AI borrowing boom splits credit market into haves and have-nots
The AI borrowing boom is reshaping credit markets, and it is creating two tiers of borrowers. Hyperscalers and data centers with permits, power and steel can still fund themselves, while new clouds and highly levered AI businesses face a closed bond market. Priya Misra of JPMorgan Asset Management sees Treasuries as the hedge because much of the economy now depends on the AI trade, and she expects capital expenditure to peak this year or next while supply keeps coming.
Takeaways
- High yield has split into two markets. Triple-C debt has widened sharply and JPMorgan Asset Management is cautious because a higher cost of capital squeezes the most levered companies. Double-B and single-B issuers have little AI exposure and widened only 20 basis points, making them remarkably resilient.
- AI borrowers raised $70bn in a week. Oracle, Meta, Alphabet, Microsoft and Apple issued roughly $70bn of bonds in seven days. New issue concessions ran at 15 basis points against a normal three to four, and the deals were three to four times oversubscribed.
- Credit market splits into haves and have-nots. Hyperscalers are fine. Data centers that already have permits, power and steel have become more valuable because new ones are harder to build, while the bond market is closed to new clouds and highly levered AI businesses.
- AI is 15% of investment-grade market. AI debt makes up about 15% of the total investment-grade market, so the concentration is not extreme. Misra still treats it as a risk factor that portfolio managers must measure and manage.
- Bond inflows strong on 6-7% yields. Investors can earn 6% to 7% in high quality fixed income while the equity market moves 15% in a day in certain stocks. Misra sees bonds as a safe haven while the AI trade consolidates or unravels.
- Capex is peaking this year or next. Misra expects heavy supply because hyperscalers have talked about their capital expenditure needs, and she sees capex peaking this year or next. Funding will not come only from bonds, so she expects equity issuance too.
The playbook for AI risk in a bond portfolio
- The portfolio manager is the first line of defence, looking at how much of the AI risk factor sits in the portfolio.
- Misra treats a reduction in risk as sometimes the best hedge when a portfolio holds too much AI exposure.
- The AI technology works, but the economics can be questioned.
- Because much of the economy depends on the AI trade, owning Treasuries as a hedge in case the economy slows makes sense.