@kate Wow, that's staggering! It's interesting that the slowdown knee happened so late; if I had to guess at a single cause, it would be the end of Dennard scaling in cutting-edge processes, but that would put the knee around 2005 rather than 2010, and that's certainly not what the data shows.