Amazon, Microsoft, and the other AI hyperscalers issued $159 billion in bonds through the first quarter of 2026, a 47% increase from the same period in 2025. The debt fueled continued buildout of data center capacity, chip procurement, and power infrastructure as the race to monetize foundation models entered its third year. Microsoft alone accounted for $41 billion of the total, with Amazon close behind at $38 billion. Meta, Alphabet, and Oracle combined for another $52 billion, while newer entrants including CoreWeave and Lambda Labs raised $28 billion through convertible notes and senior secured debt.
The issuance came at a weighted average coupon of 4.72%, roughly 180 basis points below where similar credits traded in late 2023. Investment-grade ratings held firm across the cohort. Amazon and Microsoft maintained their AAA equivalent ratings despite leverage ratios climbing to 2.8x and 3.1x EBITDA respectively. Credit analysts at both Moody's and Fitch cited durable cash flows from cloud services and multi-year enterprise contracts as sufficient cover for the incremental debt load. Bond buyers included sovereign wealth funds, insurance balance sheets, and asset managers running core-plus fixed income mandates. Demand exceeded supply by 3.2x on average across the sixteen largest deals.
The concentration is now visible in index composition. AI-related credits account for 15% of the Bloomberg U.S. Corporate Bond Index as of March 2026, up from 9% twelve months prior. That figure excludes chipmakers like NVIDIA and TSMC, which would push the exposure past 18%. Portfolio managers running broad IG mandates are carrying duration and credit risk to a single thesis: that capital expenditure on AI infrastructure will generate returns sufficient to service the debt and justify the equity multiples. The math works if utilization rates hold above 68% and inference pricing stabilizes near current levels of $0.14 per million tokens. It breaks if either assumption fails. Fixed income allocators are effectively long the same secular bet as equity investors, but with asymmetric downside and no upside participation beyond par.
Allocators should watch two forward indicators in the next ninety days. First, the April bond calendar. If hyperscalers return to the market before June with another $40-50 billion in issuance, it signals accelerated capex plans and tighter credit spreads despite rising supply. Second, the utilization data embedded in Microsoft and Amazon's June quarter earnings. If Azure or AWS report sequential declines in compute hours sold per dollar of infrastructure deployed, credit analysts will revisit leverage tolerance. Fitch has already placed four smaller AI infrastructure credits on negative watch. A downgrade cycle would reprice the entire stack.
The debt is now large enough to move the index. When hyperscalers return to the primary market, IG spreads tighten 4-7 basis points on average across the complex. The capital is patient, the ratings are stable, and the buyers are deep. But the bet is singular, and the concentration is no longer theoretical.