AI-driven corporate debt sales are significantly impacting US Treasury yields, with corporate offerings totaling almost $1.5 trillion year-to-date, a 36% increase from the previous year. This surge, largely from tech hyperscalers borrowing hundreds of billions of dollars for AI infrastructure, has prompted investors to question if this "crowding out" effect is now driving up interest rates in the bond market, traditionally a concern for government borrowing leaving little room for companies. PIMCO notes that these AI-linked bonds are now competing with government bonds for investor capital, contributing to a global trend of rising long-term yields.
The 30-year U.S. Treasury yield recently hit approximately 5.3%, a level not seen in nearly two decades, with similar rises in European and Japanese long-term yields. This increase is attributed to a combination of rising sovereign debt loads, increased AI-related corporate bond issuance, and lingering inflation concerns. While some argue that the direct impact of AI debt on overall Treasury yields is overstated, with policy expectations being a more dominant factor, the concentration of longer-dated issuance from hyperscalers is pushing up risk premiums in those specific segments. Hyperscalers have accounted for over 30% of 30-year issuance, totaling about $32 billion, compared to 13% of 10-year supply.
Investors are demanding higher yields for this AI-related debt, even for investment-grade offerings. For example, QTS Realty Trust Inc. sold $3.9 billion in bonds for a Microsoft-tied facility, yielding about 7.23% for high-grade ratings. Similarly, BlackRock Inc. is seeing a 7.53% yield on blue-chip securities issued in July for a Texas data center project. This suggests that the substantial volume of AI debt, particularly in the longer-dated segments and even for high-quality borrowers, is creating a scenario where investors require more compensation, thus contributing to the upward pressure on bond yields.
