Devs don't want to manage VMs. Patching, gold loads, SSH configs, k8s, so on. You write code, you deploy, and you pay only when the function executes. It scales as much as you want it to, as much as it needs to, and quicker even than EC2.
There are very clear benefits to FaaS. It is certainly not for everyone, esp. if you already have a well functioning deployment pipeline.
Which devs?
I'm a developer and I enjoy most of these things. Working with k8s is a pleasure, managing VMs can be quite simple (and more often than not, is). I don't do much patching, but I "SSH" into our k8s pods quite frequently. I've never thought, gee I wish I had a lambda to simplify this. I know I _also_ don't represent all devs, but I'm never seen this comment made outside of it being an argument supporting FAAS.
If you get minimal returns for managing 5 extra services, it doesn't make sense most of the time.
I do agree though that the cost should not be the thing the poster brags about with this architecture. It's the tradeoff between simplicity, cost, and functionality.
I work in the finance industry. Most devs don't seem to know or care about IAM, proper network security, or keeping VMs and container images up to date.
IAM in AWS is by no means "easy" but when you get it right, it's pretty damned good.
To me the main beneficiaries of all FaaS offerings are actually the service providers themselves, as they can make better use of their existing capacity and at the same time charge a hefty premium for the service. I'm yet to see a single convincing usage of AWS Lambdas that makes economical and technical sense.
It's also filled with "once you've done X Y is simple!" like:
> Once you have the template ready, you can simply deploy your code by running: serverless deploy
which is true for all automatable tasks. The same can be said for any VM or shared hosting.