Abstract The seasonal cycle contributes substantially to soil moisture temporal variability in many parts of the world, with important implications for seasonal forecasting relevant to agriculture and the health of humans and ecosystems. There is considerable spatial variability in the seasonal cycle of soil moisture, yet a lack of global observations has hindered the development of parsimonious theories explaining that variability. Here, we use 6 years of global satellite observations to describe and explain the seasonal cycle of surface soil moisture globally. An unsupervised clustering algorithm is used to identify five distinct seasonal cycle regimes. Each seasonal cycle regime typically arises in both hemispheres, on multiple continents, and across substantially different local climates. To explain this spatial variability, we then show that the observed seasonal cycle regimes are reproduced very well by a simple but physically based water balance model, which only uses precipitation and downwelling surface shortwave radiation as inputs, and includes no free parameters. Surprisingly, no information on vegetation or land cover is required. To our knowledge, this is the first characterization of the seasonal cycle of surface soil moisture based on global observations.