A new Goldman Sachs analysis shows that investor concerns over AI-related debt are unfolding very differently across the investment-grade (IG) and high-yield (HY) markets — even as AI-driven data-center expansion fuels a record surge in issuance.
🔍 Investment Grade: Issuer-Specific Concerns
Despite strong fundamentals in the broader credit market:
- A basket of AI-related sectors (excluding direct AI issuers) outperformed non-financial bonds by +15 bps YTD.
- But when direct AI issuers are included, the same basket underperformed by –70 bps.
👉 This signals credit investors are cautious about specific balance sheets, not the entire sector.
Neuberger’s IG credit team notes the shift creates a “haves vs have-nots” dynamic, reinforcing the importance of security selection and analyzing on-balance-sheet vs off-balance-sheet risk.
🔍 High Yield: Sector-Wide Risk
In HY markets, concerns are broader and structural:
- AI-linked HY bonds tracked sector peers for most of 2025.
- Since early November, direct AI issuers have sharply underperformed, reflecting rising skepticism about execution and leverage.
⚠️ Execution Risk Is a Major Red Flag
Some institutional investors are steering clear entirely:
- Mirabaud (CHF 30B AUM) avoided all recent IG and HY AI-linked deals over concerns about:
• untested data-center delivery timelines
• opaque contract structures
• potential need for equity-like returns to justify risk
“If the infrastructure isn’t proven yet, investors need equity-level compensation — not debt-level,” Mirabaud said.
🏦 Regulators Are Watching
The Bank of England issued another warning this week:
AI’s capital-intensive buildout, financed increasingly by debt, could pose stability risks if valuations correct.
Why This Matters
- AI demand is accelerating → but capital structures are stretching.
- IG investors are selective → HY investors are cautious across the board.
- Execution risk in data centers is becoming the critical driver of pricing and risk premia.