Molecular simulation has reached a high degree of versatility following the advance of massively-parallel high-performance computing. On the basis of physically realistic models of the intermolecular interactions, thermodynamic properties can be described and predicted with a high accuracy. Transport processes and heterogeneous systems can be reliably simulated as well. This requires robust molecular models and simulation methods with a rigorous foundation on statistical mechanics. The state of the art is commented here with a focus on applications in process engineering.