I also beg to differ. Images are pretty complicated. Just simple downscaling is obviously not that hard (unless you're like me and you worry about which colorspace you do it in, with which filter, what kind of performance for huge images, etc). But other things like lossy compression quality setting selection, automatic cropping for responsive art direction, etc etc: those things are not trivial at all.
Don't get me wrong: I'm a big fan of FOSS, and I think it's great to have in-house image management solutions based on FOSS tools like ImageMagick, mod_pagespeed and thumbor. And if your website is relatively small and the image workflows are relatively simple, that might very well be the best solution.
But I also think that for companies who just need some image management solution for their website or app, it can make a lot of sense to use an external service. It is simply a matter of dev effort and risk management. It makes more sense to have your devs focus on the core business. Building some simple in-house image management infrastructure is not that hard, but what when the next ImageTragick happens, or some new image format arrives, or browsers get new responsive image features which require more variants or server-side cleverness? Do you really want your devs to have to think about that stuff? Maybe, maybe not. If not, then moving it to an external cloud service does make sense.
Disclaimer: I work at Cloudinary; I'm also the creator of FLIF (https://github.com/FLIF-hub/FLIF).