Brazil has the world's most populous semi-arid region and climate change represents significant ecological and socioeconomic challenges for this area. To better understand the impact of these changes, it is crucial to analyze the dynamics of climate variables and evapotranspiration (ETo), a critical climate variable. This study aimed to model ETo rates considering climate change scenarios in the Brazilian Semi-arid region (BSR). The modeling was based on tests of five machine learning algorithms: Bayesian Regularized Neural Networks (BRNN), Cubist, Earth, Linear Regression (LM), and Random Forest (RF). A dataset with 20 covariates was used to represent the current scenario. In the future prediction, covariates from two shared socio-economic pathways were used (SSPs 126 and 585). The best statistical performance was achieved by Cubist (R² = 0.98 and RMSE = 0.08 mm day-¹ in the holdout-test). The current daily average ETo is 4.77 mm day-¹, while in future scenarios, it can increase by 3.56% in SSP 126 and 15.51% in SSP 585. ETo rates are expected to expand territorially; ranges from > 0.60 mm day-¹ should increase 8% in SSP 126 and 40% in SSP 585. The applied model suggests that ETo may increase in future scenarios in the BSR, which could affect biodiversity levels and intensify social conflicts.