The performance of conventional model predictive control (MPC) applied to electric drives typically depends on the accuracy of the parametric system model. However, harsh working conditions in real applications, e.g., electric vehicles and rail transportation, result in a strong time-varying effect on the parameters of the motor and impact the performance of MPC. In this article, a real-time capable data-based predictive control is introduced to permanent magnet synchronous motor (PMSM) drives as an alternative to conventional model-based algorithms. In the proposed method, raw input-output data are online collected and stored to directly determine the optimal control action skipping the stage of parameter identification. A direct data exploration allows an accurate predictive modeling by setting up a mapping relationship between input and output measurements. Another peculiarity of the proposed data-based algorithm is that it avoids any complex parameter configurations, resulting in a concise control paradigm. The proposed data-based controller is evaluated on a PMSM drive setup and compared with the conventional model-based method to demonstrate the high robustness and satisfactory control performance during transients and steady state.