[vidarh]

I've built high-availability systems that served millions with a <10 person team years before AWS even existed, at a time where our server racks had less combined capacity than my laptop does now, and the dual fridge sized storage array we used had less storage (and IO capacity) than the M.2 drive in my laptop does now.

The part of that solution which was related to making the system scalable was written by two of us, who also did other things (it involved a partionable backend storage service, and a user registration service, that combined to let us migrate users between servers to even out load and partition storage; everything else was stateless).

This idea that AWS is necessary to build to scale with small staff just does not match reality. My years of consulting also showed me that I'd earn more from clients who insisted on AWS - they typically spent far more time and resources on devops (and spent far more on hosting overall).

AWS is convenient, and it's great when you can afford it, but it's expensive and still requires substantial devops effort.