This makes sense. Any innovation in hardware you do in research will be pissing in the wind against hardware drivers like mobile devices, gaming, cryptocurrency, “mainstream machine learning” etc. that push billions maybe trillions to those innovations. I can see why researchers want to play it safe. There are some companies looking into optics for neural nets. I guess they are betting on a small chance to be a unicorn. Sounds familiar :-). Maybe VCs can make those hardware bets for a winner takes all.