Normalization doesn’t change dynamics. It just adds |(0 - the loudest point)| equally to every sample*. The dynamic stays the same.
*Sample as in sample rate, not as in sound sample.