[archgoon]

You can break a 160 bit EC with a 1000 qubit quantum machine. An equivalent 1024 RSA bit key needs about 2000 qbuits. (The following paper has a cool table and formula for working out how many you need for each)

https://arxiv.org/pdf/quant-ph/0301141.pdf

My incredulity is in the idea that Google and the NSA apparently have 1000 qubit machines that can do that, but not 2000 qubit ones (and you should be very confident that will remain the case in future).

Note that once you're at the "I have a stable 1000 qubit machine that can perform complex calculations" realm, you can consider doing things like entangling it with that other 1000 qubit machine you have lying around, bringing yourself rapidly into the 2000 qubit realm. Getting to the first part is really hard. If you can pull that off, you're probably only a few years away from the second. The main thing that's stopping us from doing that today is that the (physcial) qubits we've made so far aren't stable enough; getting to a single logical qubit is anticipated to take 1000 physical ones.

https://en.wikipedia.org/wiki/Physical_and_logical_qubits

[aborsy]

The few thousands qubits that you mentioned is logical qubits. With error correction you need tens of millions of physical qubits to break RSA or ECC.

Google and NSA can produce more of the same thing. But they probably have no major secret sauce for things such as error correction that the public does not have.