Agitation torque associated with oil lubricant is one important factor of torque loss in bearings under sufficient lubricating conditions. So far, efforts on reducing agitation torque were taken mostly by means of conventional experimental trials. Aiming for speedy, low-cost development, a calculation program for predicting the amount of agitation torque and oil distribution tendency in rolling bearings has been developed using computational fluid dynamics (CFD) analysis.At first, since rolling bearings are axially symmetric, sector models of bearings were adopted. To verify the method, torque losses and oil quantities in ordinary-sized bearings have been measured. Calculated values based on sector models are qualitatively in good agreement with measured results. The difference between the absolute values of measured and calculated torque may be caused by the difference between the vertical model used in CFD analysis and the horizontal torque-testing rig used in measurement.To improve the accuracy, full models of the bearings have been developed and verified by experiment. The measured values are about 1.1 times as high as the calculated ones. It seems reasonable that this method can predict the agitation torque quantitatively in practical accuracy.Using this technique, the influence of bearing cage shape on oil flow and agitation torque has been reliably investigated, resulting in further development of low-torque ball bearing. Oil flow in this bearing is controlled and optimized by specially designed plastic cage. Thus, agitation torque is reduced significantly in comparison with conventional products.