I don't doubt necessarily that there will be a 24GB HBM2e 80CU card coming out soon. But the chances of it being for consumers appears to be virtually nil.
Why on earth would you put 24GB of extremely costly memory on a consumer card? You wouldn't.
This has to be Arcturus, the next evolution of GCN (whether it'll be called GCN we don't know).
Navi / RDNA have demonstrated far less memory bandwidth constrained performance than any of the GCN cards, particularly the latter ones.
Unless RDNA2 & the shift to 7nmEUV means they've extracted enormous performance uplift over Navi (read double or more per CU), there would be absolutely no rhyme or reason for doing this. IMO it could easily go up 50%, as Navi / RDNA1 was the pipe cleaner, and 7nmEUV should bring some significantly higher clocks. But double? Come on.
Unless their RT implementation needs masses of memory bandwidth .. but again, that seems pretty unlikely.