The movement of animals affects the biodiversity, ecological processes, and resilience of an ecosystem. For the animals, moving has costs as well as benefits and the use of a given landscape provides insights into animal decisions and behavioral ecology. Understanding how animals use the landscape can thus clarify their effects on ecosystems and inform conservation measures aiming at preserving and restoring the ecological functions of animal dispersal. Here, we investigated the habitat preferences of African savanna elephants (Loxodonta africana) using GPS data from 155 individuals collected between 1998 and 2020 in Northern Kenya. In particular, we assessed how “energy landscapes”, i.e. the cost of locomotion due to the slope of the terrain and the animal body mass, together with elevation, vegetation productivity, water availability, and proximity to human settlements influence the habitat preferences of elephants. We found that the energy landscape is the most consistent predictor of elephants’ preferences, with individuals generally avoiding energetically costly areas and preferring highly productive habitats. We also found that other predictors such as elevation, water availability and human presence, are important in determining habitat usage, but varied greatly among elephants, with some individuals preferring habitats avoided by others. Our analysis highlights the importance of the energy landscape as a key driver of habitat preferences of elephants. Importantly, the enerscape modeling environment allowed us to develop testable hypotheses from rather coarse-grained data covering elephant movements and a few environmental parameters. Energy landscapes rely on fundamental biomechanical and physical principles and provide a mechanistic understanding of the observed preference patterns, allowing to disentangle key causal drivers of an animal’s preferences from correlational effects. This, in turn, has important implications for assessing and planning conservation and restoration measures, such as dispersal corridors, by explicitly accounting for the energy costs of moving.