This view misunderstands the nature of modern science and R&D. Interesting R&D is shockingly expensive, and you need mega-corps to do it. The cost of a new fab is $10 billion now, and it's been increasing exponentially. The 747 cost $7.5 billion in inflation-adjusted dollars to develop. The 787 cost almost $30 billion. The future of science and technology is not Edison inventing the lightbulb in his private lab. It's mega-corps doing mega-R&D.

This used to be less true in CS, but that's changing due to machine learning. Google has a huge leg-up in self-driving-car R&D because of the enormous amount of data it has from Google services. In "hard" R&D, even "startups" are spending the kind of capital that previously used to only be in the province of established conglomerates.

I agree you need significant amounts of resources / capital to do it. I disagree you need megacorps to do it. It's inefficient for the means of production to be held by a private company looking out for its own interests and wealth and not the good of society: even if you want the military to be more capable, the military-industrial complex is a bad way to do it. And the argument that startups need to spend money on their own fabs is an argument these shouldn't be in private, competitive hands.

At least the U.S. government does an absolutely terrible job with the capital-intensive infrastructure it touches. Having it be in charge of something like a nuclear reactor or a chip fab would end technological progress.

Well, the device I'm typing this on was sold by a massive, vertically integrated technology company. So I guess that's one helpful usecase.