In this paper, we present an approach to solve the nonconvex optimization problem that arises when designing the transmit covariance matrices in multiuser multiple-input multiple-output (MIMO) broadcast networks implementing simultaneous wireless information and power transfer (SWIPT). The MIMO SWIPT design is formulated as a nonconvex optimization problem in which system sum rate is optimized considering per-user harvesting constraints. Two different approaches are proposed. The first approach is based on a classical gradient-based method for constrained optimization. The second approach is based on difference of convex (DC) programming. The idea behind this approach is to obtain a convex function that approximates the nonconvex objective and, then, solve a series of convex subproblems that, eventually, will provide a (locally) optimum solution of the general nonconvex problem. The solution obtained from the proposed approach is compared to the classical block-diagonalization (BD) strategy, typically used to solve the nonconvex multiuser MIMO network by forcing no inter-user interference. Simulation results show that the proposed approach improves both the system sum rate and the power harvested by users simultaneously. In terms of computational time, the proposed DC programming outperforms the classical gradient methods.