In this correspondence, we seek to maximize the signal-to-interference-plus-noise ratio (SINR) of cognitive users (CUs) by designing an optimal receive combiner (ORC), taking into account interference from the primary base station (PBS) to CUs [inter-PBS interference (IPI)], multiuser interference (MUI) among CUs, and desired channel gains. By employing the Rayleigh-Ritz quotient result, we formulate an optimal receiver design and then determine the ORC by solving tfshe derived standard eigenvalue problem. The ORCs are derived as closed forms according to the presence or absence of interference beam information at each CU. When the interference beam information is available, the maximum SINR is given as the largest eigenvalue of the received SINR. In the absence of interference beam information, the ORC based on prediction is proposed based on the prediction of the worst IPI and MUI directions and asymptotically analyzed in terms of the number of transmit antennas. Simulation results demonstrate that the ORC maximizes the sum rate of the cognitive multiuser multiple-input/multiple-output system compared with the reference receive combiners.