And while I agree that there's high variability in quality of work, both between and within disciplines, I would argue that there have been mountains of really fundamentally useful things that have come out of CS academia (and EE / computer engineering academia to support it), both historically and recently. Everything from programming language theory and optimizing compilers, to modern computer architecture, to all of the fundamental algorithms and data structures, to the basis for modern machine learning and distributed systems and ... I could go on. Industry applies and polishes it, but (in my experience at least) there's simply no way that most of industry would tolerate the multi-year risky projects that lead to useful breakthroughs.