Someone needs to show that you can actually make money with something like this. The problem with these sorts of ventures is that besides being capital-intensive and risky, it can be quite hard to capture the economic benefits of the development in a way that justifies the investment.

There's also the issue of whether Silicon Valley has the requisite infrastructure to facilitate these sorts of projects. If you want to do projects in robotics, DARPA has both money to give you, deep connections with the universities developing the basic science, and a program manager to manage the project that typically has a PhD in the field. There's a lot of downsides to DARPA, namely that everything has to have some sort of military angle, but there's a wealth of technical expertise sloshing around there.