This paper investigates simultaneous wireless information and power transfer (SWIPT) in a multiple-input multiple-output (MIMO) broadcast system with three nodes: one transmitter and two receivers (one for energy harvesting and one for information decoding). The system considers two scenarios: separated receivers with different MIMO channels and co-located receivers with identical MIMO channels. For separated receivers, the optimal transmission strategy is derived to achieve different tradeoffs between maximal information rate and energy transfer, characterized by the boundary of a rate-energy (R-E) region. For co-located receivers, an outer bound for the R-E region is shown due to the limitation that practical energy harvesting receivers cannot decode information directly. Two practical designs for co-located receivers are investigated: time switching and power splitting. The achievable R-E regions for these designs are compared to the outer bound. The paper also presents a system model for SWIPT, formulates the problem of finding the optimal transmit covariance matrix, and derives the optimal solutions for both separated and co-located receivers. The results show that the optimal tradeoff between information and energy transfer is characterized by the boundary of the R-E region, and that the performance of practical receiver designs can be analyzed in comparison to this bound.This paper investigates simultaneous wireless information and power transfer (SWIPT) in a multiple-input multiple-output (MIMO) broadcast system with three nodes: one transmitter and two receivers (one for energy harvesting and one for information decoding). The system considers two scenarios: separated receivers with different MIMO channels and co-located receivers with identical MIMO channels. For separated receivers, the optimal transmission strategy is derived to achieve different tradeoffs between maximal information rate and energy transfer, characterized by the boundary of a rate-energy (R-E) region. For co-located receivers, an outer bound for the R-E region is shown due to the limitation that practical energy harvesting receivers cannot decode information directly. Two practical designs for co-located receivers are investigated: time switching and power splitting. The achievable R-E regions for these designs are compared to the outer bound. The paper also presents a system model for SWIPT, formulates the problem of finding the optimal transmit covariance matrix, and derives the optimal solutions for both separated and co-located receivers. The results show that the optimal tradeoff between information and energy transfer is characterized by the boundary of the R-E region, and that the performance of practical receiver designs can be analyzed in comparison to this bound.