I. AI's Toxic Financial Trifecta
Recent AI jitters have been fueled by a potentially toxic trifecta of financial constraints: elevated borrowing costs, uncertain investment returns, and concerns about whether sufficient capital exists to finance everyone's ambitious AI spending plans. Consider the following:
(1) Elevated borrowing costs. The 10-year Treasury yield is hovering near the upper end of its "old normal" range of 4.0%-5.0% (chart). That puts it at the highest level since 2007. Overall, credit spreads (the extra yield paid over equivalent US Treasuries, such as the 10-year note) have not blown out uniformly across the board. Instead, the market has bifurcated sharply between the pristine balance sheets of mega-cap tech companies and the leveraged balance sheets of pure-play infrastructure/data center borrowers.
Yields on AI debt have not broadly surged relative to Treasuries for the dominant Big Tech issuers, whose balance sheets continue to be treated virtually like "proxy sovereigns." However, the credit market is increasingly imposing credit discipline, demanding significantly higher risk premiums and spreads from the leveraged second- and third-tier infrastructure borrowers that are building out the physical plumbing of AI. Increased concerns about the credit quality of AI borrowers have been reflected in widening credit default spreads (chart).
