Just how do you imagine they'd "CDN people's non-AMP content"? There's no mechanism by which they could tell the browser to load nytimes.com but to replace the URLs of random resources with different ones.
They'd need to host the actual page on Google.com. And after solving all the problems that doing this introduces, you've pretty much got AMP already.
Even if you can't package up and ship all of your traditional site to Google's CDN, you could do most of the burdensome/heavy bits. But then Google doesn't get to control your website and define the way it's allowed to look, which is what AMP is really for.
So it was not possible when AMP launched, is not possible now, and might or might not be possible sometime in the future after some specs are finished, but only in some browsers. Doesn't sound very practical, to be honest...
I also can't imagine the amount of shit Google would have taken if they'd started just randomly doing that kind of thing for existing web pages. Instead they introduced a totally new mechanism (i.e. AMP) where the caching was a core concept from the start.
> Even if you can't package up and ship all of your traditional site to Google's CDN, you could do most of the burdensome/heavy bits. But then Google doesn't get to control your website and define the way it's allowed to look, which is what AMP is really for.
But "heavy/burdensome bits" are exactly the things that matter the least for this use case. Ideally they would not exist at all. If they do, they should not be speculatively prefetched.
It'd also mean that these pages are now tied to Google's CDN, no matter what. Have a user click through the link from some other source than a Google search result? They'll still end up loading the resources from them. Is that really what you want?
They'd need to host the actual page on Google.com. And after solving all the problems that doing this introduces, you've pretty much got AMP already.