The new field of stochastic thermodynamics allows us to analyze the thermodynamic behavior of dynamic systems arbitrarily far from thermal equilibrium, and has produced many powerful theorems concerning phenomena completely absent in traditional statistical physics. However, to date stochastic thermodynamics has (mostly) been applied to systems with only one or two subsystems, and a limited number of degrees of freedom. Here I present some preliminary results concerning stochastic thermodynamics of distributed systems with multiple, heterogenous subsystems. I focus on how the interaction network among the subsystems affects thermodynamic behavior of the full system. I first present results concerning Bayes nets, then concerning (loop-free) Boolean circuits, and then concerning multipartite processes, in which any of the subsystems may undergo a state transition at any time.
These results start to lay the foundation for the thermodynamic analysis of distributed computational systems, ranging from brains to concurrent processors to digital circuits.