Ray Dalio, the founder of Bridgewater Associates, told The Atlantic's chief executive Nicholas Thompson in a video interview published on October 14, 2026 that a bubble indicator he built places the current AI boom in the same order of magnitude as the 1929 and 2000 bubbles - and that a productivity miracle would not by itself prevent a crash, because the US government spends about $7 trillion a year while taking in about $5 trillion.

In Short

A famous investor named Ray Dalio said that the money pouring into AI looks like the giant stock market bubbles of 1929 and 2000. He thinks the danger is less about whether AI works and more about people needing real cash when the paper wealth from AI companies has to be sold, while the US government is also borrowing heavily. He said people should not put everything they own on a single bet like AI, in either direction.

The interview and its date

The conversation appeared on the YouTube channel of Nicholas Thompson as an episode of The Most Interesting Thing in AI, a podcast Thompson hosts. The video page captured on October 17, 2026 shows a publication date of October 14, 2026, a running time of 49 minutes and 56 seconds, and 49,822 views at the time of capture. The transcript itself carries no date, so the October 14 date comes from the video page rather than from anything said in the conversation. When the recording took place is not stated.

Thompson introduced the program with a disclosure that matters for how the episode should be read. "This podcast is from my team on the business side of The Atlantic and independent of The Atlantic's editorial staff," he said. The video description repeats the point in shorter form: "Produced independently of The Atlantic's editorial staff." The interview is therefore a business-side production of The Atlantic, not a piece of the magazine's journalism.

The episode was framed as a counterweight. A few weeks earlier, Thompson said, Azeem Azhar had argued on the same show that AI is not a bubble. "One notable skeptic is Ray Dalio," Thompson said. He described Dalio as the founder of "the world's largest hedge fund," credited him with predicting the 2008 financial crisis, and noted that Dalio left the firm "last year" and now publishes through a newsletter, Principled Perspectives, and books including How Countries Go Broke. The video description lists chapters running from "The forces that form bubbles" at 2:20 to "What Ray would do if he had an unlimited budget to invest in AI" at 48:27.

Five forces, one of them technology

Thompson opened with Dalio's framework of five forces that drive history: debt, internal politics, the world order, acts of nature and technology. His question was whether technology now outweighs the other four. Dalio declined the premise. "Any way you cut it, they're intertwined," he said.

The first force, debt and money, does not change because of AI, in Dalio's account. "If you're holding a bond, you expect to get paid back. And if you're owing, you have to pay. So that doesn't change with AI," he said. "My main point is that everything's mechanical."

The second force is the internal political and social order, which Dalio tied to wealth gaps and values gaps that produce "greater polarity, populism and so on." The third is the international order. Dalio said the multilateral system the United States built after 1945 - the United Nations, the World Bank, the World Health Organization - is ending. "That's over," he said. "We're now coming into a more the way the world used to be, which is power matters."

Acts of nature form the fourth force. Dalio said climate is an issue and predicted that a "super El Niño in the year ahead or two" would show it. Only then did he reach technology, which he described as "man's inventiveness of particularly new technologies." His conclusion was that AI "matters, but it's very much tied to these other forces. So we can't look at it separately."

How Bridgewater systematized decisions

Thompson asked Dalio to describe how Bridgewater used rule-based systems long before large language models. Dalio said he "controlled it for 50 years" and stepped out after 48. His method, he said, was to write down his criteria for decisions, back test them and then computerize them.

He gave one example. "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," he said, the question became whether it was a good decision to be short bonds rather than long bonds when growth was high and inflation was rising. The number of such rules ran to "hundreds," according to Dalio, built as a certain number of rules per market.

Dalio compared the process to a computer chess game. When the system's move differed from his own, he asked why. "I would say probably 65, 70% of the time it's because I forgot or couldn't program that complexity," he said, with the remainder leading him to systematize something new. Over time, he said, the two "become one and the same." Thompson cited a 9% return during the 2008 crash; Dalio did not dispute the figure.

