We measured the chemical weathering of crest soils derived from granite along a rainfall gradient in semi-arid Kruger National Park. The soils had nearly identical clay minerals, despite straddling a mean annual rainfall between 400 and 800mm. We identified illite, montmorillonite-chlorite, and kaolinite in all the soils, and only recorded vermiculite in the driest soils. Absolute clay percentages were in the 5-15% range on all sites, despite the difference in rainfall. The clay percentages and clay mineralogy of the soils imply that climate plays a minor role for the identity of weathering products. However, when clay percentage is corrected for bulk density, soil depth, and rock fragment, bulk clay percentage is higher in the driest soils. This suggests that the higher production rate of clay in the wet soils is offset by disproportionately greater leaching losses. Zirconium shows a three-fold residual enrichment relative to parent material in the wet soils consistent with weathering-related losses of mobile soil constituents (base saturation < 50%). The drier soils had [Zr] only slightly different from the parent material with marginal base losses (base saturation mostly 100%). We interpret these geochemical and mineralogical patterns as evidence for dual origin of clay due to a difference in rainfall: the soil developing under wetter conditions has lost most of the properties inherited from the parent material whereas the drier soil maintains a strong parent material signature. Therefore most of the clays in the wet soil result from transformation driven by higher rainfall while the clays in the dry soils are inherited directly from primary minerals. We speculate that clay turnover is equal at both ends of the rainfall gradient, but the turnover is driven by different processes: erosion in the dry soils and high rates of chemical weathering in the wet soils.