Our objective was to evaluate a near-infrared reflectance spectroscopy (NIRS) used in the feed industry to estimate the potential for grains to increase the risk of ruminal acidosis. The existing NIRS calibration was developed from in sacco and in vitro measures in cattle and grain chemical composition measurements. To evaluate the existing model, 20 cultivars of 5 grain types were fed to 40 Holstein heifers using a grain challenge protocol and changes in rumen VFA, ammonia, lactic acids, and pH that are associated with acidosis were measured. A method development study was performed to determine a grain feeding rate sufficient to induce non-life threatening but substantial ruminal changes during grain challenge. Feeding grain at a rate of 1.2% of BW met these criteria, lowering rumen pH (P = 0.01) and increasing valerate (P < 0.01) and propionate concentrations (P = 0.01). Valerate was the most discriminatory measure indicating ruminal change during challenge. Heifers were assigned using a row by column design in an in vivo study to 1 of 20 grain cultivars and were reassigned after a 9 d period (n = 4 cattle/treatment). The test grains were dry rolled oats (n = 3), wheat (n = 6), barley (n = 4), triticale (n = 4), and sorghum (n = 3) cultivars. Cattle were adapted to the test grain and had ad libitum access to grass silage 11 d before the challenge. Feed was withheld for 14 h before challenge feeding with 0.3 kg DM of silage followed by the respective test grain fed at 1.2% of BW. A rumen sample was taken by stomach tube 5, 65, 110, 155, and 200 min after grain consumption. The rumen is not homogenous and samples of rumen fluid obtained by stomach tube will differ from those gained by other methods. Rumen pH was measured immediately; individual VFA, ammonia, and D- and L-lactate concentrations were analyzed later. Rumen pH (P = 0.002) and all concentrations of fermentation products differed among grains (P = 0.001). A previously defined discriminant score calculated at 200 min after challenge was used to rank grains for acidosis risk. A significant correlation between the discriminant score and the NIRS ranking (r = 0.731, P = 0.003) demonstrated the potential for using NIRS calibrations for predicting acidosis risk of grains in cattle. The overall rankings of grains for acidosis risk were wheat > triticale > barley > oats > sorghum.