The API serves nearly 1 billion calls per month, with lots of variations and spikes. That, combined with my lackluster ops skills, makes AppEngine's autoscaling a perfect fit for my use case and skill-level.

Yes, it's probably more expensive than if I used EC2, but I've never been awoken in the middle of the night to need to reboot a server, troubleshoot a memory issue, or scale up my instances.