Just as the Newtonian friction model abstracts the energy required to push a weight on a surface, the recently proposed “Information-Friction” model captures the energy required to push information on a computational substrate. 

In this talk, we use the info-friction model to derive fundamental limits on energy consumption for noisy computing. We show under a gaussian-noise model of computing that the total (gate + info-friction) energy of computing diverges to infinity as the target error-probability is lowered. This also means that in this model, the capacity of noisy computing is zero.