Despite being very common in the territory of Chihuahua, Chihuahua, Mexico, to experience drought, its consequences continue to severely impact the population without prior warning. Machine learning has proven to have a significant capacity for predicting time series, and the Standardized Precipitation Evapotranspiration Index (SPEI) is emerging as the most accurate drought indicator. In this study, predictive models were developed using Artificial Neural Networks (ANN), Long-Short Term Memory (LSTM), and Support Vector Regression (SVR) for estimating SPEI. Temporal scales of 12 months (SPEI 12) and 24 months (SPEI 24) for the period 1901-2020 in the mentioned territory were considered. This was done in order to simulate the behavior of drought cycles and enhance the ability to anticipate consequences. The accuracy indices used to evaluate the models were the mean squared error (MSE), mean absolute error (MAE), mean bias error (MBE), coefficient of determination (R2), and Kendall coefficient. In total, 956 experiments were conducted using the three methods, varying parameters such as the number of neurons, kernel, and polynomial degree. The two best models for each method were selected, and the average results revealed MSE = 0.0051, MAE = 0.0537, MBE = 0.0218, R2 = 0.8495, and Kendall coefficient = 0.7592 for SPEI 12; and MSE = 0.0024, MAE = 0.0375, MBE = 0.0162, R2 = 0.9218, and Kendall coefficient = 0.8558 for SPEI 24.