Earlier quoted context omitted.
There is hardware, it isn't imaginary. It just isn't very robust or extensive yet.
There is hardware which purportedly does a quantum computation, but it's completely useless even as a proof of concept, unless you have a burning need to randomly sample from the distribution it happens to generate.
> but it's completely useless even as a proof of concept
We might have different definitions of "proof of concept", but I think that's exactly what it was (if you're referring to Google's experiment).
This FAQ explains this quite nicely: