color me skeptical. The unit that tesla had was the nvidia px2 platform with somewhere around 20tflops of compute (very rough measure of performance). Nvidia has already released and should be in full production of their volta based Pegasus platform with around 320 teraflops of performance, or about 16x performance. In tesla’s pr stuff they said their old system could do about 200 fps and their in house design could do 2000fps (frames per second) again a very rough performance measurement because we don’t know their workload. You will notice that nvidia already has a product that is greater than 10x the performance of the px2 product that tesla is comparing to. Telsa doesn’t have any of the software stack behind it. Nvidia supports every major deep learning platform and is the go to for basically everyone doing deep learning. I don’t see this being a threat in its current form.