Lithium is a key strategic metal with important reserves in pegmatites. However, the formation processes of Li-pegmatites remain doubtful, as most previous models are either qualitative or require large numbers of cyclic fractionation-extraction events that have little direct field evidence. Here, we propose that highly Li-enriched melts in the Himalaya may separate via an unusual combination of rapid crystal growth relative to diffusion of Li and water, and deformation in the context of the South Tibetan Detachment System (STDS). Models of boundary layer formation for felsic liquids show that rapid crystallization (1×10−7 to 1×10−8 m/s; c. 1 mm/day) of anhydrous minerals forms H2O+Li-rich, mobile boundary layers, whereas H2O+Li-poor, immobile boundary layers form adjacent to micas. Shearing can preferentially segregate inviscid, H2O+Li-rich, boundary layers, producing extreme compositions and compositional gaps relative to parent magmas. The rock record of some magmas may not always reflect homogeneous liquid fractionation, rather preferential extraction of hydrous boundary layer liquids that develop adjacent to anhydrous minerals. This model explains whole-rock chemistry and field observations from the STDS, where Li-rich pegmatites and potential coeval source magmas occur with a distinct compositional gap. If our model is correct, the c. 2200 km-long STDS of Himalaya may represent an important zone of Li enrichment and site for future strategic Li-pegmatite reserves. Preferential extraction of boundary layers implies some evolved felsic magmas may not record crystallization of hydrous minerals.