Besides being the main source of academic esteem, there is a serious structural issue where Elsevier et al receive publicly-funded knowledge, yet privately control the supply. The model seriously warps the production of research (and, arguably, the reproduction - p-hacking, fiddling with your regressions/covariants).
Using Sci-hub can be considered a form of protest. Perhaps 'break copyright law' is not a universalisable rule in the spirit of Kant, but the same is true of most forms of activism.