You do realize that quantum computing is only good at solving quantum problems, right? Most problems would still be most efficiently solved via a classical computer even if quantum computing was fully developed.
What you said is only under the assumption that you can't build a reasonably fast space change from binary to n-bit with all its operations. From a pure computational/mathematical standpoint, there's no competition between the two. And the issue with transitions... I'm not sure there, but I'd imagine the concept of "clocks" won't apply the same way so it's hard to gauge the concept of "speed" from how you understand a CPU today and a potential one based on Quantum principles.
And sure you'd need years to move the tooling and general expertise, but that's the case with every new technology that's a huge departure from old ones. Perhaps I'm being too optimistic, but I think it's well founded.
Just for context: I am one of those that believes you can use Quantum computers for regular tasks. Just because they're better at something more specialized, doesn't mean you can't* make them more general purpose.