It doesn't matter if you fetch 300 cycles in the future, you'll eventually saturate cache bandwidth. It's worth noting that the M1 has a reorder buffer of more

It doesn't matter if you fetch 300 cycles in the future, you'll eventually saturate cache bandwidth. It's worth noting that the M1 has a reorder buffer of more 600 micro-ops, which essentially means it can be 600 instructions in the "future" (it's only the future from the point of view of instruction commit).