Howard Marks: Claude Rates Frontier AI Labs 2 Out of 10

Howard Marks asked Anthropic’s Claude to rate how predictable each part of the AI industry is, and it gave frontier labs such as OpenAI and Anthropic a 2 out of 10. The Oaktree Capital co-founder published the scores in his 11 November memo, “A Layman Thinks About AI.”

A line from the memo is circulating on X this weekend: “Perhaps the most startling characteristic of AI is the speed at which it makes progress,” quoted by Unusual Whales on 21 November. The memo itself is 10 days old.

The memo takes on the question behind our AI bubble page. Marks writes that answering it is “largely a matter of faith.”

Business Pill · Concentration Explained

An evergreen Business Engineer explainer. It teaches the general idea only and does not discuss this memo.

The key insight: Howard Marks’s 11 November memo quotes Claude rating frontier labs 2 out of 10 for predictability and AI as a whole about 3; Marks concludes that whether AI spending pays off is largely a matter of faith.

What Claude Told Marks

Marks wrote that he asked Claude to rank AI “on a scale from 1 (unknowable) to 10 (fully predictable).” Claude scored AI as a whole “about a 3.”

It split that score in two. The direction is “maybe a six or seven: the technology will keep improving and spread through the economy.” What investors get paid for is “around a 2: how fast it happens, how big the profits are, and who keeps them.”

By part of the industry, Claude gave regulated power utilities serving AI load a 6. Data centers leased long-term to big tech companies got 5 to 6, hyperscalers 5 and chips 4. Speculative or “neocloud” data centers got 2 to 3, and frontier labs 2.

Marks called the labs’ 2 the surprise, which he takes to mean “practically unknowable.” Claude’s reason, as he quotes it: “These businesses are close to binary. Capital needs are huge, prices per unit of capability keep falling fast, and it’s unclear whether a lasting moat exists.” Marks added: “Ironic, isn’t it, given that Claude belongs to Anthropic?”

Claude's predictability scores for parts of the AI industry, as quoted by Howard Marks: utilities 6, frontier
Claude’s predictability scores for parts of the AI industry, as quoted by Howard Marks: utilities 6, frontier labs 2. Ranges are plotted at their midpoint. Data: Oaktree memo, 11 Nov 2026.

Why Speed Is the Risk

Marks writes that he is “struck by the fact that today’s leading model labs, OpenAI and Anthropic, were unknown to most of us just a few years ago. If companies can forge ahead that fast in the world of AI, can they likewise be supplanted by others, or at least challenged?”

He points to the internet era: Microsoft, Amazon, Apple and Google came through as winners, while Napster, Yahoo and Netscape “fell by the wayside.” His test for anyone valuing an AI company is to guess its earnings per share in five years and ask whether they would bet the answer lands within 50% of that guess.

The Debt Question

Marks cites Goldman Sachs: AI-related debt issuance has totaled “nearly $600 billion globally so far this year,” with around $400 billion in the US investment-grade market, over one-fifth of all supply. That is about two-thirds of the total, by our arithmetic.

He expects AI to need “an average of $1 trillion of new capital per year,” and asks whether hyperscalers that borrow at any price are pushing up rates the government pays. He also flags lease “out clauses,” chip collateral that may not last as long as the debt, and “circular transactions” between chipmakers, labs and builders. We explain how that borrowing works here.

Is It a Bubble?

Marks says the technology “isn’t an illusion.” Whether the spending earns its return, he concludes, “is largely a matter of faith.” His key question: “will AI provide sufficient value to users to warrant sales of compute and prices for compute high enough to result in adequate returns on the trillions of dollars being invested in it (and for which companies)?”

The Structural Read

As we read it, the memo’s sharpest point is the split: the direction of AI is predictable, the profits are not.

As we read it, the debt section is the most concrete part: it names the risks lenders carry, from lease out clauses to chip collateral.

Howard Marks, Oaktree memo, 11 Nov 2026

“Perhaps the most startling characteristic of AI is the speed at which it makes progress.”

Three Implications

FOR INVESTORS Marks’s test: guess an AI company’s earnings per share in five years and ask if you would bet on landing within 50%.

FOR LENDERS Marks flags lease out clauses, chip collateral and circular transactions.

FOR THE LABS Claude gave frontier labs 2 out of 10, its lowest score, per the memo.

What Is Not Established

The scores are what Claude itself calls “judgment calls,” not data. The memo is dated 11 November and says OpenAI and Anthropic have both “filed for IPOs.” Marks writes that he is not trying to keep up with daily news, since his views relate “to the long-term outlook for AI, not what security prices will be tomorrow.”

This Week In The AI Supercycle

A World Without an Order, and an AI Central Bank Without a Successor

This week’s issue: the week AI memory stayed largely sold out while the people holding the shares paid: why the AI Supercycle is not a dot-com replay, read through geopolitics, macro and technology, with the gap between building and financing the widest we’ve recorded.

Read the full issue →

Business Engineer Framework

The AI Supercycle

The Business Engineer map of where AI capability turns into business leverage.

Read the Map of AI →

The Bottom Line

In his 11 November memo, Howard Marks asked Claude how predictable AI is and got a 2 out of 10 for the frontier labs, including Claude’s own maker. He writes that the right response is “calibrated humility” and “a willingness to act in the presence of uncertainty,” and that “intelligent investing is largely about price.” This is not investment advice.

97,000+ executives read Business Engineer for the AI strategy frameworks cited by ChatGPT, Claude, and Perplexity.

A note on sourcing. Every quote and score comes from Howard Marks’s Oaktree memo of 11 November 2026, which reports Claude’s answers as Claude’s own judgment calls. The Goldman Sachs debt figures are as cited by Marks. The two-thirds share and the memo’s age are our arithmetic. This is not investment advice.

Sources: Howard Marks, ‘A Layman Thinks About AI’, Oaktree Capital memo, 11 Nov 2026 · Unusual Whales on X, 21 Nov 2026

Scroll to Top
FourWeekMBA