Predicting the mechanical performance of concrete from composition and processing remains a grand challenge. Our work focuses on predicting damage evolution in ultrahigh-performance concrete (UHPC), where analysis is challenging due to a lack of separation in the length scales of heterogeneity and a limited availability of data to describe the strength and failure of individual phases. Cement-hydration modeling informs segmentation of CT images from mechanical tests, providing a highly resolved 3-D representation of microstructure and damage evolution. Mesoscale simulations of damage in UHPC reveal how the relative strength of cement phases affects patterns of intra- and inter-granular fracture. We find that the nano-indentation hardness of individual phases is not necessarily a predictor of the relative shear strength needed to match experimentally observed fracture patterns. The mesoscale model predicts complicated material responses as emergent phenomena from relatively simple models of individual phases - a path towards forward modeling UHPC constitutive response.