I don't have a very clear understanding of what it is these cloud providers are offering but I'm pretty sure that as of today no one is close to having a quantum computer that can perform a commercially useful task faster than a classical computer.
I would never discourage anyone from learning more about a subject that interests them but I suspect the "with the hope that eventually this leads to work" part is unlikely in any reasonable time frame and quite possibly forever. The main reason I say that is that as far as I can see there's little reason to expect QC to ever have the kind of general utility that today seems to provide virtually endless amounts of work to millions of programmers. After decades of theoretical research there's really only one algorithm that is believed to provide super-polynomial speedup over classical algorithms. That is Shor's algorithm for factoring integers, which is isn't even a problem that would have much practical interest if it weren't for the historical accident that most current online cryptography ends up being based on it.
The other kind of problem that a QC might be useful for is simulating quantum systems in physics and chemistry but my guess is that most of the people who will be doing that will be physicists who have learned to program rather than programmers who have learned quantum mechanics.