I have been tought that physics allow 1.5nm to be the smallest possible size of a silicone transistor, because it uses the smallest possible amount of atoms to create a working transistor. The technology process size of chip companies is often something different than size of the transistor, TSMC usually says what is the resolution at which they can make changes to the atomic structure, so they can go well below 1, we might even see 0.9nm or 0.8nm technologies in a decade.

But thw truth is that silicone technology is coming to its limit. Up until now smaller manufacturing processes were responsible for the most improvement by far, lowering the nanometers to a half meant 2x the amount of transistors, 2x the performance and half the power consumption and manufacturing price. It is the main reason AMD was able to overtake Intel back when intel was stuck at 14nm and AMD was at 11nm and then 9-10nm. Intel had better more advanced chips, but the manufacturing process was just too good for AMD and worse 10nm chip was 5-10% better than 14nm Intel, not to mention AMD started making 16 core processors whereas Intel had 10 max, which was just enough in most games (except for Total War strategies where big maps with 16 AI players could use 16 cores at 100%).