> clearly never make it to a real commercial product (except maybe a DoD toy), but are quite content to have Intel pay for it for literal decades.

OP posted that context?

Surely there is, especially within the constraints of a commercial company, a reasonable a) direction and b) limit to research.

Research for the sake of it is called academia.

Nominally, corporate R&D should in fact have a good chance of making it into a real product if the technology works. It should not be a certainty that the technology will work.

Intel's R&D produces a lot of technology that works but never makes it into a product.

I think this is right, it's a gamble. Discoveries and advancements are made by people tinkering with things without the mind to commercialise it.