Let's just say they solve particular case(s) of problems via quantum annealing techniques and be done with it ... http://en.wikipedia.org/wiki/Quantum_annealing ... there is no doubt in the future going to be other ways to solve other problems but in the meantime the technique works.
About as much as any computer with no software. It's more a push into a new technology for the sake of it. Applications will come, or they won't, we will have to see.
They are expensive as any computer fabricated at low volume, they aren't faster than current clusters. It may still be a good specialized architecture, but didn't prove itself on any problem.