I just read an article that R. Stanley Williams wrote a few months ago, about his team's development of the "memristor" - a nanoscale crossbar switch that changes its resistance depending on how much current has flowed through it.
At one point, Williams describes the difficulties of building a nanoscale circuit: "Like everything else at the nanoscale, the switches and wires of a crossbar are bound to be plagued by at least some non-functional components. These components will be only a few atoms wide, and the second law of thermodynamics ensures that we will not be able to completely specify the position of every atom." (Dec. 2008 IEEE Spectrum, p. 33 column 1.)
I went and looked up the second law, just to make sure... and it says no such thing.
One formulation of the second law is: "In a system, a process that occurs will tend to increase the total entropy of the universe." So, if the circuit is considered as an isolated system - for example, once it has been constructed and sealed in an integrated circuit package - then if an atom in a circuit were to move, it would be more likely to move out of place than into place.
During fabrication, of course, the circuit is not an isolated system. The second law doesn't prohibit spending energy in the fabrication machinery in order to move atoms into precisely the desired place. That sort of thing happens all the time, in nature as well as in technology.
There is an obscure corner of the laws of entropy that says that in a sufficiently massive crystal, a certain percentage of the atoms must be out of place: a sufficiently large crystal at room temperature can't be perfect. I calculated once how large the crystal must be before the laws of entropy required it to contain a single atom out of place. I forget, now, whether it was thousands or maybe billions of tons - the point is that there's nothing in physics which requires atoms to be misplaced on the scale of integrated circuits at room temperature.
To blame the second law of thermodynamics (or any of the other physical laws) for the errors that creep into modern-day semiconductor fabrication is simply incorrect. Semiconductor fab errors are due to limitations in technology, not physics.
To blame physics for atoms out of place spreads the misconception that atomically precise devices can't be built. That is as fundamentally incorrect as the old ideas that airplanes couldn't fly faster than sound, or railroad engines couldn't go faster than fifteen miles per hour. All these ideas are equally baseless, but were used at the time as excuses for limiting engineering efforts - and projections of technological capability. It's a shame to see even an accomplished nanotechnologist spreading such ideas.