I'm not sure that's the best guess. While it would be the obvious choice, I don't think much could be said to be "earthshaking" there, and Knuth isn't one to throw out hyperboles. I think it will be something new, almost certainly in a different field, and it will involve looking at an unsolved problem in a way different than it's previously been approached from, much like TeX. I can't wait to see what he announces.