The hyperscalers placed $244 billion in bonds through 2026, more than double prior-year issuance. Microsoft, Amazon, Google, and Meta led the wave. Spreads are widening. Demand is no longer automatic.

The surge reflects capital intensity no operating cash flow can cover. AI infrastructure—data centers, GPU clusters, power contracts—requires front-loaded expenditure measured in tens of billions per quarter. Equity dilution remained unattractive at current valuations. Debt became the path. Investment-grade ratings allowed access to pension funds, sovereign wealth accounts, and insurance balance sheets that treat these names as quasi-sovereign. Until recently, those buyers absorbed every tranche. That absorption is ending.

Spreads widened 18 to 35 basis points across the curve in March alone. The 2033 and 2036 maturities saw the sharpest moves. Demand at auctions thinned. Bid-to-cover ratios fell below 2.1x on recent Microsoft and Amazon offerings, down from 3.4x averages in 2025. Buyers are not refusing participation. They are demanding compensation for duration and concentration risk. Portfolios already hold multiples of prudent single-issuer limits when hyperscaler exposures are aggregated. Adding another $80 billion in H2 2026—the run-rate implied by current issuance—tests institutional mandates and VAR models alike.

Second-order effects are measurable. Corporate bond ETFs show redemptions exceeding $12 billion in March, the largest monthly outflow since September 2022. Retail and wirehouses are rotating out of credit. Real-money accounts are extending duration elsewhere—utilities, industrials—where correlation to a single technology thesis is lower. The hyperscaler credit complex is becoming its own asset class, and allocators are treating it as such: a concentrated bet on AI monetization timelines, not a diversified bond portfolio.

Operators should track Q2 earnings calls for capex guidance revisions and any mentions of "capital efficiency." The May 15 to June 10 window will clarify whether issuance continues at current pace or moderates. Watch the 10-year Treasury around 4.35%—if it breaks higher, hyperscaler spreads will widen further regardless of credit quality. Sovereign buyers have started rotating into shorter maturities. If that shift persists through June, the bid structure that supported $244 billion in issuance will not support the next $244 billion.

The market priced hyperscaler bonds as if AI infrastructure were sovereign-equivalent. It is not. The correction is neither a default risk nor a credit event. It is the recognition that $244 billion is a portfolio problem, not a portfolio solution.