As electronic devices operate increasingly close to the thermal energy scale kT, fluctuations become an intrinsic component of their dynamics rather than a perturbation. In this regime, computation must be understood as a nonequilibrium physical process.

In this talk, I will show how stochastic thermodynamics provides a consistent framework to describe nonlinear electronic circuits in the presence of noise, with a focus on subthreshold CMOS technologies. By enforcing thermodynamic consistency at the level of stochastic dynamics, this approach allows one to characterize energy exchanges, dissipation, and fluctuations in computational devices.
I will discuss how this perspective naturally leads to probabilistic computing elements that exploit thermal noise, and argue that stochastic thermodynamics offers a principled way to think about computation in energy-limited, noisy regimes.