A rather shallow analysis. He first asserts (he doesn't explain) how fundamental limits to further miniaturization of silicon devices are based in heat (the chip would melt itself) and leakage (due to quantum principles electrons won't stay localized in such small features - "can't say anymore if an electron is in the chip or outside the chip").
Then he moves on to mention molecular circuitry whereby current flows through a molecule instead of only through metals and silicon - but he never explains why the above two fundamental problems wouldn't apply to molecular circuits. It's not like organic molecules are more thermally stable than metals and silicon, or that they are bigger than patterned silicon features.
What a waste of time.
there is some research that shows that carbon nanotubes do not generate heat in or on the actual tube, but rather the area around it. I do not know how much that would affect the whole thing (there is still the problem of where does the heat go, what would stop it from going to the tube?). and mores law only states that the amount of transistors will double every 18 months. this has little to do with the "speed" of computation of complex problems.
You could likely argue that the current main stream system (x86) is old and inefficient. current mobile processors like the arm architecture in a way prove this. although i have no idea how well they scale up to high frequencies.
but really, we do not know how this will all pan out until they actually start to make some chips that do useful calculations reliably.