Yes, sounds a lot like the "I think there is a world market for maybe five computers" prediction by IBM's president in 1943.
Practical quantum cryptography is actually not that far away IMHO as the theory is solid and there have been many successful test deployments in the field already.
Sure the equipment is specialized but not much more than e.g. the electronics that you need for a mobile baseband station or other specialized applications. You mostly need a good single-photon source and some high-quality optical components and there's no fundamental obstacle to producing those in large quantities and (eventually) for an affordable price. Also, high-quality fiber-optical cables are expensive but not prohibitively so, hence intra-city quantum links seem absolutely feasible and even inter-city connections are doable (and as you pointed out even satellite or line-of-sight communication is possible).
In my understanding the main issue that most cryptographers have with schemes like quantum key distribution is that they require classical cryptography in the bootstrapping process, so by their argument they can never be better than those classical schemes. They often forget to mention though that the adversary needs to be able to crack the classical cryptography during the key exchange process (i.e. within a few milliseconds to seconds), so the method is still much safer than any classical key exchange method except if you're facing a god-like adversary.