AI hyperscalers are flooding the bond market with new debt to finance the acquisition of chips, construction of data centers, and other infrastructure. This has resulted in a significant increase in U.S. investment-grade corporate bond issuance, reaching approximately $1.7 trillion year-to-date through July, a 27% increase from the previous year and on track to surpass $2 trillion. This surge in corporate debt is attracting capital that would typically flow into US Treasuries, compelling Treasury yields to rise to attract buyers.
This phenomenon, dubbed "reverse crowding out" by analysts like Ed Yardeni, means that as investors allocate more capital to high-yielding AI-related corporate bonds, there is less demand for Treasury bonds. For instance, private foreign investors purchased $390 billion in corporate bonds over the past year, exceeding their $329 billion in Treasury notes and bonds, a more than 40% decrease from the prior year for Treasuries. This shift is notable as some AI companies are issuing debt at "yield-agnostic" rates due to the high anticipated returns from AI investments, further intensifying competition for investor capital.
The impact on the broader market is a concern for financial leaders, including Treasury Secretary Scott Bessent and Federal Reserve Chairman Kevin Warsh. The higher Treasury yields resulting from this competition contribute to increased borrowing costs for various sectors, affecting mortgages, car loans, and business financing. While quantifying the exact impact is difficult, Bank of America economists estimate that the surge in corporate debt sales, partly driven by AI, has pushed up 10-year Treasury rates by about 0.3 percentage points this year. Some investment funds are already adjusting, with bond funds limiting US investment-grade debt reducing their Treasury holdings and increasing corporate bond allocations to 30%, a three-year high.
Nomura Securities estimates that the roughly $200 billion in borrowing by the largest tech companies alone is equivalent to about 25% of the US Treasury's net issuance of notes and bonds to private investors, a five-fold increase from 2025. Companies like Amazon.com Inc. and Alphabet Inc. have increased spending forecasts, and Nvidia Corp. is working to raise another $500 billion for AI initiatives. Some investors are selling Treasuries to buy higher-yielding corporate debt; for example, Alphabet's recent 30-year debt was issued at nearly a 6.4% yield, 1.15 percentage points more than comparable Treasuries.
While some analysts, like those at Goldman Sachs, suggest the spillover from AI issuance has been limited, evidenced by stable credit spreads for non-AI companies, the growing supply of long-dated AI-related corporate debt poses a potential challenge for the Treasury. This could force the Treasury to reduce its own long-term debt sales to mitigate higher borrowing costs, a scenario that is becoming a real possibility according to Jonathan Cohn, head of US rates desk strategy at Nomura. The market is showing some signs of fatigue, with S&P Global noting that hyperscalers are paying higher premiums, and market participants are becoming wary of quickly rising leverage.
