Energy-efficiency is a critical issue in wireless sensor networks (WSNs). Cooperative MIMO and data aggregation can conserve energy in WSN nodes by reducing transmit power through cooperative communication and by eliminating the inherent redundancy in raw data, respectively. Despite a few recent studies on the combination of these two techniques, none of them answered the question of how the network can be effectively partitioned into clusters of cooperating nodes, i.e., the cluster formation problem—even the state-of-the-art oversimplifies the problem under some optimistic (and unrealistic) assumptions. In this paper, we utilize a coalitional game-theoretic approach to model the cluster formation under cooperative data aggregation. Based on some coalition valuation and merge-split rules, a new and efficient clustering algorithm, named G-MIMOA, is proposed for network partitioning in a distributed manner. The resulting partition is proved to be Dp-stable, thus converging to the globally optimal clusters. Extensive simulations are conducted to evaluate the performance of our proposed approach and demonstrate its effectiveness. According to the results, a careful combination of cooperative MIMO and data aggregation in G-MIMOA can improve them by more than 7x increase in network lifetime. Moreover, G-MIMOA achieves more than 20% higher lifetime compared to the best known alternatives.