But with this kind of payment to interns, why not just spend the money on someone who already has knowledge and experience and will produce?
It's way cheaper to pay someone $6k or whatever for a summer to recruit them and then pay them a new-hire salary than to recruit someone "who already has knowledge and experience" and pay them the appropriate salary.
It's also more scalable and repeatable because you have an internship program that provides potential hires in batches funneled through colleges, whereas the experienced people you are talking about generally need to be found, courted and recruited individually.
Many places will set aside specific, "interesting" side projects solely for the interns (while the rest of us engineers wish we could work on these projects, but are busy with the mundane tasks of keeping the lights on, which I imagine if you gave an intern this task they wouldn't be so thrilled to go spread the word when they go back, so yeah there's definitely marketing involved.)
Interestingly, this contrasts heavily with the school-to-prison pipeline[1] increasingly occurring -- kids ending up in prison immediately after finishing their schooling.
But it's also an easy way to get rid of people who don't work out. If it's clear an intern is not performing well or not cut out, it's easier to weed them out as an intern at the end of the summer, than firing them later on.
Those people already have jobs, and they're tired of clueless tech recruiters trying to contact them. To hire them, you need to wait for their company to go out of business, or for some kind of management debacle to happen inside of it, or you need to know them personally.
It doesn't really matter whether you're hypothetically willing to pay them $120k or $240k or $480k a year if they don't answer your emails because they're already quite happy where they are.
As you say, bidding hard on interns is worthwhile for the chance to snap up a few long-term winners.
I've told many of them that they are doing themselves a disservice. I see many carry intern shenanigan baggage with them for years, limiting opportunity because they aren't taken seriously.
Ultimately it comes down to the belief that investing heavily in a Stanford/MIT/Berkeley/CMU undergrad with no other job experience is a smarter bet than finding a less-pedigreed graduate who is already in the work force.
The existence of this whole recruitment scheme - while not exactly evil in and of itself - suggest a level of credentialism that we like to pretend doesn't exist in our industry.
Given all that, it suddenly starts looking worthwhile to spend 3 months and $30k vetting promising recruits.
Having gone through the internship gauntlet myself with an AmaGooFaceSoft company in my undergrad years, though, I could count the number of interns from non-top schools on one hand, out of a total cohort of ~300 interns.
The internship mechanism would be more convincingly meritocratic if its ranks weren't almost exclusively stacked with MIT/Stanford/Berkeley/CMU/other-top-school students.
Think of it as a long-term interview and/or cheap test-hire.
Interns often produce the best work in a group because they have fresh ideas and have the time and space to focus hard on a single hard problem, while the normal members of staff have to make things happen day to day.