I have a fairly solid physics background and I understand that there is a physical limit for a reliable transistor. The problem is that it's not the same thing as the theoretical limit we currently have, nor is it the same as the physical limit for the size of a usable transistor. Our understanding of the science continues to progress and as we learn more and more about the principles we discover reasons why the the theory is wrong. Additionally, there are some very clever engineers working on the project who prove over and over again that they can improve on their current techniques, fabrication tools, and materials to push the theoretical limit. Then, where these things fail, there are even more folks standing by with ever more sophisticated error correction techniques to efficiently and consistently use transistors that are not physically reliable.
Don't get me wrong, there will be a time when the our current transistor model will stop advancing. What I'm saying is that until the point where advances have actually ground to a halt, or hell, markedly slowed down, I don't want science and tech writers cramming stories down my throat about how we'll never see another significant advance in computing power.