AI-linked corporate debt has crossed 15% of the total U.S. corporate bond market, creating a concentration pattern that replicates the equity overhang investors have spent two years trying to hedge around. The figure surfaced in recent concentration risk analysis and marks the first time a thematic debt cohort—rather than a traditional sector like energy or financials—has achieved double-digit market weight without regulatory attention.
The debt comes from the same names driving equity concentration: Microsoft, Alphabet, Amazon, Meta, and the cloud infrastructure builders financing model deployment at scale. Combined, these issuers have placed $340 billion in AI-adjacent debt since early 2023, according to Bloomberg terminal data cross-referenced with disclosed use-of-proceeds language. Amazon alone is raising $14 billion in Canada this week, the largest Canadian corporate offering on record, with $4.75 billion in 30-year paper—a duration bet that assumes model training economics hold for three decades. RBC, TD, Scotiabank, and JPMorgan are leading.
The concentration matters because corporate bond investors thought they were buying diversification away from equity volatility. They are not. The same five names now govern price action in both markets, and the debt sits lower in the capital structure with less room to run if AI economics disappoint. Duration risk compounds the problem: most of this paper is 10-year or longer, meaning rate sensitivity is high and refinancing windows are distant. If model monetization stumbles or capex pivots, bondholders eat the first loss while equity holders debate the pivot.
Financial bond issuance added 13% to total corporate financing in April, led by bank paper funding the lending side of AI infrastructure. That means the concentration is bifurcated: tech names funding buildout, banks funding the loans against it. Both depend on the same revenue assumption—that enterprises will pay escalating SaaS and API bills indefinitely. The structure has precedent: telecommunications debt in 1999, energy high-yield in 2014. Both times, thematic concentration broke when the use case lagged the infrastructure spend.
Allocators should track two things in the next 90 days: refinancing calendars for the 2025 maturity wall in AI debt, and any shift in use-of-proceeds language from "data center expansion" to "general corporate purposes," which is the canary. If Microsoft or Alphabet starts pulling forward maturities or shifting to shorter paper, the duration bet is reversing. Separately, monitor credit spreads on the 30-year Amazon tranche against comparable industrials—if that spread compresses below 80 basis points, the market is still pricing this as safe infrastructure, not thematic concentration.
The parallel to equity concentration is not metaphorical. It is structural, funded, and now sitting in the portfolio vehicles that were supposed to be the ballast.