It's a single point of failure, sure, but you have to consider... How likely is it that AWS goes down? How likely is it that your internal system maintained by a single person goes down?