3 years is a bare minimum. They would need to staff up hardware design, verification, performance modelers, compiler writers, OS and software teams, and licens

3 years is a bare minimum. They would need to staff up hardware design, verification, performance modelers, compiler writers, OS and software teams, and license all the necessary EDA software (simulators, waveform viewers, etc). My guess would be a minimum $20 million to get it to a solid FPGA prototype in 3 years. Then if that were successful they could spent another $25 million and get it into silicon at a good process (20nm or below).