Junk Bonds Are Flashing Yellow — Why the Index Looks Calm but New AI Paper Is Pricing at 9.75%

Spread-watching by index masks what's happening at the new-issue margin. The ICE BofA US High Yield OAS has widened only modestly — 33 bps YoY to 3.15% — and the BB cohort, the bulk of the index, has moved just 15 bps wider to 1.94%. Spread-watching alone says \"yellow.\" But absolute yield is what borrowers pay and savers earn. With the 10-year at 5.28%, a BB issuer now faces an all-in cost of 7.2% on a 5-year, a B-rated issuer 8.4%, and a CCC issuer north of 17%. The market for new BB paper has been repricing aggressively even as the index hasn't moved much — CoreWeave's $1.75 billion of 9.75% notes due 2031 sold in April and now trade closer to 9.9%, and SoftBank's $11.1 billion September sale cleared at 9.75% on the seven-year tranche.

The cross-section of the US credit market tells a more textured story than the headline HY index. Investment-grade (BBB- or higher) trades at an OAS of just 82 bps — barely above the post-GFC floor. BB spreads at 1.94% sit close to their post-GFC tights. Single-B paper has widened to 3.15%, but CCC and below trades at 12.52%, with a non-performing slice (defined as spreads above 1,000 bps) at 2,818 bps per Morgan Stanley. The market is bifurcated: performing paper keeps trading on spread, while the distressed tail trades on recovery assumptions. iShares iBoxx $ High Yield Corporate Bond ETF (HYG), which holds the bulk of the index, reflects the calm — but VanEck Fallen Angel High Yield Bond ETF (ANGL) and Invesco Senior Loan ETF (BKLN) carry the cycle's first casualties.

Two decades of rating migration have reshaped the HY index. As of March 31, 2026, BB-rated bonds made up roughly 62% of the ICE BofA Global High Yield Index, up from 39% in 2007, while CCC-rated issuers shrank to 7% from roughly 15% over the same period. That composition is the case for optimism: BB issuers carry a long-term default rate below 2%, and the index today is a structurally safer portfolio than it was during the global financial crisis.

Fitch's fallen-angel watch list shows the pipeline is real. Eight issuers were added in Q1 2026, replacing three that were downgraded into HY during the quarter and another that exited via upgrade. The pace of fallen angels has not been panic-level, but the trend is one-way: more BBB- issuers landing in HY than rising back to investment grade, lengthening the lower-rated cohort every quarter.

AI-related high-yield paper did not meaningfully exist two years ago. As of October 2026, the count stands at 15 HY data-center bonds totaling roughly $39 billion outstanding — plus CoreWeave's $6.5 billion of regular-way HY. Roughly $30 billion of that has been issued since the start of 2026. AI now accounts for 24% of new high-yield supply, up from 6% a year ago. The cohort is small in absolute index terms (roughly 2.6% of the US HY index today) but the marginal-pricing impact is large because most of the bonds came to market in the last 12 months and trade at the wide end of the BB scale.

JP Morgan estimates $4.1 trillion of AI-related debt will be issued through 2030 to fund the buildout. If even a fraction of that lands in the HY index at 9%+ coupons, the BB cohort's average yield keeps drifting higher — and the index OAS hides the composition risk inside the headline number.

The 10-year Treasury yield has been the single biggest driver of absolute HY yields in 2026. It sat near 4.0% a year ago, briefly dipped to 3.8% in early 2026, then pushed above 5.3% on October 1 — the highest level since 2002 — before easing to 5.28% by October 7. The 30-year hit a fresh 24-year high in the same week. Every 100 bps of Treasury move shoves HY absolute yields 100 bps higher without changing spreads at all. That mechanic is what pushed the HY index's all-in yield from 7.22% a month ago to 8.1% today even as the OAS moved only 20 bps. Nuveen's 2026 year-end 10-year forecast is 4.95%; the gap between that and today's 5.28% frames how much further HY could compress on its way down — or how much further it could rise if the Treasury auction cycle turns tail.

The headline default rate has yet to break — and that is precisely what bulls point to. Moody's trailing 12-month US speculative-grade default rate held at 3.3% in December 2025 and 4.4% at the end of June 2026 (global figure; US-only is lower). The first-half 2026 speculative-grade default tally reached 64 — down from 68 a year earlier. S&P's June 2026 forecast has the US trailing-12-month SG default rate at 3.7%, drifting to 3.75% by June 2027. Moody's Q1 2026 outlook was more cautious, forecasting HY spreads to widen to 470 bps over the next 12 months — a near-doubling from today's 315 bps.

Distress is showing up in places the rating agencies haven't downgraded yet. The H1 2026 tally of 372 large corporate bankruptcies is the highest first-half total since 2010. Q2 alone recorded 26 large bankruptcies, the third consecutive quarter above 25. Corporate bankruptcies overall climbed to 26,919 in Q2 2026, up from 25,960 in Q1. The composition matters: the filings cluster in small and mid-caps — the kind that don't move the headline index default rate but do feed the pipeline of issuers falling from BBB into the BB cohort over the next 12 months.

State Street's Michael Arone, Federated Hermes' Joey Gallo, and the strategists quoted in the October 9 CNBC \"flashing yellow\" piece all named the same triggers. A recession is not the consensus base case — but the asymmetry is the point. HY spreads widen fastest in sharp downturns, and the index has only 33 bps of YoY widening to absorb before stress becomes obvious. The lever that moves spreads is the 10-year Treasury. The lever that creates the dollar loss is rising defaults among AI and private-credit borrowers whose 9%+ coupons are already priced in.

Days to quarters: the move from yellow to red will be telegraphed by the BB OAS, not the headline HY index. BB at 194 bps is the line to watch; a push through 225 bps would force the HY OAS above 4% on its own. The first-derivative signal is hyperscaler CDS, where Torsten Slok of Apollo has already flagged widening. The second is new-issue concessions: SoftBank's 9.75% is the new benchmark for what BB paper can demand; the next test is whether the marginal AI neocloud can clear 10%.

One to three years: the structural question is whether JP Morgan's projected $4.1 trillion of AI-related debt through 2030 lands at investment-grade yields or in the BB cohort. If hyperscaler paper trades like BBB/BB — as John Queen of Capital Group notes is already happening — then the IG/HY index boundary becomes the credit cycle's most-watched seam. Moody's forecast of 470 bps HY spreads within 12 months is the ceiling; the long-run BB average of 250–300 bps is the floor for any normalization. In between, iShares iBoxx $ High Yield Corporate Bond ETF and SPDR Bloomberg High Yield Bond ETF trade off yield, while VanEck Fallen Angel High Yield Bond ETF trades off the downgrade pipeline.