I think what you wrote makes sense, and you missed one critical aspect : shared access to excessively expensive capitalized facilities and equipment.

One example from 1985 onwards that i can think of is NSF funding of supercomputer centers. 40 years ago, SIMD / vector processors with boatloads of memory were not ubiquitous, nor were shared memory multicore / multiprocessors, a situation which differs with the reality today.

This NSF funding established the 5 supercomputing centers

https://en.wikipedia.org/wiki/National_Science_Foundation_Ne...

and then further downstream effects include popular access to creations from the supercomputer centers, such as Mosaic from NCSA, and an expansion of ideas outside the compuserve / aol paradigms.

I think similar situations apply for other engineering disciplines, mechanical and chemical and physics and so on. Probably true for the arts in various forms: people don't have personal pipe organs to learn Bach on, for a crazy example, but universities do.

For various industries, learning requires physical equipment too expensive for individuals, historically and still.