I don't see the point. None of them except possibly Intel themselves would have the economies of scale to put the next M1-challenging super x86_64 iteration on

I don't see the point. None of them except possibly Intel themselves would have the economies of scale to put the next M1-challenging super x86_64 iteration on a bleeding edge process node, which would be required if one actually wanted to challenge M1. Also have we forgotten that there actually is a third x86 vendor? Via / Centaur's existence and dormancy suggests that having more x86_64 licensees out floating around won't really make much of a difference.