Definitely a subfield thing in part. Highly theoretical CS PhDs (complexity theory, algorithms, etc - the folks who are mathematicians in computer scientist's clothing) almost always require postdocs, both because they tend to be less heavily recruited by industry, and because the vast majority of departments are shying away from theory in favor of things that bring in more grant money and industry partnerships.

Departments aren't stupid, though. They realize that grabbing star grad students with highly marketable skills in industry as they are walking out of their dissertation defense is probably the only shot they'll have at getting them. They will often "try out" industry or take a job at Google or MSR on the advice of their PI, noting that it's just as good as a postdoc but pays four or five times as much. Problem is, they then have to walk away from their $180k/year to take a faculty position that might not pay half that and could very well dead-end in a few years if the department leadership changes (or any of a dozen other precipitating factors) and they don't get tenure.

All that said, I personally know of tenure-track professors at two of the schools you mentioned, plus ones at quite a few other top-flight CS departments (Cornell, Toronto, Washington, Cal), who were hired straight out of grad school.

I think you have a bias of looking at students from select top schools. A PhD student in almost any branch is going to have a really hard time getting a tenure track position unless they have an awesome research record (publications in top tier conferences/journals).

both can be true. super rough approximation for CS faculty job candidate consideration: Ph.D. outside of the U.S. followed by a postdoc in U.S. at School X ~ Ph.D. in the U.S. at School X