Plenty of reasons, but the big one would be integration, especially RAM. Apple M series processors are exclusively designed for Apple products running the Apple OS, none of them extensible. It means it can be optimized for that use case.

RAM in particular can be a big performance bottleneck, Apple M as way better bandwidth than most x86 CPUs, having well specified RAM chips soldered right next to the CPU instead of having to support DIMM modules certainty helps. AMD AI MAX chips, which also have great memory bandwidth and the most comparable to Apple M also use soldered RAM.

Maybe some details like ARM having a more efficient instruction decoder plays a part, but I don't believe it is that significant.