I think you’re agreeing with me. The university pays the student some absurdly small amount of money. Even ignoring the actual value added by that student, the

I think you’re agreeing with me. The university pays the student some absurdly small amount of money. Even ignoring the actual value added by that student, the university collects enough money for doing so that it comes out ahead. Note that this may not be the case for teaching assistants. One can get a teaching assistantship without being attached to any particular grant. This money actually comes from the university (but the university may still charge fees to the student). As a bit of trivia, the NSF GRFP is not a normal grant — it’s a highly prestigious grant directly to a graduate student. And it, at least historically, didn’t put up with this nonsense. GRFP gave the student a stipend and the university a smallish, fixed fee. This fee was well below the normal graduate fees. However, hosting NSF students is considered a privilege, most research groups want them (they’re free from the PI’s perspective), and as a result, the universities simply eat the “loss” and waive the balance of the fees.