The relevance to AI, in Dalio's telling, is that cause and effect still apply. "The causes happen before the effects," he said. "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."

Why Dalio separates wealth from money

The core of Dalio's argument about a bubble rests on a distinction he repeated several times: wealth is not money. Transformative inventions draw investment because people can see them working, he said, citing the 1920s, when households got electricity, refrigerators, radio, cars and films for the first time. "Everybody looks around and they say that is obviously a thing to bet on," he said. "But everything has a price and is that price too high?"

Companies, according to Dalio, face a choice under deep uncertainty. "I need to compete for capital. I don't know whether I'm over investing or underinvesting," he said, paraphrasing the executive's position. The safer course, from inside a company, is to spend as much as possible, because underinvestment guarantees losing. Thompson added that a company could end a year with an annual run rate a thousand times larger than at the start. "That's right. It's hard," Dalio replied. "And therefore do distinguish a good technology from a good investment and people classically fail to do that."

He illustrated the wealth problem with a funding round. "You can make an offering, I'll raise $50 million at a billion dollar valuation," he said. "Only put up $50 million, but all of a sudden you're worth a billion dollars." That valuation is wealth on paper. "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."

The pop comes when holders must convert. Dalio pointed to wealth taxes as one trigger and named California's proposed billionaire tax. "Everybody has to pay it. They don't have the cash, they got to pay it," he said. He also asked whether assets sit in "strong hands or weak hands" - whether owners can absorb losses or will sell when they find themselves "half as wealthy."

Dalio rejected a common definition of a bubble. "A lot of people think, well, will it pay off in the end? It's not that. The end never happens," he said. "The question is, do you have a crash or does something go down in the intervening period?"

Triggers on the table

Thompson listed events that could force wealth into cash. He said Anthropic was about to go public, which would eventually lead employees to sell shares "so they can buy the house or the yacht or give it to the cause." He cited the Situational Awareness fund in July, when, he said, Leopold Aschenbrenner had to raise a large amount of money to cover leveraged bets. He also mentioned the SpaceX IPO and an expected OpenAI IPO. These are Thompson's characterizations in the interview; the transcript does not include financial details for any of them.

Dalio agreed with the direction. "So we have a bunch of dynamics that look to me like some big pins going towards a bubble," Thompson said. "It looks to me that way too," Dalio replied. He then described the concentration that worries him: "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."

The bubble indicator

Dalio said he built a bubble indicator that combines these factors. "This is a bubble measured analogous to the 2000 and the '29 bubble in that order of magnitude," he said. Asked to compare it with 2008, he said 2008 was smaller as a bubble but carried heavy debt, and was concentrated in housing. "2000 was a bigger bubble than 2008, which was a big bubble because of the debt," he said. "The subject in 2008 was it was houses."

The indicator's components and weights were not disclosed in the interview, and Dalio himself allowed for error. "But let's say I'm wrong," he said before continuing. The comparison is his own measure and not an independent calculation.

The deficit as the "plaque"

Thompson pushed back. Tariffs, the Iran war and the Situational Awareness episode had not popped the bubble, he said, so why could the same dynamics not sustain it for three more years? Dalio answered by turning to the bond market and to US government finances, using the analogy of a cardiologist reading an x-ray.

Credit, in his description, works like a circulatory system. When borrowed money produces enough income to service the debt, the system is healthy. When debt service rises relative to spending, "it's like a plaque in the circulatory system building up. And then that's what causes an economic heart attack."

He then gave figures, all his own:

  • The US government spends about $7 trillion a year and takes in about $5 trillion.
  • The deficit is about $2 trillion, which he described as overspending "by about 40%."
  • Total debt is "six times the amount of money that's coming in."
  • Interest is about $1 trillion a year, or roughly half the deficit.
  • About $10 trillion of debt matures each year and has to be resold.

The second problem, according to Dalio, is demand for that debt. "One third of our buying of the debt is foreigners," he said, adding that the largest foreign buyers were China and Japan. Foreign buyers are concerned about the volume of debt and about the possibility of sanctions "the way Russia was sanctioned," he said. "So we have a supply that's going like this and we have a weakening demand."

