A reliable prediction of grinding forces and surface morphology is critically important to the design of a grinding process. However, due to the complex microstructure of a grinding wheel which contains randomly-sized and randomly-distributed abrasive grains, a practical prediction model has been unavailable. This paper aims to develop a novel stochastic model to take into account the random nature of the abrasive grain size and their distribution in a grinding wheel, and hence to predict grinding forces more accurately. In addition, the evolution of the ground surface morphology, the real-time undeformed chip thickness, the interactions between the abrasive grains and the grain-workpiece contact mechanics are integrated into the modelling. Nanoindentation and tribology experiments were conducted to determine the micromechanical properties of the workpiece and abrasive grain. Grinding experiments were then performed to validate the predictions of the model. It was found, both theoretically and experimentally, that the stochastic model thus established can reliably predict grinding forces. It was also demonstrated that the proposed model can be effectively used to reveal the underlying mechanisms of grinding processes.