OK, fair enough - I had forgotten about one-time pads. I really should have specified "all encryption based on multiplying two large primes," since that's the vast majority of commercially-significant encryption. I'm not even sure if there's a theoretical quantum attack on elliptic-curve algorithms or not.
So, can this thing crack all non-quantum encryption, then? I seem to remember reading about how that would only require 32 qubits or so. And whether it can or can't, if commercial offerings have come this far, how long has the NSA had a version that can crack all encryption?