M3 is not M2.Oops! Apple's M3 uses the same node as these upcoming CPUs. Nor do transistors care about "overall power consumption", but only the localized temperature they themselves experience. Cerebras makes a CPU that doesn't simply consume 100 watts or even 1,000, but a staggering 23,000 watts. It works just fine ... because the temperature at any point on that (extremely large) chip is similarly constrained.
Small CPUs have more problems with local temperatures for obvious reasons. That Cerebras AI chip is over 100 times bigger than CPUs we are talking about, no comparable.