A surprise $25 billion bond offering from Amazon has pushed artificial intelligence-related borrowing toward $270 billion, yet thinning order books, wider yield concessions, and shrinking free cash flow expose how heavily the cloud giants now rely on the bond market.
The sale, characterized by Bank of America as a surprise transaction, lifts Amazon’s total borrowing to $92 billion over recent months. Across the broader market, AI-related debt now stands at approximately $270 billion over the same period—nearly double the $136 billion issued in all of the previous year. Hyperscaler bonds account for an estimated $194 billion of that total, and Abishai Financial Asia views the sale as a marker of how sharply the sector’s funding needs have shifted.
This offering comes only months after an earlier $37 billion deal, an interval Bank of America considers unusually compressed. Amazon has signaled to underwriters that it intends no further issuance in the near term. Order books closed at 2.5 times the amount offered, down from 3.2 times on the earlier deal and the softest cover for any hyperscaler since a comparable $30 billion sale from Meta late last year. To attract buyers to maturities as long as 40 years, Amazon conceded 18 to 21 basis points above comparable securities.
The individual deal sits within a broader repricing of technology credit. Morgan Stanley expects annual AI-related issuance to rise to $570 billion. Daniel Coventry, Director of Private Equity at Morgan Stanley, described a sector that has “moved decisively beyond operating cash flow and into the bond market as a standing source of finance,” a shift he believes warrants closer monitoring of aggregate exposure budgets. The five largest hyperscalers have together raised $121 billion in US corporate bonds over the past year, compared to an annual average of $28 billion across the preceding four years—a 332% increase.
Amazon’s reliance on debt reflects an equally rapid rise in spending. Web Services revenue reached $35.6 billion in the most recent quarter, a 24% increase year-over-year. CEO Andy Jassy stated publicly that the division faces greater demand than it can currently supply, with chip availability the binding constraint, even as the group directs $43.2 billion toward cloud and generative AI capacity in the opening quarter alone. Committed capital expenditure now stands at $200 billion, while trailing free cash flow has contracted to $1.2 billion from $25.9 billion a year earlier. Capital spending now absorbs 94.5% of operating cash flow.
Comparable pressures run across the sector. Oracle has raised $27.7 billion over the past year, and Meta recently arranged a $27 billion private credit facility with Blue Owl to fund a data center campus in Louisiana. Alphabet has placed the first 100-year bond from a technology issuer in decades. S&P Global warns that credit quality faces erosion should the anticipated recovery in cash flow fail to materialize. With Oracle already carrying a BBB rating on a negative outlook, Coventry observed that “an investment grade label alone is no longer a sufficient guide to the risk embedded in these balance sheets.”
The strain reflects a widening gap between what cloud providers earn and what they spend. The five largest are on course to commit some $750 billion to capital expenditure—equivalent to 38% of revenue—and set to approach $1 trillion within four years. Quarterly operating cash flow of $26 billion sits against outlays of $44.2 billion. Rate expectations offer only a partial offset, with global investment grade credit currently yielding around 4.4%. Coventry maintains that “disciplined exposure management, rather than passive reliance on a rating, is the more defensible posture while issuance continues at this pace.”
Set against secondary markets where roughly $50 trillion of US corporate bonds trades barely $43.5 billion on an average day, the build-out leaves institutional investors weighing resilience as closely as return. Hedging through credit default swaps has climbed sharply in recent quarters, even as secondary depth in the newest paper stays thin. Widening yield premiums and thinning cover point to a durable change in how that risk is priced, one that rewards active management of single issuer limits over confidence in a rating alone. Abishai Financial Asia reads the present wave of infrastructure-led issuance as a test of capital discipline that is only beginning.
About Abishai Financial Asia
Abishai Financial Asia Pte. Ltd. (UEN 201016239E) is a Singapore-based asset manager and research-led partner in capital allocation. Its approach compounds capital in public markets through active equity selection, bottom-up research, and disciplined rebalancing, supported by overlay tools such as systematic tilts, opportunistic hedging, and drawdown-aware controls. Governance rests on macro-aware risk budgeting, with explicit limits, concentration and liquidity guardrails, stress testing, and transparent attribution. Environmental, social, and governance factors are embedded wherever financially material. The firm is exploring compliant structures that could, subject to suitability, extend selected solutions to retail qualified investors over time. Further information is available at https://abishai.com; media enquiries may be directed to Peng Joon at p.joon@abishai.com.