In a way, it is the opposite. Code is what you execute. The transistors are the engines that do the execution. They are going to expect a chip they designed wit

In a way, it is the opposite. Code is what you execute. The transistors are the engines that do the execution. They are going to expect a chip they designed with 105B transistors to perform (speed/efficiency/whatever) in the same ballpark as a high=end GPU for their AI workloads. It is like knowing the kind of engine a car has. Not all V8 gas engines produce the same power, but knowing that it is a V8 instead of an inline three cylinder does give you an idea of the expected performance characteristics.