Howard Marks Softens His AI Skepticism as Market Stumbles Challenge the Sector’s Aura of Certainty

Recent market setbacks tied to artificial intelligence have weakened the sector’s aura of infallibility, Oaktree Capital Management co-chair Howard Marks said in an interview with Bloomberg TV. His comments place a prominent investor’s caution inside a broader reassessment of AI enthusiasm, where impressive technical demonstrations now compete with questions about valuation, reliability and business returns.

Marks has spent decades studying speculative cycles and investor psychology, making his comments notable beyond any single technology stock. Bloomberg’s account describes a partial bursting of the belief that AI companies and projects should be treated as economically untouchable. Other accounts portray his position as more nuanced: he has revised his view of AI’s capabilities while continuing to avoid a definitive call on whether the sector represents a bubble.

Marks’ reassessment followed a tutorial from Anthropic’s Claude, according to accounts published by Fortune and republished through Yahoo Finance. The exercise left him impressed enough to describe a profound level of awe in a client memo, a striking shift from his earlier skepticism about the technology’s significance.

The change concerns what AI systems can do, especially their growing autonomy and ability to perform complex intellectual tasks. Marks reportedly prepared an earlier memo examining whether AI represented a financial bubble, then revisited his conclusions after challenges from his son, who works in venture capital. MarketWatch’s headline similarly characterizes the shift as a 180-degree turn after the Claude tutorial, though Marks’ reported comments retain clear limits around what AI can understand and how investors should price it.

Marks’ updated view separates AI’s technical power from the investment case surrounding it. He has acknowledged that investors underestimated what AI systems can accomplish, while declining to treat that capability as proof that current valuations are justified. Bloomberg’s reporting on his comments centers on skepticism returning after recent stumbles, while the podcast summary describes an “AI upgrade” focused on autonomy rather than valuation.

That distinction matters because technology adoption and shareholder returns follow different paths. A system can perform increasingly sophisticated work while companies supplying it face intense competition, heavy computing costs and uncertain pricing power. Marks’ refusal to label AI definitively as a bubble or not a bubble reflects the investor’s emphasis on disciplined judgment under incomplete information, rather than a simple endorsement or rejection of the sector.

Marks still draws a line between AI’s ability to process information and the judgment investors use when conditions change. The podcast account says he questioned whether second-level thinking, the ability to reason about how other participants will interpret events, can be taught to an AI system. That concern preserves a central part of his investment philosophy even after his view of AI’s capabilities improved.

His perspective also connects the debate to experience rather than raw computing power. A separate account highlights Oaktree’s deployment of roughly $450 million a week during the 2008 financial crisis, when the firm had little dependable data and had to act amid extreme uncertainty. The example frames investing as a discipline involving judgment, timing and emotional control, areas where impressive demonstrations from Claude do not settle the question of machine insight.

The episode gives investors two separate signals: AI deserves more respect as a technology, and AI-linked assets deserve more scrutiny as investments. Marks’ comments arrive after market disappointments weakened confidence in automatic gains, making room for skepticism without requiring a rejection of the underlying innovation.

That recalibration also affects how executives and fund managers assess spending. Enthusiasm can support aggressive infrastructure commitments, but weaker market confidence raises the standard for proving demand, margins and durable competitive advantages. Bloomberg’s report on renewed skepticism and the accounts of Marks’ changed outlook point to a market entering a more selective phase, where technical progress alone carries less weight than evidence of durable economic value.

AI markets are likely to be judged through a sharper split between demonstrated capability and realized cash flow. Marks’ revised opinion gives the technology greater credibility while his continuing reluctance to make a valuation call keeps the investment debate open. The same tension will shape decisions around AI stocks, private funding and corporate adoption.

For companies building AI products, the immediate test is practical performance: reliable output, useful autonomy and a business model that survives expensive infrastructure. For investors, the test is whether those gains translate into defensible earnings rather than another cycle of expanding expectations. Marks’ message is therefore neither a retreat from AI nor a blank check for it. It is a demand for better judgment after the sector’s first wave of certainty has worn down.