This paper aims to analyze the throughput of a multiple-input multiple-output (MIMO) multiple-access (MA) wireless network. Some recent research studies showed that, in the concerned scenario, the joint processing of all the received signals by access point (AP) cooperation can dramatically enhance the system throughput because of the macrodiversity. However, in practice, the channel estimation quality for the paths from the macrodiversity providers is usually poor. Unfortunately, existing literature on the performance of the joint processing in the MIMO cellular networks is deficient in channel estimation errors. To fill this literature gap, this paper provides an analytical framework for calculating the large-system throughput of a MIMO cellular network with channel estimation errors. Unlike most large-system results, analytical results of this study can be applied in the scenarios with any given input distributions (not necessarily Gaussian) from transmitters. In addition, several issues including 1) how much of the coherence interval should be spent on training and 2) the impact of AP cooperation are highlighted by the analytical results.