The outer square is not given or fixed, it is the result of the arrangement inside. You pack the squares as tightly as you can and that then results in an enclosing square of some size. If someone finds a better arrangement the outer square will become smaller
To be fair, the large square can not be cleanly divided by the smaller square(s). Seems obvious to most people, but I didn’t get it at first.
In other words: The size relation of the squares makes this weird solution the most efficient (yet discovered).
The outer square is not given or fixed, it is the result of the arrangement inside. You pack the squares as tightly as you can and that then results in an enclosing square of some size. If someone finds a better arrangement the outer square will become smaller