Large technology companies, often referred to as "hyperscalers" (including Alphabet, Amazon, Meta, Microsoft, and Oracle), are increasingly relying on the bond market to fund their colossal investments in AI infrastructure, leading to a significant increase in their debt. Morgan Stanley estimates that these companies will raise $400 billion from the US high-grade market in 2026, a substantial leap from $170 billion in 2025 and $44 billion in 2024. Apollo Global Management projects that by 2030, hyperscalers will constitute half of the ten largest borrowers in the US investment-grade corporate bond market. This borrowing surge has pushed up borrowing costs for some while exposing the bond market to greater risk from AI-related investments.
Concerns are mounting among investors that the vast capital expenditures on AI might not generate proportional returns, potentially creating a bubble similar to the dot-com era. For instance, Bank of America notes that the capital expenditure for Oracle, Alphabet, Microsoft, Amazon, and Meta is currently exceeding their free cash flow. Oracle, in particular, has seen its debt rating downgraded by S&P Global to just one notch above junk bond status due to deteriorating finances linked to its heavy AI spending, and its credit spread jumped over 0.75 percentage points after an $18 billion bond issuance. These financial pressures have led to Oracle's stock falling over 35% this year.
The demand from bond investors for these tech giants' debt is also showing signs of weakening. Amazon, for example, had to offer an additional 18 to 21 basis points of yield on a recent $25 billion bond sale because investor demand was only 2.5 times the bonds on offer, down from 3.2 times in March. JPMorgan strategists highlight that the AI and data center sector now accounts for 14.5% of the JPMorgan US Liquid Index, a benchmark for the nearly $10 trillion US investment-grade bond market, surpassing the share held by banks. Some analysts, like Savita Subramanian of Bank of America, draw parallels to the dot-com bust, cautioning that while hyperscalers have stronger business models, their immense borrowing needs are "a little nerve-wracking."
