The idea of additional AMP cache servers is that bing, yahoo, facebook, twitter, ... could set them up for their outbound links, so they can implement the same

The idea of additional AMP cache servers is that bing, yahoo, facebook, twitter, ... could set them up for their outbound links, so they can implement the same preloading behavior as an optimization (since they tend to have servers that are closer to users, and the connection already exists and can be reused). Think of the Google AMP cache as an extended web site snippet on the google.com search, not as your primary delivery mechanism: To make this safe and useful for all parties (publisher, user, link service providers such as google search or twitter), AMP defines an html/css/js subset that is considered safe but still functional - so that (for example) analytics can still be made to work, which is important to publishers, but without working in the AMP cache hoster's domain context, which is important for their security. Another AMP cache provider is Bing: https://blogs.bing.com/search/September-2016/bing-app-joins-the-amp-open-source-effort https://blogs.bing.com/search/September-2016/bing-app-joins-...