Making processes faster changes a lot the dynamics of how things are done. You won’t have small commits and incremental changes if your build takes hours, you won’t test in a staging environment if setting one up takes you a lot of time and effort. Work on reducing friction pays off quickly.
It’s not exactly an apples to apples comparison. Good luck with managing that database when you need higher reliability and availability. That’s where the pain starts. Also, AWS integrates more stuff, which as you grow or add to your stack compounds in terms of how useful having everything there is. Otherwise, you’ll set up your in-house Hadoop cluster with Spark on top?