Interesting, if GPUS/ASICS have been improving in energy efficiency, then why is it that total consumption has been exponentially increasing?

Because so many people want to run the models.

You see this in other sectors where demand outstrips improvements in economy. Individual planes use substantially less fuel than they did 50 years ago, because there are now fewer engines on planes and the remaining engines are also more efficient; but the growth in air travel has substantially outpaced that.