Back in 1996, Alan Greenspan mused about "irrational exuberance" in markets. Three decades later, Howard Marks, the co-chairman of Oaktree Capital, is pretty sure about one half of that phrase. "There's no question about the fact that we have exuberance," he said. The real question, he adds, is whether today's exuberance is actually irrational.
Marks, a lifelong value investor, argues that nobody can really answer that question right now. He points out that no one has told him exactly what AI will be able to do, when it will do it, who will benefit, or how much profit it will generate. In his view, artificial intelligence is "a concept" whose parameters can't be pinned down. It might be "the most powerful force any of us have ever seen," but it's also the least specifiable.
Elevated But Not Expensive
Marks is quick to note that the market is elevated, but not unhinged. The S&P 500 trades at roughly 23 times earnings, compared to an 80-year average of about 16. That's about 50% higher, which sounds like a lot. But it's still well below the 32 times earnings at the peak in 2000, and it's nowhere near the multiples of 60 to 90 that the Nifty 50 stocks fetched in the late 1960s. Back then, you had companies like Xerox (XRX), IBM (IBM), Polaroid, and Eastman Kodak (KODK) trading at those nosebleed levels.
Even the Magnificent Seven, excluding Tesla (TSLA), are sitting in the 30s. "That doesn't sound so expensive to me," Marks said. He also cautioned that looking at multiples alone is "too simplistic." Today's giants are less capital-intensive than the industrial behemoths of the past. Their product is intellectual rather than "a piece of metal," so incremental profitability is much higher. The growth rates, he said, are unlike anything he has witnessed in his career.
But beneath the surface of the equity market, Marks sees a long tail of easy money. Forty-plus years of falling interest rates and 17 years of economic expansion since March 2009 have been "salad days" for alternative investments. That kind of environment can encourage risk-taking that might not be fully justified.
Every Revolution Has Its Bubble
Marks places AI in a long lineage of transformative technologies. Think railroads in the 1860s, radio in the 1920s, computers in the 1950s and '60s, and the internet in 2000. Each one brought excitement, a winner-take-all race, and capital that "flowed in like water"—too much of it, into too much infrastructure, at prices that were too high.
"If this technological innovation with its exuberance doesn't produce a money-losing bubble, it'll be the first," he said. And to those who say "this time it's different," Marks notes that's what "they always say."
He also makes a crucial distinction between efficiency and profit. He cites Warren Buffett's observation from 2000 that the internet would add to efficiency but not necessarily to profitability. If hyperscalers, OpenAI, Anthropic, and a bunch of startups all "engage in battle" at enormous cost, who actually captures the gains? If AI is mainly a labor-saving device, Marks suggests the real beneficiaries might be shipping companies, retailers, and warehouses—the customers of a price war, not the combatants.
A Spectrum, Not a Binary
Marks frames AI exposure as a spectrum, not a simple yes-or-no bet. At one end, you have the hyperscalers like Amazon (AMZN), Alphabet (GOOGL, GOOG), Meta (META), and Microsoft (MSFT). These companies have moats and vast cash flow, but their businesses are so diversified that even if AI "octuples," the gains might be blunted.
Then there are the established leaders like NVIDIA (NVDA), OpenAI, and Anthropic. Marks thinks these companies are unlikely to be obsoleted, but he'd bet that anyone forecasting Anthropic's earnings for 2036 won't land within 50% of the truth. And at the far end of the spectrum, you have startups, which he describes as "lottery tickets." Most holders will lose everything, but a few will become incredibly rich.
The task for investors, Marks says, is to recognize where an asset sits on this spectrum and calibrate your activities accordingly. Accept that much of this is closer to speculating than analytical investing.
Forecasts Need Probabilities
Marks insists that dealing with the future requires two things: a forecast and a judgment of its probability of being right. Investors who are "highly confident" about AI's trajectory are "probably making a big mistake."
But inaction is its own error. "One of the greatest mistakes you can make is being not optimistic enough," he said. Another is declaring the future too unclear to invest at all.
Borrowing from Nassim Taleb, Marks contrasts investing with dentistry. When you fill a cavity correctly, it works every time. Investing has no such physical rules. Even Warren Buffett credits his record to about 12 great decisions. So if you're the kind of person who wants to be successful every time, Marks suggests you "become a dentist."
In the end, Marks isn't saying AI won't change the world. He's saying that the path to that transformation might be bumpier than the current market enthusiasm suggests. And for investors, the key is to understand what you're betting on and how much you're paying for the privilege.
