I see it as another instance of the problems Intel faced with Broadwell and Ice Lake. In each case, it seems they had optimized their old microarchitecture and process so much that it was hard for a new core on a new node to overtake it. In this case, the overhang is perhaps bigger than before, especially when it's on a node they don't even control.
Where I think AMD has succeeded is in consistency. They have simply been putting in the work and delivering incremental improvements on a fairly regular cadence, and it adds up. Other than chiplets, they haven't been making bold moves like doing hybrid architectures with massive P-cores and tiny E-cores or ditching hyperthreading, but they've been very good at delivering gen-on-gen improvements. If they improved their cache-stacking to the point where the 9950X3D can now have cache on both dies, that should deliver a full generation worth of improvement, and in their flagship CPU!
How do you figure that Apple has been sitting still? They had a couple generations where their cores didn't change much, probably due to losing so much talent to Nuvia, but I thought we were again starting to see some microarchitecture improvements from them.