Currently, the chips everyone's cranking out are 14 nM chips, and Intel's been developing 10 nM chips, but isn't quite there yet. http://www.computerworld.com/articl...-7nm-chip-and-leapfrogs-over-competitors.html
Holy crap! Clearly, Moore's Law is not dead yet! Still, (barring something truly revolutionary) we're coming to the end of shrinking semiconductor devices. Around 5 nm is going to be the limit, since quantum tunneling will start to interfere with device operation.
They didn't think they could break the 10 gigabyte magnetic hardrive barrier, then they figured out how to exploit electron spin states. When they push up against 5 nm, they'll find a new quantum trick, and push through.
That's going to be tough. It's hard to keep electrons from acting like electrons. But, yes, I expect progress will continue in some other way. Either will stack devices in layers or transition to some quantum or optical system or...who knows?
State is a function of a space-time curvature; nanometers represents only the spatial component but at any given point or frame of reference of space there can be more than one point of time and vice versa. Communicating between past-present-future states in a single spatial reference should therefore be theoretically possible and especially on small scales where the shift is near planck-time. So after you have maximized the amount of transistors that can physically occupy a space you can start to change its states in different instances of time and call upon those different states when needed. In reality we probably already do this anyway on a macro scale where probability or the theory of very large numbers results in a reality that we experience which is more like the aggregation of a bunch of possible states into one average result.