[bkcooper]

The thresholds needed to achieve Quantum Error Correction are extremely close to being achieved nowadays

Yes, they are quite close to surface code thresholds. However, being close to the thresholds means that you need to be on the pessimistic side of how many physical qubits you need per logical qubit. As I mentioned in my other comment, even Martinis (whose group is, as you say, very good) is still several orders of magnitude short in terms of the number of physical qubits needed to implement one logical qubit, let alone the hundreds to thousands of logical qubits necessary to do interesting computations.

Packing more qubits in seems to me like it's going to be a very challenging problem. Looking at Martinis's recent Nature paper, 9 qubits are taking up something like 2 x 4 mm. Making these smaller would be nontrivial for many reasons: they each have their own microwave coupling line (which requires a certain amount of length); the coherence properties of superconducting qubits seem to care about how much surface you have relative to bulk metal, which is a loser for shrinking the devices; presumably jamming them together presents crosstalk issues; etc. Realistically, you also need to add a bunch of other types of electronics down there to handle multiplexing a la D-Wave. I don't know that these obstacles are insuperable, but I'm definitely a little skeptical that this road leads to useful technology.