~50% - I'm in "big data" right now, partly because I get to worry about the algorithmic complexity of things all the time. As the scale of a task goes up and the feasibility of just throwing more machines at it decreases, you can suddenly justify spending a lot of programmer time on interesting things.