On the one hand I completely agree with everything you wrote.
On the other hand, it sounds like what you are proposing is that there should be socialism for sufficiently talented people (I am not actually averse to this, although it seems fairer if it could somehow be extended to everyone).
Private capital will only fund ideas which a) have quantifiable benefit which b) can be captured. This aligns well with patentable pharmaceuticals or adtech, not so well with theories of quantum gravity. Stuff which benefits everyone conventionally has to be funded by government (although a patronage model does not seem impossible).
In the ideal case, how would we work out who are the sufficiently talented (they get socialism, let's call them the eloi), and who will who make their shoes and clean their toilets (it's capitalism for them, let's call them the morlocks)? What fields would be eligible for this kind of support and who would choose that?
It sounds like your complaint is that the existing systems for doing this (grants, tenure, research impact, prestige) are mis-allocating. I agree that existing systems aren't working well. Is it just that there is not enough funding, or does there need to be a rethink somewhere?
It seems fundamentally to be an information problem. That is, it's hard to figure out who and what are worth funding. The information problem is more difficult if the research has a longer time horizon.