Something interesting to think about is as a hobbyist I'll be buying that $100K scope at a hamfest or off ebay in about 10 or 20 years for basically the cost of shipping. Then at home purely as a hobby I can entertain myself cheaply and endlessly doing stuff that was cutting edge in 2016, which is still going to be an enormous market in 20 years.
There are two problems. One is that companies like to think they're cutting edge. So gaining expert level skills on, say, late 70s high speed TTL is not interesting to hiring managers. The second problem is companies that like to think they're cutting edge, almost always are not, unless they're household names. Everyone knows google uses go and python but "the rest of the real world" runs on perl and cobol, like it or not. Oddly enough the hiring managers often don't know how far behind the times they are, so you get the equivalent of an interviewer bragging about the EE equivalent of planning to implement CVS someday soon, or trying that newfangled agile design methodology someday, not knowing their plans for the future are already obsolete in the market. Half of companies are below median, after all.
TLDR is there's a lot of posing in stem, and at least some of the people doing the loudest posing are actually doing stuff right out of 1975, if not earlier. Usually because if something makes money there's no need to fix it solely to strike a stunning pose. In public of course the image will always be that theres no way we couldda dunnit without that $500K budget, partially to intimidate competitors, partially to appear more powerful, mostly just traditional neophilia.
In terms of pure theory, I can see what you're saying but a lot of STEM curriculum includes lab work that's hard to replicate for an auto-didactic.
For applied sciences like biology, electrical engineering, applied physics (not theoretical physics), the universities will have expensive equipment that students can use in their lab assignments. Biology students examine specimens in $20k microscopes, physics students schedule time on multi-million dollar particle accelerators, and engineering students working on high speeed GHz digital circuits can use $100k Agilent scopes. Some universities also have specialized and very expensive labs with robotics and virtual reality rooms.
The humanities graduate wanting to "reinvent" himself/herself as a "electrical engineer" can buy the same college books and watch hours of youtube videos about circuits but the lab work is really hard to replicate. That's not to say that there aren't a lot of self-taught electronics engineers out there... there are ... but most I've seen learned it first as children and added to their hobby and knowledge over a decade. That's a very different learning trajectory than a 30-something trying to self-study physics without a particle accelerator in his garage.
Also, the STEM students have access to professors and their professional networks into businesses. Their recommendations opens doors for internships. This is another form of "real-world" lab work to strengthen the acquired knowledge.
Computer programming is unique in that it's a STEM that self-learners can get very deep and practical in -- without expensive lab equipment.