Howard Marks sees a hidden risk in the AI boom

02 Nov 2026, 18:33 · The Fresno Bee

Artificial intelligence investors face a strange choice.

If AI lives up to its potential, investors can own the businesses developing the technology and perhaps profit greatly. Or they can lend such companies money in exchange for a certain interest rate.

Howard Marks is beginning to wonder whether the second transaction makes sense.

The billionaire co-chairman of Oaktree Capital Management discussed the growing AI debt boom on the Oct. 29 episode of The Economist’s “Money Talks,” which focused on the risks of a potential financing blow-up.

Marks is worried about more than just excessive borrowing by AI firms. He’s concerned that although their prospective returns remain limited, lenders could be taking on undue risk.

Theoretically, stockholders of an AI firm may benefit from years of increasing earnings and value if the company succeeds.

Typically, a lender gets the agreed-upon principle and interest. And although debt often takes precedence over equity in a company’s financial structure, lenders may still lose money if things go wrong.

That creates a more significant dilemma for the typical investor: Are some purportedly safer strategies aimed at profiting from the AI boom riskier than they seem?

Howard Marks questions the reward for taking AI risk

Marks has spent decades investing in credit, distressed debt, and companies where assessing downside risk matters as much as finding upside.

That makes his AI argument unusual.

In essence, a seasoned debt investor is asking whether debt is the best option for taking on some of the largest risks associated with the technology.

In the past, Marks argued that buying shares in an AI firm might make more sense if investors are taking on basic business-model risk, since shareholders share in the profits if the company succeeds.

The same limitless upside is not available to debt investors.

This does not imply that stocks are safer. In the event that a company fails, stockholders often rank below lenders and risk losing their whole investment.

The difference has to do with making up for uncertainty.

AI firms are attempting to forecast demand, processing expenses, and technologies that might undergo significant change in a few years. In contrast, some loan financing extends for decades.

Marks wants to know whether a fixed return sufficiently compensates investors for accepting that degree of uncertainty.

For months, Oaktree has brought attention to the problem. Marks wrote in February that the speed at which AI was developing made it very difficult to forecast how the sector will look even a few years from now.

“If you want to invest in AI, you should probably buy the stocks,” Howard Marks said on the podcast.

Bloomberg / Getty Images

AI companies are asking Wall Street for enormous sums

The theoretical issue is quickly turning into a practical one.

Recently, Blackstone and Wall Street banks agreed to provide approximately $63 billion to fund networking equipment and Broadcom (AVGO) processors for Anthropic.

Roughly $43.5 billion of that package consists of senior first-lien loans arranged by Bank of America, Citigroup, and Morgan Stanley, according to Reuters.

Broadcom is also providing support against declines in the residual value of the equipment.

That is one example of a finance structure for AI that is becoming more complex.

Large financial reserves were originally used by IT corporations to finance a large portion of the AI boom. These days, banks, debt markets, private credit funds, and special-purpose financing entities play a much larger role.

More Wall Street:

The pressure is starting to show.

Bloomberg reported that yields on some debt financing AI infrastructure have climbed into the double digits, while lenders and banks are finding it harder to distribute some of that risk.

Additionally, the price of default insurance for investment-grade corporate debt has hit its highest point since March.

These indicators do not imply that an AI credit crisis is unavoidable. However, they imply that investors are starting to ask for more reimbursement for funding the expansion.

Average investors may own more AI debt risk than they realize

An AI business will never get a direct loan from most individual investors. However, the funding boom may still affect their portfolios.

Funding for AI infrastructure is increasingly coming from corporate bond funds, credit funds, insurers, and other organizations. As a result, retirement investors may be indirectly exposed to the same credit markets that are funding the rise.

It is simpler to see the equity side. Major positions among widely held stock funds include Alphabet (GOOGL), Microsoft (MSFT), Amazon (AMZN), Meta Platforms (META), Oracle (ORCL), and Broadcom.

As a result, an average investor may be exposed to AI risk in two very distinct ways.

In exchange for potentially significant rewards, stockholders willingly tolerate volatility. And investors in debt often anticipate reduced risk and steadier returns.

If the distinction between the two becomes hazy, Marks’ caution is important.

When lenders are expected to assume technological risk without benefiting from the same upside as stockholders, borrowing AI becomes less appealing.

Additionally, there is a significant counterargument. In the event that a firm has difficulties, debt holders often get contractual interest payments, have more legal protection, and take precedence over ordinary investors.

These safeguards may nevertheless make high-quality AI-related debt appealing to conservative investors at the appropriate rate.

Price is crucial because of these factors.

Marks didn’t advise steering clear of all AI bonds and loans. Rather, he suggested that investors ensure the risk and reward are equal.

That difference may prove one of the most crucial tests of the boom as billions of dollars pour into AI infrastructure.

Investors are looking beyond whether AI succeeds to ask if they’re being paid enough for the particular risk they are taking.

The Arena Media Brands, LLC THESTREET is a registered trademark of TheStreet, Inc.

This story was originally published November 2, 2026 at 10:33 AM.