*It's going, but does Moore's Law taper off slowly, or does it suddenly crash? Also, might there be a retreat from Moore's Law, where the most advanced and generally capable chips are too expensive to find a market?
*The machinery would still get faster, but it would be purpose-built and fine-tuned for market-centric applications that will pay. General-purpose machines might be in abeyance, or even outlawed Cory-Doctorow style.
*I think one could argue that this is a final triumph of capitalism over technology here; it's not that engineers couldn't fulfill Moore's Law, more that the businesses that employ engineers couldn't afford to invest in it.
*The biggest new supercomputers are increasingly Chinese. Maybe the Chinese can afford Moore's Law – but would they sell their China-centric creations to anybody else? And if so, would other people dare to buy it?
*Also – it's not that you can't get plenty of highly capable computation nowadays. It's just – you don't own the hardware. Instead, you rent it from the likes of Amazon, from the commercial networks. The Law-As-A-Service.
*People love "deep learning" all of a sudden, and those systems can run on alarmingly small platforms. From a chip-engineering perspective there's something quite creepy and decadent about deep-learning.