I wanted to comment on this, since I feel I'm in an unusually good position to understand it.
First, the claims are quite real, and quite attainable. This setup really can ensure that nobody can eavesdrop on your conversation without your knowlege, at least in the middle of transmission. Of course, as soon as you detect anyone eavesdropping, you stop transmitting, leaving the interceptor with only a few bytes (ideally) of data. Combined with conventional encryption, that data is virtually worthless.
There are some theoretical attacks that may enable somebody to intercept and fake some percentage of the photons en route; but the problem is, if the interceptor can't fake 100% of the photons (or some obscenely large percentage), the reciever and sender can compensate with their protocol and still detect the interception with extremely high probability, extremely quickly. It is exceedingly unlikely that it is even possible to forge 100% of the photons correctly; it may even be provable that this is the case.
Upshot: This isn't rocket cars and jet packs; it's quite feasible. (And IMHO the most significant thing we'll see from "quantum computing".)
Of course, you still must worry about the integrity of the endpoints, and all of the relays, but again, since this will inevitably be layered on top of conventional private-key encryption, this is no worse then things already are today.
The upshot is that this is a nice step forward, essentially eliminating entirely one form of attack. Of course, there are still others; corrupt a relay station and steal a key via some other means, and you can still intercept messages, but for some situations (like, say, the military), it's quite feasible to secure the devices and stop worrying about the lines being tapped.
Will consumers see this? That remains to be seen. I don't see why the equipment itself should be that expensive, but it may be difficult to use, as it must be line-of-sight, or go through a series of relays. We'll have to see how it plays out.