The crux of the problem is the napkin math was wrong because he obviously forgot to a) include his billable time on learning such stuff and b) include the price of further staff to manage such things.
Personally I think there are tons of devs who can be good at ops, you just need to quickly come to the realisation that many of the aspects of ops are fairly essential non-optional functions of your job role, and should sit behind the writing of new code if your company has anything of value.
Once you come to these realisations you can quickly understand that despite a dedicated server being cheaper than paying for an email service, the time sink and required technical knowledge will quickly more than even the cost out, leading to this actually not being about ops for any other reason than an ops guy wrote it.