Uhh, precedent. Double the resources, double the ability. This is well known.
That is magical thinking. It has no place in proper engineering practice.
There are zillions of reasons that interfere with ability to work faster in larger problems - yet they can be summarized with the words "non-linear growth". Try solving the Travelling salespeople problem twice as big with merely twice as fast hardware, it will slow to a grinch.
It's not like AI is going to run on some unknown substrate.
We know the substrate of brain power, gray cells. This doesn't mean that we understand the way they work together to create intelligence. If we ever create an AI, it will be so complex that we'll likely be in the same situation with respect to how it works.