Much of nvidias marketing material covers this if you want to believe it. They at minimal claim that there will be a million fold increase in compute available specifically to ML over the next decade.