Too many tasks: top academic scientists are administrators

Here's an interesting piece of evidence to support the article from the field of CS. To a lot of graduates who aren't interested (primarily) in teaching, but rather research, research labs like MSR, IBM, and even recently Google Research (they do research in a small subset of topics, to my knowledge directly related to ads, auctions, and data analysis) are looking extremely attractive precisely for the reason that all the overheads are taken care of. This leaves them to do what they enjoy doing.

The only downsides are the absence of graduate students, who bring in fresh ideas, and can be a board to bounce your ideas off of, and occasionally teaching advanced courses which helps you consolidate your research and understanding better. These are partly alleviated by the presence of interns and guests, and the various adjunct faculty positions offered in universities.