The massive increase in corporate debt issuance to fund artificial intelligence infrastructure, particularly from "AI hyperscalers," is encountering growing investor fatigue and "indigestion," according to financial analysts. BNP Paribas data indicates that AI-linked debt issuance has soared to $220 billion in 2026, a substantial rise from $12.5 billion in the same period last year. This surge is pushing investors to demand higher yields, raising concerns about a potential tipping point if AI spending continues its rapid escalation. For example, Amazon's recent $25 billion bond sale priced at roughly 120 basis points over Treasuries, double the spread from a year prior, and tech bond spreads are now 89 basis points, nine basis points wider than the overall investment grade market.
Despite the strong credit quality of companies like Amazon and Alphabet, the sheer volume of issuance is creating challenges. Traditionally, investment-grade bonds would see tightening spreads in the secondary market, but recent AI-linked deals from companies like Nvidia, SpaceX, and Amazon have seen their debt spreads widen shortly after issuance. This poor secondary market performance makes money managers hesitant, prompting underwriters to adopt defensive strategies such as favoring "real-money accounts" like pension and insurance funds, which tend to hold bonds long-term, and spacing out debt sales. Alphabet, for instance, announced its latest bond sale would be its last in the U.S. market this year.
Institutional investors face practical limits, with many capping exposure to individual issuers at approximately 2% to 3% of their assets. As a handful of AI companies repeatedly issue large amounts of debt, these limits are becoming a significant constraint. While the investment-grade corporate bond index yields around 5.4%, offering attractive returns that support some demand from foreign investors, pension funds, and insurance companies, the market's capacity to absorb continuous, large-scale borrowing is being questioned. Analysts warn that if these companies keep tapping the market frequently, concessions will grow larger, and spreads will continue to widen, indicating that the era of unlimited investor demand for tech debt is over.
The demand for higher yields is also evident in some investment-grade AI-related debt attracting junk bond investors. For example, QTS Realty Trust Inc. sold $3.9 billion of investment-grade bonds yielding about 7.23% for a data center project tied to Microsoft, a yield higher than many middle-tier junk bonds. Similarly, BlackRock Inc. offered a 7.53% yield on blue-chip securities for a Texas data center project. This trend highlights the increasing concessions companies must offer to secure funding for their AI ambitions, with some bankers predicting another $50 billion to $60 billion in hyperscaler debt to hit the market after the U.S. Labor Day holiday.
