Water-saving technologies have long been seen as an effective method to reduce irrigation water use and alleviate regional water shortage. However, growing reports of more severe water shortage and increasing application of water-saving technologies across the world have necessitated reassessment of agricultural water-saving. This study develops a simple method based on satellite-based ET partitions to estimate water withdrawal, water consumption and return flow from the 1980s to 2010s, and quantifies water-savings across globe and four hot-spot irrigated areas at both field and regional scales based on water accounting framework. The results show that global irrigation water flows keep increasing from the 1980s to 2010s, with over 50% increase from the expansion in irrigated lands. While water-saving technologies are found mainly applied in originally old irrigated lands, traditional flooding irrigation is still dominant in newly-developed irrigated lands. Non-beneficial water consumption (soil evaporation) is effectively reduced by water-saving technologies, but return flow has increased at the same time. At field scale, water-saving technologies fail to save water because the accumulated increased return flow is more than the accumulated decreased non-beneficial water consumption. At regional scale, however, water is saved because the return flow percolated to fresh aquifers is seen as beneficial rather than loss. At the same time, the accumulated increase of beneficial water consumption (crop transpiration) exceeds regional water savings, which explains the paradox between wide application of water-saving technologies and more severe regional water shortage. This study provides key new evidence for the paradox of irrigation efficiency and helps reconsidering water-saving technologies and their impacts on regional water resources.