The gap here for me is that AWS, GCP, and the like are perfectly equipped to build infra "from the edge in" as well. It's just not an advantage for Cloudflare -- I'd actually argue that AWS has the comparative advantage in building "from the edge in" given the clip at which they can turn out datacenters.
With the new Outpost model as well, I don't see too many hurdles for AWS to just start deploying their whole feature set in any old colo, in any old country, in a matter of months. They do have to set their minds to it though. At the very least, they should definitely have their storage services in every country with actual data locality laws.
I'm sure I'm missing something, but on the infra side I don't see what Cloudflare is better at in this regard. I'd also argue that AWS storage migration is just as good or better than Cloudflare's offering here. I won't comment on how users perceive/value each service because I don't use either much.
What I definitely DO agree with, is that infra "from the edge in" is a major threat to existing public cloud business models. At the point where you have your whole stack portable from colo to colo (or, gasp, as a dAPP), it starts to sound more like public cloud as a library, rather than as a centralized managed service.
Open source communities, unbundlers, and new-age cloud companies together will have a field day building 'library' versions of common cloud services and I think Cloudflare will pair nicely at that point. The question is when that becomes a reality.