India has the world’s largest number of school-aged children. The majority live in rural areas, many of which are highly flood-prone. Previous studies document that in such areas, floods are associated with lower enrollments, attendance, and learning, in some cases with differentiation by gender, caste/religion, and family SES. Previous literature suggests that components of community infrastructure have positive associations with children’s learning. However, previous literature has not addressed whether better community physical and social infrastructures are associated with (1) smaller flood-related learning losses on average, (2) different learning for marginalized versus other children in the absence of floods, and (3) different vulnerabilities to floods for marginalized versus other children. This paper finds that (1) most aspects of community physical and social infrastructure are not associated with lower flood-related learning losses on average, but proximity to towns and several components of social infrastructure are associated with lower flood-related learning losses on average, (2) community physical and social infrastructure components have heterogeneous associations, in some cases increasing, in most cases not affecting, and in other cases reducing disparities in learning between marginalized and other children in the absence of floods, and (3) community physical and social infrastructure components have heterogeneous effects, in some cases increasing, in most cases not affecting, and in other cases reducing disparities in learning between marginalized and other children in the presence of floods.