If I may, I found the NSF GRFP process unsatisfactory. For example, I was rejected for two reasons: 1) because I did not publish prior to applying, and 2) because the reviewers thought that my research, although it had intellectual merit, did not have broader impacts (i.e., "the potential to benefit society and contribute to the achievement of specific, desired societal outcomes").
I proposed research on how to improve upon string searching, which according to the reviewers, did not benefit society (??).
The reviewers all have PhDs, I believe. But somehow, I do not feel they are qualified to review the applications. They are looking for specific things cross the check-boxes.
The second point is that stipends are taxed. So, $3k looks like a lot, but post-taxes, it's trimmed. Also, if you look at the OP's posting, $300k for ~5 years for a student. That does not mean the student receives $60k/year for living.
The third point is that someone immediately out of undergrad at 20 might find PhD / below poverty level stipends enticing, but once you make 150k+, you really don't want to hand control of your life over to your advisor's whims.
The fourth point (which is a corollary to the third) is that I know quite a few miserable PhD students, whose advisors control large aspects of their life. For example, a few of them were house-sitting their advisor's newly purchased mansion overnight in sleeping bags. These were foreign students.