Ensemble-based global climate models （GCMs） are commonly used for predicting hydrological variables, but the diversity of GCMs introduces significant uncertainty in model predictions. Thus, this study employs the long short-term memory (LSTM) model combined with ensemble predictions from seven GCMs to forecast monthly groundwater storage (GWS) in the northeast region under four scenarios from 2022 to 2100. Results indicate that the GWS in the northeast region exhibits a general increasing trend, with the highest growth rate of 0.20 mm/a under the SSP585 scenario through equal-weight averaging. However, significant spatial and temporal differences are observed in the output results of different GCMs, with the maximum difference in trends reaching 0.33 mm/a. Using 44°N as the boundary, all GCMs show a consistent trend of decreasing GWS in the south and increasing GWS in the north compared to the historical period under different scenarios. The findings of this study can shed some light on the application of GCMs in climate simulation and the prediction of future groundwater storage in northeast China.