Moore's Law will eventually catch up. The Pebble is currently the PalmPilot of smartwatches; perfectly suited to the current state of the art. However, progress being what it is, eventually the technology will be good enough that Android (or iOS) on a smartwatch will become the better solution.
Go back in time 15 years, then build a computer as powerful as a modern smartphone. It wouldn't be easy, but it could be done. Now try to power that computer with a modern smartphone battery. Not gonna happen.
Everyone is interested in maintaining the growth in compute devices diversity and capabilities that Moore's Law has enabled in the past decade. To do that we need to continue to pack more and more transistors on chips.
As you pointed out, todays problem is not the transistor size, but rather it is the energy consumption. Both dynamic and static power are not decreasing proportionally with node size, and therefore we have gotten to the point of asking "why add them if we can't use them?". This is the problems of the power wall and the observation that we must continue to turn off more and more of our silicon if we can't scale dynamic and static power characteristics of transistors with their size. This problem threatens the well being of all semi conductor companies.
Do you think that Intel will just surrender and go bankrupt? If they don't defeat this problem, soon companies will have no reason to buy chips from Intel. So what if they have 2x the cores on the chip if you can only keep half of them active? Without breakthroughs in energy storage and heat dissipation, how can google/apple/... create exciting new compute devices if the processors continue to sip as much power as they did two years ago? Do we decrease the compute capabilities (i.e. give up on the additional transistors) or run them slower than we are used to and is that acceptable?
There is a lot of interest in continuing Moores law. Todays problem is power, like you pointed. Tomorrows problem will be different. With 100 billion dollar livelihoods on the line, I have no doubt they will be solved in the coming years!
Software is another part of the equation - how can the overall efficiency be better if the increase in energy efficiency is being counteracted by increasingly less efficient software using more computations to perform a given task? Exponential increases in hardware have not translated to anywhere near the same gains for overall productivity in average use cases.
> As you pointed out, todays problem is not the transistor size, but rather it is the energy consumption.
If >10 GHz processors were available, that's where the excitement would be. Instead, because we've hit such a wall in raw compute power, the bleeding edge is in things like multiprocessor architectures, power consumption, and sensor integration.
