At this point, when you are doing big AI you basically have to buy it from NVidia or rent it from Google. And Google can design their chips and engine and systems in a whole-datacenter context, centralizing some aspects that are impossible for chip vendors to centralize, so I suspect that when things get really big, Google's systems will always be more cost-efficient.
(disclosure: I am long GOOG, for this and a few other reasons)
Any computable function f on one variable x has a program. That function is a program of size p. The input x also has a data size d. BB(p+d) >= f(x), by definition, for all f and x. If you think you might have a (function, input) pair (and corresponding (program, data) pair) for which this is not true, see the previous sentence.
(disclosure: I am long GOOG, for this and a few other reasons)