You are thinking about serverless costs all wrong

The article makes an important point: AWS (well, lambda in this case) may not be cheap but it allows you to cut cost related to paying people to do things more cheaply. Yes, if you have the resources, the time, and the in house knowledge, you might be better off running a few cheap servers in some data center. But on the other hand, you risk wasting enough time (and thus money in salaries) that it would pay for the potential savings in amazon bills for years to come.

The math is pretty brutal. If you are paying on the order of 10-15K for a single person month that can essentially take care of quite a bit of Amazon bills. We're doing around 2K/month for our setup. We used to run in Hetzner at around 300/month. AWS is definitely more expensive for us.

I've considered doing the work to reduce the bills a bit but I can't justify putting 2-3 weeks into reducing cost by say 30% or so. I would say this should be feasible in principle given that we are over provisioning and running a few things in a less than optimal way. Using less instances, utilizing them more properly, maybe eliminating a few of the routers, etc we should be able to cut down quite a bit.

An inescapable fact about Lambda pricing. It rounds up to the nearest 100ms interval. I target 10-25ms typical in my HTTP functions, which means I’m throwing 75-90% of my money into the fire if I choose Lambda.

It seems like most of the advantages listed for serverless also apply to App Engine. How do the costs compare? When would you pick one over the other?

Can you test AWS lambda properly nowadays?