Precisely because Apple controls the entire ecosystem.
- AMD/Intel can't do anything about slower memory because they have to use DIMM
- AMD/Intel can't switch to ARM because 1. They can't just let go of x86 and 2. Windows - they can't just force an OS migration to ARM which is entrenched in the industry
- They can't build a unified memory architecture because, again, they have to use DIMM (although, AMD has something called smart memory access which is a step towards this, I guess, not entirely sure)
Even so, I'd say they have made a lot of progress in the last few years. Intel's Iris Xe is really good, even if not as good as M1. I haven't used AMD's 5500u or similar, but they also get good ratings on graphics performance. And with next generation chips moving to RDNA2, things can only get better.
Memory bandwidth will probably become less of a concern as DDR5 matures as a technology.
But even after all this, nothing can help the fact that ARM architecture is simply more efficient than x86. So we'll have I wait and watch on that front.
To be sure, the fact that Apple's products are not upgradable is a part of their performance/efficiency. You can't have that kind of efficiency unless you use an SoC instead of the modular architecture of PCs. And SoCs can't be made upgradable.