Grain size is a key microphysical property of ice, controlling the rheological behavior of ice sheets and glaciers. In this study, we develop a new model for grain size evolution in ice and show that it accurately predicts grain size in laboratory experiments and in natural ice core data. The model provides a physical explanation for the power-law relationship between stress and strain rate known as the Glen law and can be used as a predictive tool for modeling ice flow in natural systems.