Elucidating the functional traits that govern the leaf photosynthetic capacity is crucial for predicting vegetation responses to climate change. However, the drivers of leaf photosynthetic capacity in water-limited ecosystems remain poorly understood. Here, we measured light-saturated photosynthetic rate (Asat), leaf mass per area (LMA), area-based leaf nitrogen and phosphorus content (Na(r)ea and Pa(r)ea), and hydraulic traits (sapwood-and leaf-specific conductivity, Ks and KL; Huber value, HV) across 18 dominant woody species in a semi-arid savanna ecosystem of southwest China. Our results showed that deciduous species exhibited higher Asat, consistent with a resource-acquisitive strategy, whereas evergreen species showed lower Asat, reflecting a conservative strategy. Contrary to leaf economics spectrum predictions, Asat correlated poorly with leaf structural and chemical traits (LMA, Na(r)ea, or Pa(r)ea) across species. Instead, Asat was strongly and positively correlated with hydraulic traits (Ks, KL, and HV) across species. In conclusion, leaf habit delineates broad ecological strategies, interspecific variation in leaf photosynthetic capacity within this semi-arid ecosystem is poorly predicted by leaf structural and chemical traits, but is strongly associated with hydraulic traits under water limitation across all species. Our findings underscore the necessity of integrating hydraulic traits and leaf habit as central predictors in models of plant function and carbon-water relations in dryland ecosystems.