> I think it’s Moore’s law approaching its limits and the direction of chip improvements isn’t core speed but the number of them

Note that this is still Moore's law. You're thinking of Dennard scaling. That basically ran out 12 years ago.