In this paper, we propose a generalized framework that combines the cognitive radio (CR) techniques for spectrum sharing and the simultaneous wireless information and power transfer (SWIPT) for energy harvesting (EH) in the conventional multi-user MIMO (MuMIMO) channels, which leads to an MuMIMO-CR-SWIPT network. In this system, we have one secondary base-station (S-BS) that supports multiple secondary information decoding (S-ID) and secondary EH (S-EH) users simultaneously under the condition that interference power that affects the primary ID (P-ID) receivers should stay below a certain threshold. The goal of the paper is to develop a generalized precoder design that maximizes the sum-utility cost function under the transmit power constraint at the S-BS, and the EH constraint at each S-EH user, and the interference power constraint at each P-ID user. Therefore, the previous studies for the CR and SWIPT systems are casted as particular solutions of the proposed framework. The problem is inherently non-convex and even the weighted minimum mean squared error (WMMSE) transformation does not resolve the non-convexity of the original problem. To tackle the problem, we find a solution from the dual optimization via sub-gradient ellipsoid method based on the observation that the WMMSE transformation raises zero-duality gap between the primal and the dual problems. We also propose a simplified algorithm for the case of a single S-ID user, which is shown to achieve the global optimum. Finally, we demonstrate the optimality and efficiency of the proposed algorithms through numerical simulation results.