Comment by dist-epoch

Why doesn't NVIDIA also build something like Google TPU, a systolic array processor? Less programmable, but more throughput/power efficiency?

It seems there is a huge market for inference.