AI hyperscalers issued $159 billion in corporate bonds during the first quarter of 2026, a 47% increase from the same period last year. Amazon, Microsoft, and their peers now account for 15% of the entire corporate bond market, a concentration level last seen in telecom debt before the 2001 crash.
The acceleration reflects a straightforward constraint: AI infrastructure build-outs are consuming cash faster than operating income can cover them. Amazon's $14 billion Canadian offering — the largest corporate bond in Canadian history — includes a $4.75 billion 30-year tranche, priced to lock in rates before the next Federal Reserve decision. Microsoft, Alphabet, and Meta issued similar tranches in January and February, each sized above $10 billion. The average maturity of AI-sector debt issued this quarter is 18.3 years, compared to 12.1 years for the broader investment-grade market.
The market is absorbing this supply without meaningful spread widening. Amazon's Canadian bonds cleared at 135 basis points over government benchmarks, tighter than comparable telecom issuers in 2000 by roughly 40 basis points. Fixed-income desks cite two reasons: hyperscaler balance sheets remain fortress-grade, with debt-to-EBITDA ratios between 0.8x and 1.4x, and pension funds are starved for duration. A $4.75 billion 30-year bond from a AA-rated issuer fits liability-matching mandates that have gone unfilled since Treasury issuance flattened in 2024.
The risk is not credit quality. It is concentration. When a single sector represents 15% of the corporate bond market, systemic repricing events become self-reinforcing. If AI revenue growth disappoints — or if capital expenditure does not translate to margin expansion within the market's expected timeframe — spread widening in one name forces mark-to-market losses across pension and insurance portfolios, which then reduces appetite for the next issuance. The telecom parallel is not hypothetical. In 2000, telecom debt reached 13% of the corporate bond market. By 2002, over $300 billion of that had defaulted or restructured.
Allocators should monitor two indicators over the next six months. First, whether hyperscalers begin to extend maturities beyond 30 years, signaling they expect capital intensity to persist longer than current models assume. Second, whether investment-grade spreads in the AI sector begin to diverge from each other. Right now, Amazon, Microsoft, and Alphabet trade within 20 basis points of each other. Divergence would indicate the market is pricing execution risk, not just sector exposure.
The $159 billion is not a warning. It is a fact pattern. The hyperscalers are borrowing because they can, and because the build-out permits no pause. The bond market is still betting they are right.