In this paper, the performance of an underlay multiple-input multiple-output (MIMO) cognitive radio system is analytically studied. In particular, the secondary transmitter operates in a spatial multiplexing transmission mode, while a zero-forcing detector is employed at the secondary receiver. Additionally, the secondary system is interfered by single-antenna primary users (PUs). To enhance the performance of secondary transmission, optimal power allocation is performed at the secondary transmitter with a constraint on the maximum allowable outage threshold specified by the PUs. Further, the effective number of secondary transmit antennas is specified based on the optimal power allocation for an arbitrary MIMO scale. Also, a lower bound on the ergodic channel capacity of the secondary system is derived in a closed-form expression. Afterwards, the scenario of a massive MIMO secondary system is thoroughly analyzed and evaluated, where the harvesting-enabled secondary transmission is studied. The optimal power allocation, the effective number of secondary transmit antennas, the efficient tradeoff between transmit-and-harvest secondary antennas, and the average channel capacity of the secondary system are analytically presented. Finally, extensive numerical and simulation results corroborate the effectiveness of our analysis, while some useful engineering insights are provided.