In July, Amazon successfully issued a large-scale long-term bond worth $25 billion. The bond, with a maximum maturity of 40 years, carried an additional spread of 0.7 percentage points over the U.S. Treasury yield. Although it returned to the market just four months after issuing $37 billion in bonds in March, investor fatigue has been significant due to consecutive issuances.
Due to the accumulation of supply volumes, the bond issuance oversubscription multiple has dropped from a past level of 3–4 times to the mid-1x range. Nevertheless, Amazon is proceeding with bond issuance to meet capital expenditures totaling $200 billion for AI data centers and cloud infrastructure.
The large-scale bond issuance by big tech companies is not limited to Amazon. In 2024, the five hyperscalers — Alphabet, Amazon, Meta, Microsoft, and Oracle — issued a total of $20 billion in bonds, accounting for only about 2% of all investment-grade bond issuances.
However, amid intensifying competition in AI infrastructure investment, their combined bond issuance in 2025 surged to $94 billion. The upward trend continues into this year. As of July this year, the total volume of bonds issued by these five companies reached $132 billion, representing 11% of all investment-grade bond issuances.
AI capital expenditures by hyperscalers are expected to continue rising in the coming years. AI infrastructure spending, which was $416 billion last year, is projected to reach $777 billion this year and exceed 1 trillion won by 2027. Consequently, bond issuance by hyperscalers will also increase, worsening market supply-demand conditions.
Recently, Alphabet shocked the market by announcing that its free cash flow turned negative in the second quarter at -$5.9 billion, down from a positive $10.1 billion in the first quarter. Although Alphabet generated $39.1 billion in cash flow from operating activities, it faced deteriorating cash flows due to capital expenditures of $44.9 billion, far exceeding its operating cash generation.
Alphabet's long-term debt also rose to $98.2 billion in the second quarter, doubling compared to the end of the previous year. This indicates that AI infrastructure investment is putting pressure on the overall financial health of the company. Even Amazon and Meta, traditionally known as cash-rich firms, have recently seen their free cash flows turn negative.
It has become increasingly difficult to cover the rapidly rising capital expenditures with cash generated from operations. Only Microsoft, whose cloud business has grown rapidly, is barely maintaining a positive free cash flow. As cash flows decline, there will be insufficient funds for shareholder returns.
A more serious issue is that the surge in bond supply by hyperscalers is gradually depleting the market's capacity to absorb new bonds. These companies are absorbing market demand with credit ratings comparable to U.S. Treasuries. As a result, recent benchmark interest rates have been under continuous upward pressure.
Even the capacity to purchase U.S. Treasury bonds has diminished, pushing U.S. Treasury yields to their highest level since the 2008 financial crisis. The combination of illusions surrounding hyperscalers' network and AI growth potential is materializing into a reverse crowding-out effect that displaces Treasuries.
A massive leverage network has formed among semiconductor manufacturers, server hardware companies, cloud providers, and final model developers, linked with private equity (PE) funds, infrastructure funds, and banks. With high interest rates persisting, if a hole opens in one part of the network, there is a high risk that the entire system could be threatened. This is an area that investors must monitor closely.
in OPINION
