Abstract: Water resources carrying capacity (WRCC) is an integrative metric for assessing whether a regional water system can sustain socio-economic development while maintaining essential ecological functions under both natural constraints and human regulation. Ordos City (Inner Mongolia, China) is a representative arid and semi-arid resource-based city where water scarcity and high development intensity create pronounced water–economy–ecology trade-offs. Using panel data for 2008–2022 across seven banners and two districts, this study develops an integrated DPSIRM–TOPSIS–SDE–GeoDetector framework to quantify WRCC, diagnose spatial evolution, and identify dominant drivers. Following the DPSIRM logic (Driving force–Pressure–State–Impact–Response–Management), a multidimensional indicator system comprising six subsystems and 22 indicators was constructed to represent development intensity, water-use stress, resource endowment and utilization status, ecological impacts, response measures, and governance and management capacity. The explicit Management (M) dimension was introduced to capture policy implementation and regulation effectiveness. Data were compiled from official water-resources bulletins and sectoral plans, statistical yearbooks/bulletins, and eco-environmental records; limited missing values were estimated by interpolation to ensure consistent time series. Indicator weights were derived using a hybrid AHP–entropy scheme to combine expert judgment with objective information content, and WRCC was measured by TOPSIS closeness coefficients. Spatial dynamics were characterized by the standard deviational ellipse (SDE) to track clustering direction, centroid movement, and dispersion, and GeoDetector was employed to quantify the explanatory power (q-statistic) of candidate determinants. The results indicate a marked improvement in WRCC in Ordos over 2008–2022. The composite index increased from 0.336 in 2008 to 0.597 in 2022, a 77.7% rise, and the multi-year mean was 0.480, implying an overall “critical carrying capacity” status approaching the “weak carrying capacity” threshold (0.6). The upward trajectory accelerated after 2012; during 2016–2020 WRCC increased by about 0.030 per year on average, suggesting strengthened regulation and improved allocation effectiveness in the later period. Weight decomposition suggests that utilization stress and system conditions dominate the evaluation: Pressure (0.227) and State (0.298) have a combined weight of 0.525. At the indicator level, the ecological/environmental water-use ratio has the highest weight (0.105), followed by water supply per unit area (0.091) and the exploitation rate (0.078), underscoring the centrality of ecological allocation efficiency, supply capacity, and control of development intensity. Spatially, WRCC exhibits a persistent “higher in the east, lower in the west” gradient. Long-term mean WRCC is highest in Jungar Banner (0.511), Kangbashi District (0.505), and Dalad Banner (0.500), whereas Hanggin Banner (0.477), Dongsheng District (0.469), and Ejin Horo Banner (0.468) remain below the citywide average. SDE parameters indicate a stable northeast-oriented clustering pattern with modestly reduced dispersion: the azimuth increased from 65.6° to 69.4°, while the major and minor axes decreased by approximately 5.51% and 2.31%, respectively, relative to their 2008 values, implying a slight convergence of interregional differences. Subsystem diagnostics clarify mechanisms behind the aggregate trend. Driving force and Pressure remained on an upward trajectory (multi-year means 0.219 and 0.240), reflecting sustained growth-related demand and rising stress on the water system, whereas the State subsystem was relatively stable (mean 0.344). Governance and management capacities improved markedly: Response reached 0.303 in 2022, and Management rose from 0.048 in 2010 to 0.292 in 2022. In contrast, the Impact subsystem fluctuated strongly (mean 0.132; peak 0.189 in 2018), indicating that ecological feedbacks remain a key constraint. GeoDetector attribution identifies the ecological/environmental water-use ratio as the dominant determinant (q = 0.961), followed by water supply per unit area (q = 0.731) and exploitation rate (q = 0.730). Additional influential factors include the groundwater abstraction coefficient (q = 0.657), per-capita water resources (q = 0.591), and tertiary-industry share (q = 0.513). Consistent with these findings, water supply per unit area declined slightly during 2010–2014 (from 18.44 to 18.04 m³/km²) but increased rapidly after 2015, reaching 47.75 m³/km² in 2021. Spatial contrasts in ecological allocation are also evident, with the ecological/environmental water-use ratio at 5.1% in Hanggin versus 22.9% in Kangbashi, highlighting substantial heterogeneity in ecological water allocation. Overall, Ordos has shifted from pronounced overload risk toward a near-threshold carrying state, driven mainly by strengthened response–management capacity and improved ecological water allocation, yet vulnerability persists due to volatile ecological impacts and continued dependence on groundwater. The proposed DPSIRM–TOPSIS–SDE–GeoDetector framework provides a transferable approach for WRCC assessment and driver attribution in arid, development-intensive, resource-based regions, supporting differentiated zoning control and targeted interventions to enhance WRCC resilience.