There's a lot wrong with this. Availability when you're managing some servers in a rack in a Colo is a whole other kettle of fish to what you get with AWS. The key is how leverage those AWS resources in different zones and regions to avoid single points of failure. But that's so much harder outside of the cloud.
This is not intended to be silverbacking, condescending, anything like that at all: Have you ever tried to provide uptime with physical equipment?
I have been able to provide the same level of uptime over ~15 years (in various environments, ranging from corporate america to the department of energy to startups) with physical servers as I have with AWS. The initial capital expenditure outlay is higher, but its cheaper over 1-3 years than AWS.
AWS isn't the only solution out there. Its great for proof of concept, its great when you don't know how much capacity you're going to need, but it is very expensive comparatively.