Arid grassland ecosystems play a crucial role in maintaining regional ecological balance and supporting pastoral production. However, overgrazing and climate change have led to the degradation of ecosystem services (ESs). Understanding the complex relationship between grazing intensity (GI) and ESs is essential for achieving sustainable management of grassland ecosystems. This study focuses on the arid grasslands of Xinjiang, China, using dual-scale analysis (grid and county levels), and integrates multi-source data and models such as the Geodetector to evaluate the spatiotemporal dynamics of GI on ecosystem service capacity (ESCL) and synergy level (ESSL) from 2000 to 2020. Results show significant synergy among ESs, with GI playing a key interactive role in influencing them. Both ESCL and ESSL exhibit an inverted U-shaped relationship with GI: at low GI (GI < 0.4), ESCL and ESSL demonstrate synergy, but when GI exceeds 0.6, a trade-off emerges. The impact of different grazing intensities on ESCL and ESSL varies significantly across spatial clusters, with medium grazing levels generally sustaining the sustainability of grassland ecosystems. In fragile regions like Southern Xinjiang, maintaining low GI is crucial for protecting ESs, while in more resilient areas like Northern Xinjiang, medium GI (GI 0.4-0.6) helps optimize the benefits of ESs. The findings provide scientific support for grazing management strategies in Xinjiang's arid grasslands and offer theoretical guidance for the sustainable use of grassland resources in arid regions globally.