This paper investigates the potential of coordinated air conditioning control for a simulated community of 900 homes with a high penetration of rooftop solar photovoltaic (PV) panels. The simulated community of homes is created from an extensive data set including home energy audits, homeowner surveys, and electricity meter measurements from actual homes in Austin, Texas, USA. Coordinated air conditioning control in the homes is simulated using a rolling horizon model predictive controller to minimize the peak demand of the community using both centralized and decentralized control methods. By manipulating thermostat set points, the controller takes advantage of the thermal mass of the buildings to store thermal energy. In all cases considered, the centralized controller outperforms the decentralized controller, but both lead to significant reductions in peak electricity demand. We find that decentralized control achieves nearly the same peak reduction as the centralized control method when all homes have rooftop PV. We also find that coordinated air conditioning control achieves a marginally smaller benefit as the penetration of rooftop PV increases.