He presented it as an argument to check rather than accept. "Don't believe me, you don't have to believe me," he said. "You could look at the x-ray, you could decide what for yourself."

Why a productivity boom may not be enough

Thompson made the optimists' case. AI, he said, could raise productivity tenfold, with "geniuses in a data center," and the country could grow its way out of the debt. "I've got plaque building up," he said, extending the metaphor. "But there's a new treatment. It's in the works. It's about to get approved by the FDA."

Dalio accepted the mechanics but questioned the timing and distribution. If incomes rise enough, he said, debts become easy to pay. The problems are whether productivity becomes income, whether the government captures enough of it through taxes, and whether that happens quickly enough. He put the government debt at about $280,000 per household.

He also said competition tends to pass productivity gains to buyers. "A lot of productivity miracle is not transferred to income because there's always the competition and there's always the means that it becomes cheaper, there's a price competition and so on," he said. His own reading of the year-by-year projections: "wow, that is a scary picture."

Dalio also connected the wealth gap to politics. AI, he said, "creates a very large differences in wealth," which ties back to his second force.

Government stakes and the "3% solution"

Thompson raised the idea, discussed "maybe this past spring" according to him, that the US government could take a stake in OpenAI similar to its 10% stake in Intel, or create a sovereign wealth fund holding AI companies. Dalio said the new world order changes how he views such interventions. "To believe in entirely a free market system that directs everything has always never been the case and not necessarily is the best way," he said, while describing himself as "very much inclined to be free market."

His reasoning was competitive. Consumers left free might "buy $1,200 handbags," he said, and those resources may not help in a contest where "if you win the AI battle, you're going to win not only the economic, but you're going to be the battle for military or the world dominance type of battle." He said he supports the "Trump accounts" in Connecticut.

Thompson suggested that a government stake taken before an IPO would mean less private wealth needs to become money. Dalio returned to the cash question each time. If the government pays for a stake, it borrows the money; if it takes the stake without payment, someone is taking something from someone. "I'm just a mechanic," he said. When Thompson suggested AI companies could pay taxes in equity, Dalio said that "is still not money," and Thompson conceded the point with an image of trying to buy a banana with a fractional share in Anthropic.

Asked what he would do if given full control of US economic policy, Dalio described what he calls the "3% solution": bringing the deficit below 3% of GDP. He said the target is widely shared, naming Treasury Secretary Scott Bessent (rendered "Scott Bessen" in the automatic transcript). His suggested mix was proportionate. "Raise the taxes by four or 5%, whatever, and cut your spending by four or 5%," he said. Politicians on both sides agree with the mechanics, according to Dalio, but tell him they cannot win office without promising "I will not raise your taxes and I will not cut your benefits."

When Thompson asked him to design policy for America's position in AI, Dalio said taxes have to work so that "the majority of the people are going to benefit," and that there would probably be taxes associated with AI. He did not specify them. Thompson said he had spent a lot of time with Dalio and had no idea how he votes. "Just to let you know, I'm apolitical," Dalio said.

How China views AI, according to Dalio

Dalio, who said he spends considerable time in China, described three ways Chinese leaders think about AI. First, they treat it "like electricity or running water or infrastructure," where "profit is not the main objective" and efficiency is. Second, they focus on placing "simpler, cheaper versions of AI in manufactured products" rather than on the frontier model race. "That frontier model and going to the highest levels of capabilities is not where most of the benefit is," he said.

He backed this with a figure. "The Chinese right now own 32% of world manufacturing... manufactured goods, more than the United States, Germany and Japan combined," he said. The interview does not give a source for the number. Third, he said, China sees AI as essential for both "economic warfare or military warfare."

Personal advice and Digital Ray

Thompson said he had used Digital Ray, an AI avatar of Dalio, to prepare questions. Dalio said his priority now is to "pass along whatever I have that might be of value," including an AI platform for investing "that anybody can use." He did not give a launch date, a name for the platform beyond his description, or pricing. "That's what my hope is," he said.

