The purpose of this research study is to investigate the applicability of deep learning, and more especially the long-short-term memory (LSTM) model, to the prediction of droughts, one of the extreme weather events caused by global warming. The research considered the standardized precipitation index (SPI), a drought index that is commonly used in the scientific community to estimate the drought severity. The study was conducted using one of the global climate models from the six-phase coupled model intercomparison project (CMIP6), the latest version of the CMIP models, namely the Canadian Earth System Model version 5 (CanESM5-1). The data were collected using historical and future simulations. The shared socio-economic pathway (SSP5-8.5), which represents the worst-case scenario for future climate change forecasts, is the foundation for the future data. The CanESM5-1data were utilized throughout the nineteenth, twentieth, and twenty-first centuries (1850–2100). The location of this study is the Canadian prairies which include Alberta, Saskatchewan, and Manitoba. To train and test the LSTM machine learning algorithm, the data were split between 70% training, 15% validation, and 15% testing sets. Seventy percent of the data were used for training the LSTM machine learning algorithm, fifteen percent for validation, and fifteen percent for testing. The outputs of the LSTM model were compared with the drought indices calculated from numerical methods. The statistical analysis was conducted using two statistical metrics: the mean-square error (MSE) as a loss function and R-squared (R2) to evaluate the performance of the model. The LSTM model demonstrated remarkable predictive accuracy, with a MSE value of 0.0035, 0.004, and 0.004 and R2 of 0.87, 0.78, and 0.74 for Saskatchewan, Alberta, and Manitoba, respectively.