I have to wonder if there's not more that can be done on this front for low cost IC prototyping. I don't think the fixed infrastructure is necessarily the problem (i.e. building the fab) as there's enough capacity for cheap chips in volume, meaning each additional wafer isn't the cost limiting factor. There are multi-project wafers (like PCB aggregators), but my understanding is that the hard cost limit currently is the NRE of making the mask set, which isn't getting amortized over a sufficient number of devices in a prototype run.
So cheap masks (or fewer masks) would be an area I'd be interested to see development.