We might have already crossed the line where the cost per transistor goes up when transistors get smaller, which may put an end to transistor shrinkage before it becomes impossible to make smaller transistor. Expensive 40-series CPUs from NVIDIA may be the harbinger of this trend.

The industry has directions to go. Other semiconductors such as Indium Phosphide can run a lot faster than Silicon, looking at SBIRs that have been funded it seems possible that 50 GHz+ processors have been made. The other one is devices that are stacked 3-dimensionally and have a huge number of parts but the parts might run a lot slower than current processors to control heat production.

Apple's M2 chips represent meaningful progress that is poorly reflected in Moore's Law. Similar to ICE engines, it is likely that the focus will become a mix of efficiency and power rather than purely focus on power moving forward. And like engines, incremental gains in horsepower per unit of energy are still useful.

yes, agreed

You brought it up