The Oaktree co-founder argues that rapid AI progress, Big Tech competition and enormous capital needs make today's leaders vulnerable

Sat 21 Nov 2026 · 15:06 EST · 3 min read

In AI

Share

Howard Marks is known for urging investors to stay humble when everyone else feels certain. The Oaktree Capital Management co-founder has now applied that habit to artificial intelligence.

In a memo dated November 11, 2026, titled “A Layman Thinks About AI,” Marks argues that fast AI progress and competition from major technology companies could threaten the lead held by OpenAI and Anthropic. Getting there first, it turns out, does not guarantee staying there.

A self-described layman asks the hard questions

The memo’s title is a modest bit of framing. Marks casts himself as an outsider to the technology. That lets him set the engineering debates aside and focus on risk, capital and the odds of getting paid back.

His central worry is durability. When a technology improves this quickly, any advantage a company builds today can be erased by the next breakthrough.

That breakthrough does not have to come from a scrappy upstart. Major technology companies have the resources to compete directly with the current frontier labs. Marks flags that rivalry as a real threat to the existing pecking order.

One of the memo’s more striking details involves Anthropic’s own product. Marks includes input from Anthropic’s Claude model, which gave frontier AI labs a predictability score as low as 2 out of 10.

Marks also looks back at earlier technology shifts to make his case. He cites Microsoft and Amazon as examples of companies that turned an early position into lasting success. Napster and Yahoo appear as the cautionary counterweights.

AI, tech, and the markets they move—in one daily briefing.

Daily. Free. Join 34,000+ readers across crypto, finance, and policy.

We respect your privacy. Unsubscribe anytime.

The bill is enormous, and it keeps growing

The second pillar of the argument is cost. Marks cites a projected annual investment requirement of $1 trillion for AI.

Spending at that scale only makes sense if the revenue on the other side eventually justifies it. Marks questions whether these huge infrastructure outlays will produce adequate returns, and he notes that heavy capital needs could weigh on profitability.

Debt markets are carrying much of the load. Goldman Sachs data indicates that nearly $600 billion in global AI-related debt issuance is expected in 2026, including around $400 billion from the US market.

Then there is concentration. About 42 AI-related companies have accounted for roughly two-thirds of the S&P 500’s price gains since January 2024.

A large share of the broad index’s recent performance rests on one theme. Many investors who think they hold a diversified fund may have more AI exposure than they realize.

What this means for investors and the AI leaders

For OpenAI and Anthropic, the memo reads less like a forecast of failure than a reminder that their position is contestable. Marks describes fast-moving technology, rivals with deep resources and very high capital needs.

Marks warns that the landscape could change sharply if new competitors arrive. For investors, his message is about posture rather than timing. The memo calls for a humble, measured approach to a sector where uncertainty runs high and the cost of staying in the game keeps rising.

That points to a few things worth watching. One is whether AI revenue starts to close the gap with AI spending. Another is how the 2026 debt pipeline holds up. A third is whether the 42-company concentration in the S&P 500 deepens or starts to unwind.

None of this makes AI a mirage. Marks’ own examples include companies that rode a technological shift to enormous success. His point is that spotting the trend is the easy part. Telling the future Amazon from the future Yahoo, while both are still burning cash, is much harder.

Disclosure: This article was edited by Diego Almada Lopez. For more information on how we create and review content, see our Editorial Policy.