Howard Marks on market cycles, contrarian investing, and why AI is now his mentor

Summary

Howard Marks, co-founder of Oaktree Capital Management, explains market cycles through a simple lens: underlying economic value follows a steady trend, but prices swing wildly because human psychology never changes. Greed, envy, and FOMO drove the tulip bubble in 1620 and the South Sea bubble in 1720, and the same emotional mechanics repeat in every excess since. History does not repeat, but it rhymes.

Contrarian by nature

Oaktree operates primarily in credit and distressed debt, not equities. Marks describes himself and co-founder Bruce Karsh as innately contrarian, most comfortable betting against the herd at its extremes. The disposition traces back to 1978, when Citibank asked him to start a high-yield and convertible bond fund. Being a lender demands conservatism, and he says he would have failed running a venture capital fund. His October 2008 memo titled "The Limits to Negativism," written after the Lehman bankruptcy, best captures that instinct reversed: recognizing when pessimism itself becomes the excess to fade.

Contrarianism can be learned intellectually but may not drive action without the right emotional wiring. Marks is never certain when he makes a call at a market extreme. "Anybody who's sure is an idiot," he says. But logic and temperament together make it workable.

AI as mentor

Marks has become a serious student of AI despite identifying as a conservative credit investor. His son Andrew, co-founder of TQ Ventures, pushed him to update an AI memo he published in December. They developed a detailed prompt together and fed it to Claude, which produced educational material in a field new to Marks. When he sent his updated memo to Claude for review rather than Andrew, Claude responded with what Marks calls a note that felt like a colleague wrote it—warm, personal, drawing on analogies to his 35 years of memos and even injecting humor. When he asked Claude to be "hypercritical," it replied: "Do you want me to be hypercritical or hypocritical?"

Marks views AI's role in investing carefully. It marshals data with thoroughness, speed, and accuracy Oaktree never had before. It frames hypotheses on questions like credit repayment probability. But humans still need to check those hypotheses, particularly because AI is weakest at analyzing genuinely novel situations where no established patterns exist. That, Marks notes, is much of what credit investing actually requires.

The parallel he draws is to indexation: passive investing displaced the underperforming active managers who charged high fees for inferior results. AI will do something similar—raise the floor, eliminate the weakest practitioners, but not replace the best.

AI's scale and speed

Marks believes AI is categorically different from every prior technology wave, including the internet. The power is vastly higher, and innovation moves faster than anything he has observed. He points to a recent ChatGPT model launch where OpenAI disclosed that AI helped design the model itself, something with no historical precedent. Unlike the railroad or internet, which were labor-saving devices that did existing jobs better, AI will design new jobs, assign new tasks, take on work it was never asked to do, and eventually operate without instruction.

He sits on the worried side of the societal debate. He can enumerate jobs AI will eliminate but struggles to imagine the new ones created, so he expects a net decline. The speed of AI innovation is outrunning society's ability to adjust, creating at minimum a significant period of dislocation. He also rejects the optimistic framing that people won't have to work. "To me, that's terrible news. We get a great deal from our work other than a paycheck."

China and the cost of scruples

Marks frames the US-China AI race as a question of strategic dependency rather than technology gaps. Rare earths offer the clearest example. The US held primacy in rare earth mining, much of it from a single California mine, and ceded it to China after environmental concerns made extraction politically unacceptable. China took the business, and the dependency followed. Germany's decision to shut nuclear reactors and buy Russian gas parallels the same mistake. Both decisions were made on economic and ecological terms while excluding the geopolitical dimension.

AI follows the same logic. If the US slows development because of ethical constraints while China continues without them, the result could be a strategically superior Chinese AI position. Marks describes it as an arms race. America has never previously faced a serious economic rival—Russia was a military and nuclear threat but not an economic one. China represents something genuinely new.

Portfolio construction

Marks dismisses any stock-bond allocation rule with specific numbers in it. The right risk posture depends on age, wealth, income sufficiency, aspirations, number of dependents, and intestinal fortitude—the ability to live with fluctuations. The directional logic embedded in such rules is reasonable: younger investors can tolerate more uncertainty, older investors approaching retirement should want more dependability. The specific numbers are not.

Oaktree's fundraising tracks the distress cycle directly. When distress is widespread, there is more to do and more to raise. When markets are placid, Oaktree raises smaller funds and operates selectively. In good times, when most investors are pressing forward, Oaktree pulls back. Marks quotes Buffett: "The less prudence with which others conduct their affairs, the greater the prudence with which we must conduct our own." The scariest market condition is the belief that there is no risk. The worst loans in banking are made in the best of times, for exactly that reason.