Quantitative detection of protein, fat, starch, and amino acids in foxtail millet using Fourier transform near-infrared spectroscopy (NIRS) was investigated. Foxtail millet samples (n=259) were analyzed using NIRS. Spectral data were linearized with data from chemical analyses. Calibration models were established using a partial least-squares (PLS) algorithm with cross-validation. Optimized models were tested using external validation set samples with coefficients of determination in the external validation ($R^2{_{val}}$) of >0.90. Residual predictive deviation (RPD) values were nearly equal to or >2.5 for crude protein, alanine, aspartic acid, glutamic acid, isoleucine, leucine, and serine. However, for glycine, histidine, phenylalanine, proline, threonine, tyrosine, and valine, the $R^2{_{val}}$ values were >0.83 and RPD values were nearly equal to or >2.0. For crude fat, total starch, arginine, and lysine, the $R^2{_{val}}$ values were >0.70 and RPD values were >1.5. NIRS is a rapid determination tool for foxtail millet breeding, and for quality control.