AI’s growing sway over stocks and bonds

A retired-investor column by Bill Schmick questions whether the artificial intelligence boom has gone too far, too fast, noting that traders have recently been selling first. The author argues that equity investors have bid up AI-related stocks for more than a year in a feedback loop in which higher earnings expectations drove higher prices, which in turn reinforced earnings expectations. Third-quarter earnings were underway with the market looking for roughly a 30 percent year-over-year jump in results. Citing Goldman Sachs' head equity analyst, the column notes that more than half of third-quarter earnings are expected to come from 10 stocks, with more than one-third from two companies.

The column says the bond market shows the same concentration. AI-related transactions accounted for over 50 percent of investment-grade corporate issuance as of the end of August, and demand was strong enough that a Microsoft bond was preferred over government bonds of the same duration. Apollo Management estimates the bond market will raise $1.8 trillion by year-end. The column links this supply, including U.S. government borrowing, to rising bond yields.

The column also points to $1.65 trillion in off-balance-sheet obligations taken on this year by Microsoft, Alphabet, Amazon, Meta, and Oracle. These include long-term operating leases for data centers not yet built, bulk GPU supply contracts, and joint ventures. Under current accounting rules, such liabilities need not be recorded until assets are delivered or go live.

Citing a recent International Monetary Fund report, the column flags hedge funds as major holders of AI stocks and bonds and U.S. Treasuries. It describes these leveraged entities as having doubled their assets in six years to about $13 trillion, with a little more than half borrowed, and a share of the Treasury market above 12 percent. The author warns that these funds invest similarly and may sell en masse if trades turn against them. Their combined assets under management plus derivatives face value were estimated at $42.2 trillion in the first quarter.

The author concludes that AI's long-term promise is real, but says the sector can overshoot. He does not expect a protracted decline like the dot-com bust, when 50 percent or more of companies failed in the early 2000s, but identifies AI concentration risk in both bonds and stocks, with big tech roughly 50 percent of the S&P 500, as the central worry.