His advice to young people was that little has changed. "You have to be on the cutting edge of the AI wave. You have to be on that crest of the wave or you're going to be drowning in the wave," he said. He restated a principle from his book Principles: "pain plus reflection equals progress." For investors generally, his conclusion from the bubble discussion was about concentration. "Don't make one thing that is highly volatile and is leveraged and so on your one thing bet," he said, adding that this applies "whether you want to bet for it or whether you want to bet against it."

Why this matters for marketing

Dalio's warning about "a lot on that one thing" has a direct line to advertising, because advertising pays for much of the AI buildout. Meta's infrastructure chief Santosh Janardhan put Meta's 2026 infrastructure spending at "well over $100 billion" in an interview published on October 8, against capex guidance of $130 billion to $145 billion. PPC Land's analysis of that interview noted that advertising made up about 97.6% of Meta's second-quarter 2026 revenue and that free cash flow for the quarter fell to $784 million.

Concentration on the customer side is also visible. In an August podcast covered by PPC Land, Ed Zitron cited UBS estimates that OpenAI and Anthropic would account for 27% of Google Cloud revenue in 2026 and over 48% in 2027, while Alphabet raised its 2026 capex guidance to $195 billion to $205 billion. Those are the same two companies Thompson named as IPO candidates in the Dalio interview.

Advertising is also one of the routes AI companies have chosen to close the gap between wealth and money that Dalio described. ChatGPT ads launched in February 2026, and CPMs fell from $60 at launch to as low as $25 within weeks, according to a Digiday report cited by PPC Land. OpenAI told investors to expect $2.5 billion in ad revenue in 2026 and $100 billion by 2030, while Reuters reported a projected loss of about $14 billion for the company in 2026. Whether that ad revenue arrives in time is a version of the question Dalio asked about the federal budget: not whether income grows, but whether it grows fast enough.

Holding companies are feeling the cost side already. PPC Land reported in August that agencies are bundling AI tooling and tokens into principal media deals, with Omnicom's third-party service costs reaching almost $2.9 billion in the first half of 2026. On employment, which Dalio listed among the exposures, Stanford payroll data showed employment for workers aged 22 to 25 in the most AI-exposed occupations fell 16% between October 2022 and September 2025, and agency owners said in September they had stopped hiring graduates as AI absorbs junior work.

The bubble debate itself is not new to PPC Land. In October 2024, a Yale Digital Ethics Center study by Luciano Floridi argued that the AI market shows classic bubble signs similar to the dot-com crash, comparing it with five earlier episodes. Dalio's contribution differs in emphasis. He does not argue that the technology will fail; he argues that the payments, the debt and the timing can break first. For an industry whose largest platforms are spending ad revenue on data centers at record rates, that distinction is the one worth tracking.

Timeline

Summary

Who: Ray Dalio, founder of Bridgewater Associates, interviewed by Nicholas Thompson, chief executive of The Atlantic, on The Most Interesting Thing in AI, a podcast produced by The Atlantic's business side and independent of its editorial staff.

What: Dalio said a bubble indicator he built places the AI boom in the same order of magnitude as the 1929 and 2000 bubbles, argued that bubbles pop when holders must turn paper wealth into cash, and said US government finances - about $7 trillion of spending against $5 trillion of revenue, $1 trillion of annual interest and about $10 trillion of yearly maturities, by his figures - leave little margin even if AI lifts productivity.

When: The video was published on YouTube on October 14, 2026, according to the video page; the recording date is not given.

Where: On Nicholas Thompson's YouTube channel, with the discussion covering the United States, China and global bond markets.

Why: Dalio's case is that the risk lies in debt, cash needs and timing rather than in the technology itself, and that the concentration of wealth, stock market value and employment in AI raises the stakes. For marketers, the same concentration shows up in ad-funded AI capex at Meta and Google and in AI companies' reliance on advertising revenue to cover losses.