Columbia Business School professor who made the point that between 2025 and 2032, the total amount of capital that will go into the AI build-out will be around $ 9 trillion, or about 3.2% of GDP a year. And that is bigger than the railway investment boom of the 1890s. That makes it the biggest capital investment boom that we've seen in US history. So just putting that in context, history also tells us that these things may be good in the long run, but there's bumps along the way.

18:10 This is The Big Take from Bloomberg News. I'm David Gurra. To get more from The Big Take and unlimited access to all of Bloomberg.com, subscribe today at Bloomberg.com slash podcast offer.

18:22 Thanks for listening. We'll be back tomorrow.