Abstract: A large number of small and medium-sized data centers with high energy consumption in urban areas have great potential for load regulation, which can be the main carriers for demand response programs. However, most researchers take such a single data center as the object of study without considering the possibility of integrating the computing resources of multiple data centers. Therefore, this paper proposes a two-stage robust optimization model for colocation data centers based on the computing resource sharing mode. First, tenants of the colocation data centers are divided into the day-ahead group and the backup group. Then, tenants of the day-ahead group participate in demand response, and that of the backup group use the Nash bargaining game to share their computing resources with the day-ahead group for calculating uncertain workload. Finally, the two-stage robust optimization model is applied to formulate day-ahead regulation and backup regulation plans. The simulation results show that the proposed model is able to improve the total payoff of the data center operator. The results also prove that the model will reduce the operator's dependence on the diesel generators when there is uncertainty in the load reduction.