The original observation is basically a winner-take-all combined with the fact that transistors scale as the square of the minimum dimension. So, if there is linear improvement in dimension, there is exponential improvement in density.

The reason why Moore's Law continued on for so long was that companies were willing to spend exponentially increasing amounts of money to hit the next technology node because of the winner-take-all nature of the product. Anybody who got to the next node forced everybody else to the next node or wiped them out of business.

This is going to be the same thing with fusion. Linear improvements in the fundamentals translate to exponential improvements (fourth power or better) in the outputs. The first folks to fusion are going to force everyone to fusion or wipe them out of business.