Some of that (e.g. datawarehousing, etc.) is easier to outsource; other parts (data acquisition from your product, ETL design, etc.) are necessarily bespoke to your company an thus not readily "buyable." I understand OP to be arguing roughly "you can get a good DBA for much cheaper than you can get a good ML Engineer (much less a good ML Engineer who's ALSO a good DBA), so there's no sense in making Database management part of the Data Scientist role."
