AI Starts Paying A High-Yield Premium In US Credit

As junk-bond and loan issuance rises, investors are pushing riskier AI borrowers to show real cash flow and stronger collateral.

What’s going on here?

AI infrastructure is spilling into the riskiest parts of US credit markets, and junk-bond and leveraged-loan investors are demanding higher yields and clearer proof that new borrowers can generate cash.

What does this mean?

Reuters reports that low-rated, AI-related debt issuance has reached $88 billion this year, citing Goldman Sachs, even as higher Treasury yields lift borrowing costs across the market. Investors are still funding the theme, but they’re concentrating on higher-quality borrowers and on deals with predictable revenue, long-term contracts, and collateral they can sell if things go wrong.

That puts data-center financings front and center, because many projects have heavy upfront spending and cash flow that arrives later. BNP Paribas, a European bank, says AI infrastructure has made up $40 billion of high-yield supply so far this year, versus $12 billion for all of 2025.

Pricing shows the skepticism. Analysts told Reuters that even BB+ borrowers are paying around 9%-10% yields, while weaker issuers can face 14%-15%. SoftBank Group, a Japanese tech investment firm, recently paid as much as 9.75% on 7.5-year bonds despite a BB+ rating, levels some analysts said look more like lower-rated debt. Credit investors don’t get much upside if an AI bet pays off, so they care most about near-term cash flow and execution risk.

One swing factor is demand from collateralized loan obligations (CLOs), which buy corporate loans and fund them by selling their own bond-like slices. Morningstar’s Elizabeth Templeton says those managers are getting more cautious on certain AI-linked names.

Why should I care?

For markets: Zenith Arc’s 8.875% coupon still saw the bonds drop to 96.75.

A weak trading debut is a real-time check on whether there’s a stable buyer base for AI-linked leveraged debt once the deal is done. Zenith Arc’s $2.25 billion five-year senior secured notes for an Oklahoma data center leased to trading firm Jane Street priced at 99.50 with an 8.875% coupon, then fell to 96.75 soon after, according to Pender Fund Management.

If big repeat buyers like CLO managers step back, the marginal buyer sets a tougher clearing price. That usually means higher interest costs and/or tighter lender protections like more collateral. So financing can still be available for lower-rated AI buildouts, but it comes with a bigger cash-flow-and-collateral premium, which can slow projects or reshape who can raise money.

—

Originally Posted September 30, 2026 – AI Starts Paying A High-Yield Premium In US Credit