http://code.google.com/appengine/docs/quotas.html#Requests

Essentially, we use a generic 1.2 GHz Intel x86 processor as a reference when calculating an app’s usage. An app uses CPU to handle requests (in the Python layer) and also when interacting with the datastore. The actual amount used by an app to handle a given request is highly dependent on what the app actually does. Apps that do a lot of heave datastore writing and Python-layer processing will use more CPU, while apps that make extensive use of our Memcache API and write sparingly will use less CPU.

I hope this is helpful!

I have the same question. Amazon’s EC2 charges for an active cpu. Its either “on” or “off”, regardless of how much work that CPU is doing. My impression from the Google AppEngine was that the pricing would be based on cpu cycles expended through handling work/traffic. So, you could consume fractions of physical CPU’s, and thus not have to start with the $80/month that EC2 demands for a CPU. I would appreciate an explanation.