Blood metabolites on individual animals give relevant information on energy, protein and other nutrient metabolism in dairy cows and they are widely used to assess their nutritional and health status, but their analysis is expensive and time consuming. Near-infrared (NIR) spectroscopy represents a rapid, non-destructive sample analysis, cost-effective alternative to traditional methods of analysis that has been successfully used in many fields and that requires small amount of sample. The aim of this study was to evaluate the use of NIR spectroscopy for analysis of a set of blood biochemical parameters included in a metabolic profile for dairy cows. The NIR spectra were acquired from 200 µL of plasma samples of 149 pluriparous dairy cows from 7 herds in different physiological phases, from dry period to mid-lactation. A Fourier-transform (FT)-NIR Analyzer (MPA, Bruker Optics, Germany), with a spectral resolution of 4 cm⁻¹ over a wavelength range of 12,500–4,000 cm⁻¹ with 32 scans per spectrum, was used in transmission mode. Reference values, obtained using accepted reference biochemical methods, were used as calibrating values to develop predictive models using a partial least square method. The validation was carried out using cross validation method (leave one-out sample procedure) and its accuracy was evaluated considering the coefficient of determination (R²) and residual prediction deviation (RPD). Predictions were obtained for cholesterol, total protein, globulin and albumin (R² from 0.95 to 0.99; RPD from 4.29 to 12.4). Nonesterified fatty acids and total bilirubin had an approximate quantitative prediction level (R² = 0.61 and 0.62; RPD = 1.61 and 1.62), whereas urea, β-OH-butyric acid and glucose had non-usable predictions (R² from 0.36 to 0.22; RPD from 1.25 to 1.13). This study supports the use of NIR spectroscopy as a substitute of some biochemical methods and has comparable performance to those previously obtained with FT-MIR (MilkoScan FT 120, Foss, Denmark). As plasma samples were obtained from healthy animals, a quite narrow range of variability were present in the data set for some plasma parameters. This could have limited the predictive potential of models and, for that reason, further studies on a larger data set are in progress.