08/11/2026
I saw something cool in two recent Howard Marks memos that I thought was worth sharing.
For anyone who doesn’t know Howard Marks, he co-founded Oaktree Capital Management and is one of the more respected investors of the last several decades. A lot of his writing focuses on risk, market cycles and investor psychology. Warren Buffett has said that when a Howard Marks memo shows up in his email, it’s one of the first things he reads.
In December, Marks wrote Is It a Bubble? about AI. He was clearly impressed with what AI could become, but he was also concerned about what investors were doing around it.
He talked about huge amounts of capital going into AI, the use of debt, questionable deals, uncertainty about who will actually make money, and the history of investors getting carried away whenever a technology comes along that looks like it could change the world.
His conclusion was pretty cautious. AI may be one of the biggest technological developments we’ve ever seen, but that doesn’t mean every AI investment will work. His advice was basically to have some exposure without getting carried away.
A few months later, he wrote another memo called AI Hurtles Ahead.
What happened between those two papers is the part I found really neat. Marks talked to people who understood the technology better than he did. Someone suggested that he have Claude build a tutorial for him based partly on his December memo.
So he did.
Marks said he viewed what Claude produced with “awe.” It understood his previous work, used his own investment concepts to explain AI to him, anticipated questions he might ask and reasoned through the answers. He came away with a different view of where AI may be headed.
By February he was writing that AI was advancing at a speed unlike previous technologies and was already being adopted much faster than most people probably realize.
And this was his conclusion:
“If I had to guess, I’d say its potential is more likely underestimated today rather than overestimated.”
That’s quite a change in tone in a few months.
And I like the way he got there.
He had an opinion based on what he knew in December. Then he kept learning. He talked to people who knew more than he did. He actually used the technology. He got new information and adjusted his thinking.
Investors have to be able to do that.
It’s easy to become attached to something you’ve bought, sold, predicted or said publicly. At some point you can find yourself defending your previous decision instead of evaluating what’s happening now.
There’s a quote often attributed to John Maynard Keynes:
“When the facts change, I change my mind. What do you do, sir?”
Marks didn’t suddenly decide every AI company is a great investment. He still has plenty of concerns about valuations, spending and whether the eventual profits justify the amount of money being invested.
He just learned more and updated his view.
That’s something every investor should be able and willing to do.
I’ll put both papers in the comments if anyone wants to read them. Start with Is It a Bubble? and then read AI Hurtles Ahead.
