Arguments in isolation are usually poor. The water usage arguments usually (always?) comes along with a bunch of other arguments, including power consumption, workers rights, consumer protection, public safety, enshittifcation, etc.

When evaluating the economical cost or morality of a thing, (just like when training a machine learning model) the more data you consider the more accurate the result (although just like statistical modelling it is worth to be wary of overfitting).