The local energy market (LEM) has emerged as an effective way to automatically regulate the power balance and promote distributed energy utilization of the distribution system. With the increasing integration of data centers (DCs) with high load-shifting flexibility into the distribution network, they have gradually become significant market players in the LEMs. However, limited research has been done on the optimal trading strategy of DCs within LEMs. A bi-level optimization model for networked DCs across multiple LEMs is proposed to bridge the research gap. Firstly, since DCs with substantial power consumption can directly impact the market clearing prices, LEM participants are classified into DC-type and conventional-type prosumers, both analytically modeled. Secondly, a LEM model incorporating DCs is created, employing the alternating direction method of multipliers (ADMM) to clear the market in a distributed method, thereby ensuring end-user privacy. Thirdly, a bi-level optimal trading strategy for interconnected DC prosumers across various LEMs is proposed to distribute workloads among DCs optimally. Fourth, the differential evolution and meta-model address the bi-level optimization problem efficiently. The simulation results indicate that this strategy enhances networked DCs’ benefits, boosts renewable energy utilization, and improves overall social welfare.