That's the story they sell to investors, but real Amazon developers know much of their internal infrastructure is actually using some old systems and not AWS. The whole reusing internal interfaces thing isn't what they're doing at all. Really what Amazon's AWS division is good at is 1) hiring and organizing engineers and 2) selling their services to companies, sometimes very aggressively.
When I worked at an internal Amazon team, we made heavy use of plenty of AWS services: MySQL on RDS, S3, and SQS for example. What we didn't use was EC2 et al. There was an internal equivalent for those.
A number of AWS services are modeled after the Amazon internal versions, but they are built from scratch and generally less rough around the edges. Then internal services migrate to the AWS versions over time. The migrations sometimes happen very slowly because teams are generally not forced to do it.
