Field capacity (FC) and permanent wilting point (PWP) are the important soil hydraulic properties which determine the availability of water for plant growth. Conventional estimation of soil hydraulic properties is tedious and expensive which are normally derived from pedo-transfer functions. Visible and near-infrared (Vis-NIR) spectroscopy is a low-cost, non-destructive alternative method for the rapid estimation of soil properties. The present study was carried out to predict soil hydraulic properties using Vis-NIR spectral data in the semi-arid region of the Northern Karnataka plateau. Spectral data of 558 soil samples were acquired using Field Spec ASD Spectroradiometer and three models (support vector machine, random forest and partial least square regression) were evaluated for prediction of FC and PWP. The models were calibrated using 2/3rd of total observations and validated using 1/3rd of observations. The validation results showed that RF and SVM provided slightly higher performance compared to PLSR model. Permanent wilting point was predicted well (R2 = 0.70–0.74, RMSE = 5.44–5.74%) compared to field capacity (R2 = 0.66–0.69 and RMSE = 7.25–7.51%). Among the soil orders, FC and PWP of Vertisols were poorly predicted by Vis-NIR spectra (R2 = 0.34&0.42) and moderately predicted for Alfisols (0.44&0.52) and Inceptisols (0.55&0.65) while excellent prediction was recorded for Entisols (0.83&0.76). The present results showed that Vis-NIR spectroscopy is helpful for fast estimation of hydraulic properties in semi-arid regions of the country with high accuracy.