Being early on the ML hardware acceleration boat is going to pay off astronomically. Embedded inferencing is going to be a society defining technology by the time we hit mid-decade. It's already being used for predictive maintenance of machinery in IIoT with huge payoffs via decrease in unforeseen total machine failure or need of heavy overhauls.
It really blows my mind how many people are still bearish on ML. It’s fair to argue timelines (although even that is becoming less true), but I think the evidence is firmly on the side of the bulls now.
And I believe AMD are good with using calculators.