[weitendorf]

Regarding cost, it can depend a lot on your traffic structure. If traffic is bursty, or substantially different between intraday peaks and troughs, it can be more cost effective. Solving this yourself costs dev time.

>reduced time spent in devops and increased time spent in development

This may be true if you’re trying to figure out how to do something you already know how to do outside of serverless, but IME many developers benefit from serverless eliminating boilerplate and nudging them away from state where it’s not necessary.

Also regarding cost and devops: I see serverless as an insurance policy against “my startup/product got a once-in-a-lifetime lucky break by going viral while I was asleep” causing you to fall over from a flood of traffic. Not only do you get to skip implementing your own scaling to go from 0-1 but you only pay substantially (vs the regular price diff) extra for it when it happens.