Bearings are at the heart of all rotating equipment. The condition of the bearings often reflects the health of the machine. In particular, spindles, which are important components in various machining operations, are highly dependent on the reliability of their bearings. Therefore, diagnosing bearing defects is critical for managing bearing health. Towards this, the present work explores a methodology for classifying bearing faults using the stacked ensemble technique (SET). The SET leverages the strength of several underlying models (e.g. logistic regression, decision trees etc.) chosen due to their unique properties. The base models are trained on bearing vibration data to make predictions, and then a meta-learner is trained to combine these predictions. The proposed methodology is implemented on vibration data generated from seeded bearing defect experimentation. The ensemble stacking approach not only improves predictive accuracy, but also mitigates the limitations inherent in individual models. The methodology classifies different bearing faults with 100% accuracy.