If we split the world into 4 tiles, (level 1 of zoom) then we would need 2 bits to give a tile address (topleft, topright, bottomleft, bottomright).
Each of those tiles would be about 20,000km in size. This is obviously too big to be practical, so we would need to split it into smaller tiles. Adding 2 more bits, we have 10,000km tiles (zoom level 2) – better, but still way too big. But of course, we can do this as much as we like. By the time we get to 32 bits (zoom level 16), our tiles are about 600m in size.
An interesting way to address any point on earth with 1mm precision with a 64 bit number, plus address square parent areas with decreasing number of bits.