[keredson]

you're right about the detail of his point (people won't buy 100x just because a 100x increase in efficiency), but the point he was trying to make is correct. given a fixed global investment, a 100x increase in efficiency won't reduce costs. it will just increase the difficulty level to compensate.

[IanCal]

But it does have an impact on the energy use. If the energy efficiency goes up 100x and people aren't using 100x as many machines then the overall energy usage drops.

[KVFinn]

>But it does have an impact on the energy use. If the energy efficiency goes up 100x and people aren't using 100x as many machines then the overall energy usage drops.

If there really was an overnight 100x improvement there would be a lot of lag, but in the long run the cost is mostly the energy. So if efficiency slowly goes up 100x people will generally use 100x the computing power, and use the same overall energy.

There's some difference because of the ratio of silicon investment to wattage cost over time, but that's the overall shape of it.