Those of us working in lithography are familiar with the ITRS roadmaps, which largely attempt to identify technological limits to the continuation of Moore’s trend. [The physicist in me rebels at labelling this trend as “Moore’s Law”, as it is a purely empirical observation.] While the semiconductor industry acts as though its players are fearful of “falling off” the trend due to so far unrelenting competition, they have given up on faster clocking while continuing to pursue increasing gate density. This pursuit of density now leads the industry to seriously contemplate constructing three-dimensional integrated circuits. At the same time, we realize that a 3D stack presents a worse heat removal problem and even more routing congestion than a 2D layout.
But what really matters to the semiconductor industry? Computing! Apart from niche markets, smaller gate size does not drive computing applications, because these are already dominated by communications, not by switching. And the happy press releases about quantum computing, optical interconnect, etc. overlook the practical limits and lack of generality that applies to these developments.
Igor Markov recently posted a preprint addressing these and many other limits to computation. It is well worth reading.