There can be 64 core Apple CPU, but that will be a very different beast, not an M1. An M1 is a SOC, a GPU, CPU (both firestorm and icestorm cores), neural accelerator, video accelerator, whatever on the same chip (the memory is NOT, that is a separated chiplet). All-in-one.
That's 16B transistors, which is more-or-less OK with the 5nm TSMC (I calculated something like 85-90% yield). If you just take one single Firestorm core (that's the "performance core"), it needs about 0.5B transistors with its cache, i.e. 64 such cores would be 32B transistors alone. No memory controller, no PCIe controller, no GPU, nothing.
This means that Apple will make a machine with discrete gpu/accelerator/cpu, with much higher power consumption, different timings, who knows what performance (I still expect very good, but don't know). The big CPU would result a 75% yield at most with current 5nm (who knows how much with 3nm, if that will be available...), and they won't produce too many, since they will produce mostly laptop chips. In the mean time, AMD will have about 98% yield per chiplet (they need 8 of them for 64 cores), while producing the same chiplet for each of their CPUs (high volume).
Based on this, I don't think that there will be more than 32 cores in the next 2 generations, mostly since their users do not need more. These are not servers, they rather need few very (say not more than 20) strong cores and some really strong GPU.