AMP performance numbers outside Google search

In one of the AC meetings (#85), the AMP outreach and marketing team gave a presentation on their overall plan. One thing they mentioned was the recommendation of using AMP as the default web development framework for publishers. The main benefits being developer productivity, avoidance of dual development, and performance.

In regards to performance, the AC felt that we need to be sure that the speed benefits of AMP are realized even when using the technology outside of Google search. When we remove the AMP cache and pre-rendering, are AMP pages still performant enough, say, e.g., when compared to a page built with React? To answer this question, we wanted to see some comparison numbers from TSC.

It would be great to see the performance numbers of AMP in the wild when used as the default web development framework. This can be published alongside AMP performance numbers that the team maintains for Google search.