>Tell me, how well did that work for Grin?

Crypto projects succeed/fail for all kinds of reasons that are completely unrelated to de-/centralization. You'll have to be more specific about what Grin's case should teach us.

>So then, _centralize_ around an ASIC?

ASICs are commodities. For BTC (SHA-256) there are at least 8 different companies producing ASICS, and even a smaller project like KAS (kHeavyHash) has >4 competing companies. Not much centralization risk on that side, at least not for mature projects (which a hypothetical ASIC-XMR would be by now).

The main challenge for ASIC-miners is the same as for CPU- and GPU-miners: cheap electricity -- and that's not something that can easily be centralized.