He’s not talking about recursion in C; C is imperative, recursion isn’t particularly fast in C (I assume there are many very low-level ways to make it kinda fast[ish] but it’s never going to be as fast as imperative loops et al). He’s talking about the concept in general and in particular about Lisp (this is why he mentions SICP).
Recursion’s not just a like-for-like, formally elegant replacement for imperative loops. It’s the most effective way to iterate over very common data structures, for example lists or trees. It makes writing parsers incredibly simple relative to writing them in imperative styles. It’s the only way to iterate in a large group of languages including Lisp - where the code is the data and the data is the code, in a tree, that you recurse over.
And yeah, he may well come across as a bit of a douche, but he’s not really wrong here, and he’s not wrong about a lot of other things; his blog archives are still highly regarded for good reason. You have to look at in the context of him writing from the perspective of trying to hire good programmers, and of him writing w/r/t how things were in the early 2000s.
What he says matters w/r/t small tech company success, and his hiring/project management methods and general advice are well known. He was the first kinda star blogger who could write well from a viewpoint informed by massive success in both tech and business. What he espoused (the checklist, no open-plan offices, treating employees [including interns] staggeringly well, good planning, etc) involved programmer autonomy, and for that to happen and lead to good software you generally need very skilled programmers. And if a programmer understands how things work, and what things have historically already been tried, and why, they can avoid reinventing the wheel or falling into the same traps everyone else who didn’t understand why falls into. It was a super optimistic view of how good a working environment could be, which sadly is pretty rare (to the level he evangelised for).
Plus his his track record is fairly astonishing w/r/t good, successful software projects: Excel Basic, Stack Overflow/Exchange, Trello, Glitch. He kinda demonstrated that what he talked about works in practice.