Is it executed on the client side?
It could be. But server side execution is more likely IMO. It depends on what you are trying to do with it. (I'm not aware of any client-side implementations of SPARQL).
Is microsoft somehow going to be able to mess this up?
Anything is possible. But if you spend all your time worrying about Microsoft you won't get anything done. (If I was an IT entrepreneur, I'd be more worried about software patents and patent trolls.)
Can you use a common relational database (SQL)?
To execute SPARQL? No. As a backend for a triple store / SPARQL engine? Yes, though there are technical issues that mean that this isn't the most performant option ... for some kinds of SPARQL query.)
Are there particular resources you recommend over others for learning?
Not specifically. What are you trying to learn that you cannot learn from the tutorials?
Do you feel it is premature to begin investing myself in this?
It depends on what you have in mind. If you are doing this for fun, its not premature. If you are interested in research, it is not premature. If you are doing this to land a job, it is possibly premature (take a look at the job ads in your area). If you want to make your first million dollars ... your guess is as good as mine.
Best way to get set up and running to start tinkering?
Download and install one of the open-source triple-store / SPARQL implementations and try some experiments. My personal preference is OpenRDF / Sesame because it has clean APIs, is well engineered and just plain works.
