This paper investigates multi-user multi-input multi- output cognitive wireless powered communication networks (WPCN), in which a secondary WPCN shares spectrum with a primary wireless information transfer system. A typical WPCN consists of two different phases. In the first downlink phase, a hybrid access point (H-AP) transfers energy to charge users, and then in the subsequent uplink phase, the users send information by using the harvested energy to the H-AP. We consider underlay cognitive WPCN without cooperation between the primary transmitter and the secondary H-AP. In this case, we formulate sum throughput maximization problem by taking the interference leakage to the primary network into consideration. The problem is generally non-convex due to coupled variables in the WPCN. To tackle this issue, we first convert the problem into equivalent convex form, and then identify the global optimal solution by applying the proposed iterative optimization algorithm. Finally, simulation results demonstrate that the proposed algorithm outperforms conventional schemes.