AI may be one of the most transformative technologies in history, but investors face an unusually difficult challenge in determining who will actually make money from it, according to Oaktree Capital Management co-founder Howard Marks.
In a new memo published Wednesday, Marks described artificial intelligence as potentially the most consequential and least predictable development for investors in at least a generation, warning that the enormous capital flowing into the industry could expose financial markets to risks that are difficult to quantify.
“AI is the biggest development of our time and could be one of the most impactful things the world has ever seen. But investors don’t have the option to abstain from thinking about it," Marks wrote.
The veteran investor questioned whether today’s spending on AI infrastructure, chips and computing capacity will ultimately generate sufficient returns, particularly as companies compete to build increasingly expensive data centers and develop more advanced models.
Marks noted that AI could require an average of $1 trillion in new capital annually in the coming years, with hundreds of billions of dollars potentially raised through debt issuance.
According to Goldman Sachs data cited in the memo, AI-related debt issuance has already approached $600 billion globally this year. Roughly $400 billion came from U.S. investment-grade borrowers, accounting for more than one-fifth of total issuance in that market.
AI’s Winners Remain Unclear
One of Marks’ biggest concerns is that investors may be underestimating how difficult it will be to identify the eventual winners.
He pointed to the rapid emergence of OpenAI and Anthropic, questioning whether today’s leading AI companies will maintain their competitive advantages over the next five to 10 years or eventually be overtaken by new entrants and established technology giants.
Marks also questioned whether AI models will become commodities as competition drives down prices, potentially shifting the economic benefits toward businesses using the technology rather than the companies developing it.
To illustrate the uncertainty, Marks asked Anthropic’s Claude to assess how predictable different parts of the AI industry are on a scale of one to 10.
Claude rated AI overall a three, while frontier AI laboratories such as OpenAI, Anthropic and xAI received a score of just two. Regulated utilities supplying power to AI infrastructure scored six, while data centers with long-term leases to major technology companies received a five to six.
A Potential Fault Line For Markets
Marks raised additional concerns about circular financing arrangements between chipmakers, cloud providers, AI laboratories and data center developers, questioning whether some transactions reflect genuine demand or companies effectively subsidizing their customers.
Despite the risks, Marks stopped short of calling AI a bubble. He believes the technology will fundamentally change the economy, but said investors still lack sufficient information to determine whether current spending levels and valuations are justified.
“Clearly, there’s a lot investors don’t know and can’t predict about AI,” Marks wrote.
His advice was not to avoid the sector entirely, but to approach it with humility, carefully assess risk and recognize the limits of forecasting.
Ultimately, Marks argued, investing in AI requires understanding not only what could go right, but how much could go wrong if today’s assumptions prove incorrect.
Photo: Shutterstock