Our friends at CycleComputing are offering a prize of compute time on a 30,000-core CycleCloud installation on Amazon worth US$10,000 to a single researcher with the best idea for applying such massive computational power to a question that seems unanswerable. (We hope, however, that the answer does not turn out to be 42.)
There's not much we can say here that the original blog post doesn't already say, but we're very supportive of this initiative and we want to make sure the word is spread as far as possible. It's open to anyone to enter, as long as the problem is seriously big in scale and the researcher proposing it works for a non-profit institution.
To enter, see details here.