AI investment risk is spreading from stocks to bond markets

In short: More investors are backing the AI boom through bonds, not just stocks, and that is concentrating risk in a small group of big tech borrowers.

What's going on

Investors often try to spread risk by owning different types of assets, like stocks, bonds, and real estate. A Financial Times column argues that with AI, many of those choices now point back to the same idea. It is like buying different products that all depend on the same supplier.

In US stocks, the concentration is already clear. The column notes that more than 35 percent of the S&P 500 is made up of large companies that are seen as key winners from AI spending.

A similar pattern is building in bonds, which are basically loans that investors make to companies. Estimates cited in the piece suggest around $2 trillion of AI financing could come from the US investment-grade bond market alone, according to JPMorgan. “Hyperscalers” (the biggest cloud computing firms) could grow to nearly 25 percent of the $7 trillion to $8 trillion US high-grade bond market in coming years, up from about 5 percent now.

Risk is also rising in lower-rated corporate debt. Barclays estimates that AI and data center related high-yield bonds have grown to about $40 billion outstanding in a little over a year, close to 3 percent of the US high-yield index.

What to watch

The column says investment booms often end not because money disappears, but because expectations get too high. Even without defaults, bond prices can drop if investors start to doubt future AI demand. Signs of cooling interest may matter too, such as reported declines in how heavily some recent hyperscaler bond deals were oversubscribed (meaning more people wanted to buy than there were bonds available).

Source: Financial Times