Howard Marks Sees Mixed Signals as Wall Street Weighs AI Bubble Risk

Oaktree Capital co-founder Howard Marks says Wall Street’s AI bubble signals remain mixed as market declines, above-forecast U.S. inflation and debt concerns at infrastructure firms fuel fresh scrutiny.

The Nasdaq-100 fell 5.5% last week, and the Nasdaq Composite lost 4.5%. U.S. consumer and producer inflation readings both exceeded expectations, while the 10-year Treasury yield briefly reached 5.33%. CoreWeave shares fell about 15% Wednesday and another 12% Friday amid debt concerns.

Bridgewater Associates founder Ray Dalio called AI a classic bubble earlier this month and warned that borrowing to fund investment could leave the market vulnerable if interest rates keep rising. Galaxy Digital CEO Michael Novogratz also called AI a bubble, while saying it had not reached its most frenzied stage.

Marks, who correctly warned in early 2000 that the dot-com bubble had peaked, said investors may fear missing gains in technology stocks while still remaining alert to bubble risks. He said the evidence does not yet establish that a bubble has formed.

In a Dec. 9, 2025, memo, Marks listed circular financing among AI companies, debt-funded expansion, high valuations and fear of missing out as warning signs. Related transactions between large cloud providers and other AI companies total about $300 billion, with $650 billion in debt involved. Nvidia, Apple and Microsoft together account for about 21% of the S&P 500; in October 1999, nearly 10 companies were needed to reach the same share. The Shiller price-to-earnings ratio, based on 10 years of inflation-adjusted earnings, is about 41, near its 1999 record of 44.

Supplier-customer ties also featured in nearly half of AI-to-AI investment value from 2021 through 2025.

Marks’s counterargument is that the current market differs from the dot-com era. AI products have reached about one billion users, and leading companies have revenue, earnings and cash flow. He said AI leaders’ price-to-earnings multiples remain below the multiples Microsoft, Cisco and Oracle had during the dot-com bubble. Marks has also described technology-driven bubbles as capable of leaving lasting infrastructure even when investors lose money.

Markets will focus on the Federal Reserve’s rate path, AI companies’ debt and whether spending on AI turns into revenue and profits.