The "node" or lambda is not all that important. It is a multiplication factor - you get sizes of parts in integer scales of node size. It allows for some parts

The "node" or lambda is not all that important. It is a multiplication factor - you get sizes of parts in integer scales of node size. It allows for some parts of chip parts to be smaller. Most transistors stay about the same, because you cannot make them much smaller, otherwise they will leak or will not be as fast as needed. If you look for transition of the same company from lambda1 to smaller lambda2 you will see that gate or transistor density rises not as theoretical square of lambda1/lambda2, but as about lambda1/lambda2 or less. It is also quite possible to screw things at smaller lambdas. For example, NVidia get their process right for same lambda only after several years, they were fighting steady state leaks, power issues and the like. And NVidia, just like AMD and Apple, was using TSMC. So if you are pissed about AMD lagging behind a node, don't be. Being at forefront of node size does not mean instant success, lagging a node size step behind does not mean you will get much inferior product.