The theory of multiple-input multiple-output (MIMO) technology has been well-developed to increase capacity in single-input single-output (SISO) fading systems. This capacity gain can often be leveraged by linearly precoding the transmit symbols utilizing channel state information at the transmitter (CSIT). In this dissertation, we study multiuser MIMO systems when perfect or partial CSIT is available. This perfect or partial CSIT can be fed back from the receiver(s) to the transmitter(s). First, we consider linearly precoded MIMO multiple access systems and obtain the capacity region when partial CSIT (an index corresponding to a codeword in a codebook) is available at each user. We also discuss how to determine the precoding matrix according to the partial CSIT to maximize the sum-rate. Second, we consider unitary precoded MIMO broadcasting systems when partial CSIT is available at the basestation. We discuss how the unitary precoding matrix is designed to maximize the sum-rate using the partial CSIT. Finally, we consider MIMO multicasting systems when perfect CSIT is available at the basestation. We propose a computationally simple recursive algorithm to determine the precoding matrix and the precoding vector to maximize the achievable rate.