@kate I can see a slowdown in supercomputer investments being a factor for a departure from the trend, but not a change in the exponential trend, which (IMHO) mostly depends on improvements in semiconductor fabrication technology making it possible to improve price-performance by 30x or 200x over a decade.

I mean, it's not *absurd* to imagine a progressive reduction in the economic scale of big supercomputer projects by a factor of 7x over a decade. You could imagine it happening if, for example, governments around the world systematically deprioritized nuclear weapons programs because they considered military conflict to be increasingly unlikely. I just don't think that's the world we live in since 2010. If that was the cause, I think you'd expect growth to eventually return to the previous exponential growth rate once spending stabilized at a lower level.