Machine learning engineers (i.e. people with graduate degrees, preferably PhDs) typically outearn both, though when you factor in the work experience that software engineers have accumulated during the time the machine learning people were pursuing the PhD, it's pretty comparable. It's unclear how long this will last now that everyone is pursuing grad degrees in ML/AI.

Data scientists who do ML don't have the same prestige (and pay) as those who have that ML stamp from universities.

Obviously I'm biased in the other direction (and in another country), but this is interesting to me.

I've always taken the ML stamp from universities as a visible sign that you're a newbie, if not purely because those qualifications haven't been around for a sufficiently long time, so its basically a signal when you've got one that you must be just starting out.