Abstract: National economic development is primarily driven by final consumption demand. Production and consumption processes determine the interregional flows of virtual water embedded in traded goods, thereby reshaping the spatial pattern of water supply and demand. Taking the perspective of consumption-based demand, this study integrates water use and net virtual water inflow and employs a multiregional input-output model to quantitatively analyze the characteristics of intersectoral, interregional virtual water flows. Furthermore, this study develops an extreme learning machine model to predict China's socioeconomic water consumption at the national and regional scales. The results show that the total volume of virtual water increased by nearly 2.95 times from 1987 to 2017, reaching 595.47 billion m³ in 2017. Specifically, the trade network expanded from China's eastern coastal areas toward central and western regions, indicating progressively strengthened interregional linkages. Per capita socioeconomic water consumption was closely associated with economic growth and industrial restructuring, with a generally increasing trend in line with socioeconomic development; meanwhile, provinces dominated by agriculture exhibited a declining trend. In 2017, China's per capita socioeconomic water consumption reached 457.8 m³, with projections indicating that it will peak at 670.4 billion m³ by 2035 and then decline slightly to 652.1 billion m³ by 2050. This trajectory is shaped by the effects of changes in commodity consumption patterns, trade structures, and demographics and reflects the effectiveness of water-saving initiatives and resource management policies in constraining overall water use.