The authors consider the uplink of a one-dimensional 2-cell network with fixed Base Stations (BSs) and randomly distributed User Terminals (UTs). Assuming that the number of antennas per BS and the number of UTs grow infinitely large, they derive tight approximations of the ergodic sum rate with and without multicell processing for optimal and sub-optimal detectors. They use these results to find the optimal BS placement to maximize the system capacity. This work can be seen as a first attempt to apply large random matrix theory to the study of networks with random topologies. They demonstrate that such an approach is feasible and leads to analytically tractable expressions of the average system performance.