Howard Marks Says AI’s Promise May Be Understated as Investment Returns Remain Unclear

Howard Marks says artificial intelligence could reshape business, but investors may need about a decade to learn whether profits from its infrastructure buildout will justify the money being spent.

The Oaktree Capital Management co-founder and co-chairman has described AI as a real technology already in broad use. In a Feb. 26, 2026, memo, he wrote that its promise was more likely to be underestimated than exaggerated. That assessment concerns the technology’s potential; it does not establish that investors are paying reasonable prices for AI-related assets.

In a September interview, Marks said investors can see enthusiasm for AI without yet being able to measure its business effects or likely profitability well enough to judge whether that enthusiasm has become irrational. “We certainly have exuberance. The question is whether it’s irrational,” he said.

He drew a line between optimism about AI’s capabilities and the prices attached to investments built around it. “However, that’s not the same as saying AI investments are on the bargain counter or even fairly priced,” Marks wrote.

Marks estimated investors may need about 10 years to find out whether AI infrastructure profits justify the spending. He has advised against putting everything into the theme or avoiding it altogether, favoring a moderate, selective approach while business results take shape.

That distinction matters because a useful technology does not guarantee that the money invested in its development will earn adequate returns. In Marks’s view, the unresolved issue is not only what AI can do, but also how much of that potential becomes profit relative to the cost of building the systems that support it.

His Dec. 9, 2025, memo described a familiar risk in technology booms: companies can invest too heavily, leaving excess infrastructure and assets priced higher than their eventual value supports. Marks warned debt could magnify the risks. Those risks are part of the broader debt and infrastructure pressures surrounding the AI buildout.

Marks treats the technology and its financing as separate questions. Broad use and transformative potential speak to whether AI matters; they do not settle whether investors will earn returns that justify capital spending or whether assets acquired during the boom are overpriced. His September comments leave open whether the exuberance is irrational.