Think of it another way - compare a modern V8 that's more powerful (more torque and horsepower) across its RPM range than a similarly displacing V8 from thirty years ago, and then note that said modern V8 is not only more efficient when putting power down, but has cylinder deactivation and is more efficient at idle. Or, can be made significantly more powerful than the older V8 peak to peak tuned, but again peak to peak, the newer V8 uses more fuel. Because it can, not because it needs to.
We're seeing massive leaps in efficiency, and we're also seeing companies be able to push their parts closer to the limit stock.
I think this is closer to the truth.
The flip side of that coin is that stock power goes up when companies hit walls on other means to generate performance increases.
If you can't engineer it in more performance as part of the architecture, or throw more cores at it, then you hope like hell you can just push more power through it and boost the clocks.
It will continue to go up until people stop buying them. For the mass market - you aren't going to see 280W 13900K's in the average desktop computer in a cubicle for a call center, or in a POS cash register, or a laptop. All the chips in that market will continue to get more efficient -- because the cost of cooling is an overhead item they can cut out as the technology does get more efficient - the performance boosts just aren't as pronounced when the technology to run it has been fast enough for more than a decade now.
That part the V8 analogy still carries through - you don't see big Hellcat-driven minivans mass produced at the car lot, you see smaller displacement and more fuel efficiency and more gears in the transmission, while staying around a hp/torque level that seems "good enough" for the size of the car. The speed limit on the highways haven't changed in a long time, after all (but Americans have gotten fatter - hence turbochargers becoming more common ---- that's a bad joke, but only because part of that is at least true).
Those big, power hungry chips really only have two markets: big workstations that do serious work (pro content creation, engineering CAD, simulations, etc) - and gamers/enthusiasts. And in those markets - they won't really care about power until they start tripping breakers. And even then, good odds on them just saying "Oh well, I'll call the electrician" and still not caring.