In June, Ray Dalio described the artificial intelligence (AI) rush as a bubble still in its early stages, with no triggers in sight. Four months later, the founder of Bridgewater Associates has changed his tone. Speaking today at the Forbes Global CEO Conference in Singapore, he called the AI boom a «classic bubble» and moved the clock forward: «We’re not there yet, but we’re getting closer.»

What has changed in the meantime is mostly the backdrop. The factor Dalio points to as a possible trigger for the burst, rising bond yields, has materialized over the past few weeks.

Why interest rates are the fuse for the AI bubble

Dalio’s reasoning starts from how the boom is being financed. According to South Korean broadcaster SBS, the financier observed that «an enormous amount of debt is being raised to finance AI» and that, if rates keep rising, the bubble will reach the point where it bursts. Data centers, chips, power grids: a growing share of the investment needed for AI infrastructure is funded with debt, including bond issuance. Every extra point of yield in the government bond market therefore feeds straight into the cost of that debt.

Morgan Stanley estimates that AI-related bond issuance will reach nearly $570 billion in 2026, with $236 billion already placed as of May 31, a pace four times faster than in 2025. According to the investment bank, capital expenditure at the big hyperscalers (Amazon, Alphabet, Meta, Microsoft and Oracle) now absorbs about 100% of their operating cash flow, against a ten-year average of around 40%.

The bond market has moved sharply on this front. Today, October 7, the 10-year Treasury yield stands at 5.31% and the 30-year at 5.66%. Over the past month the 10-year has gained more than half a percentage point and touched 5.33%, the highest level since 2002. Among the pressure factors, traders cite persistent inflation, concerns over U.S. public finances and, not by chance, the growing supply of AI-related debt.

The result is a potentially vicious circle: companies borrow to build AI infrastructure, the abundance of issuance helps push yields higher, and higher yields make the debt needed to keep investing more expensive. On top of this comes the effect on valuations. When risk-free rates are above 5%, the cash flows expected many years from now, which underpin the multiples of AI-linked stocks, are worth less today.

The challenge of turning wealth into cash

The second trigger Dalio identifies is a theme the investor has repeated for months: the difference between paper wealth and spendable money. «Everyone says “I have a billion dollars of net worth,” but try spending it,» he said. «To spend it you have to sell assets and turn them into cash, and that’s usually where the bubble starts to burst.»

According to Dalio, the occasions when this conversion becomes unavoidable could come from wealth taxes and, more broadly, from the need to monetize unrealized capital gains. The mechanism is the same as in every bubble: as long as nobody needs to sell, prices hold, but when many holders have to raise cash at the same time, demand is no longer enough to absorb the supply.

Editor’s note

This article was originally published in Italian on money.it by Flavia Provenzani on October 07, 2026 as «Ray Dalio cambia diagnosi sulla bolla dell’AI. Il punto di rottura si avvicina e la miccia sono i rendimenti dei Treasury». It has been translated and adapted for an international audience by the Money.it International desk.