Are pure mathematicians at U.S. universities expected to win research grants?
At my large R1 state university in the U.S., it might be impossible to get tenure without having an active grant at the time, even with an otherwise excellent record, since mathematics faculty disinterested (or hostile to whatever the specialty or person in question might be) would use this as grounds for doubting the quality. After all, supposedly, all the best people are funded. Supposedly, endorsement by the NSF (or NSA) is an external, objective test of quality. This is certainly a convenient assumption for arguments in certain directions. EDIT: and since a strong super-majority is needed for a "positive" tenure vote, even irrational ranting can sway otherwise uninformed or disinterested people to scuttle the vote.
Issues of external funding are apparently even more identified with research activity by engineers (and experimental scientists), who often dominate the college-wide tenure (and promotion) committees. That is, it is sometimes apparently unimaginable that a person would be doing research without a grant.
Further, at my university, in other hard-science departments, often large grants are allowed to "buy out" faculty from any teaching at all. This is rare in my math department. EDIT: thus, in the minds of some engineers, "teaching" is stigmatized, since in their own depts it's only the disenfranchised who have to do any of it. And then there're the wildly different paper-count standards for experimental science/engineering versus mathematics.
The literal "funding of research" usually is less critical, apart from the expense of travelling to conferences, and extra summer salary.
The expectation was clearly there in all three places I've been so far though the extent to which it was crucial varied a lot. From my perspective, it looks like the external funding is most crucial at the top tier places (where most people around just happen to have it, so why not you?) and at the bottom tier places (where the administration just doesn't understand any measurement units but dollars, so to sell somebody without a grant to the dean may be quite a headache) but not so crucial in the middle.
The good part about pure math. is that the grant support is not essential for performing the routine duties (you do not need to finance a lab, or anything else like that), and getting tenure is just a once in the lifetime event, so you do not need to be in a constant jeopardy as far as getting external funding is concerned. The bad part is that it looks like on top of general budget cuts, the NSF is currently also trying to give big chunks of money to few people instead of supporting more researchers (especially the young ones) with smaller amounts, so getting an NSF grant in pure math. gets harder and harder.