Andy Grove gave a keynote at hot chips, I think around 1999. Back then, almost everyone had a fab. Arm wasn't a big thing, the iphone didn't exist, smartphone generally didn't exist, and many companies had fabs. ,
He basically outlined how each process shrink would allow more transistors, but also get more expensive. His conclusion was that fewer fabs would use the leading process in each generation and that costs would almost double for each generation.
It wasn't a particular opinion at the time. Generally it was thought if you were serious about making CPUs, that you would have your own fab.
For those curious about the origin, the earliest source I have on hand is https://www.gwern.net/docs/cs/2003-ross.pdf "5 Commandments [technology laws and rules of thumb]", Ross 2003, which attributes it to Moore in the 1990s who attributes it to Arthur Rock at an unspecified time:
> Sometimes Called Moore’s Second Law, because Moore first spoke of it publicly in the mid-1990s, we are calling it Rock’s Law because Moore himself attributes it to Arthur Rock, an early investor in Intel, who noted that the cost of semiconductor tools doubles every four years. By this logic, chip fabrication plants, or fabs, were supposed to cost $5 billion each by the late 1990s and $10 billion by now.
(I shamelessly call it Moore's second law to arrogate the prestige of the first law into its invocation, and because it's hard for me to remember 'Rock's law' - who's that? Plus it puts it in good company by making it an instance of Stigler's Law, which was not first described by Stigler and thus renders Stigler's law autological.)
The difference between tool price doubling every time density doubles, and tool price doubling every time density quadruples, is pretty vast. Even though those two predictions resemble each other, I don't think it's fair to call them the same thing.
"I predict that within one-hundred years computers will be twice as powerful, ten-thousand times larger and so expensive that only the five richest kings of Europe will own them."
I'm sure that's true, but I think this decision was ultimately made based on GloFo losing AMD as a big customer. GloFlo has relied for too many years on AMD's business.
He basically outlined how each process shrink would allow more transistors, but also get more expensive. His conclusion was that fewer fabs would use the leading process in each generation and that costs would almost double for each generation.
It wasn't a particular opinion at the time. Generally it was thought if you were serious about making CPUs, that you would have your own fab.
Impressive how true his predictions were.