AI Debt Floods The Bond Market, And Fidelity Is Backing Away

This is less a story about default risk than about duration and spread compression breaking the economics of “safe” credit. When top-rated issuers flood the market with long-dated paper, the marginal buyer is effectively being paid Treasury-like returns to take equity-like headline risk without any meaningful cushion. That creates a brittle setup: the paper can stay orderly until it can’t, then reprice quickly because the market is already near zero-vol premium.

The second-order effect is that hyperscaler capex is now competing directly with every other high-grade borrower for balance-sheet oxygen. As these issuers term out more funding, they can crowd out industrials, utilities, and even quasi-sovereign names from the safest end of the credit curve, forcing a wider dispersion in financing costs across IG. That likely benefits asset-backed and structured credit with granular collateral, while punishing long-duration single-name IG and any portfolio benchmarked to generic corporate indices.

The real catalyst is not a default headline; it is either a ratings outlook change, or a further backup in long Treasurys that exposes how little spread investors were clipping relative to duration risk. The time horizon here is months, not days: supply has to be absorbed through year-end issuance windows, and the market can tolerate a lot until a soft macro print or a capex guide-up forces another leg of borrowing. If rates stabilize or fall sharply, the trade pauses; if capex keeps outrunning free cash flow, the downgrade/sticky-duration trade gets more dangerous.

Consensus is still treating AI capex as a pure equity winner and a high-grade bond non-event. That misses that the bond market is the first place where the financing model is being stress-tested, and the signal will likely show up first in wider CDS, weaker new-issue concessions, and underperformance of the longest maturities. The overdone part may be the fear of near-term default, but the underdone part is the risk that these names transition from “too safe to matter” to “too big to ignore” for credit indices.