Within the ecosystem, mangroves serve diverse functions by acting as a carbon sink, removing nitrogen, providing habitats for marine organisms, and regulating climate. Nowadays, many mangrove forests are being converted to shrimp ponds across the world, and several studies have quantified the effects of this conversion on soil carbon stocks and greenhouse gas emissions. However, its effects on the soil microbial community structure and assembly processes remain unclear. In the present study, high-throughput sequencing and multivariate statistical analyses were used to quantify the characteristics of soil microbial communities in common mangrove types in southeastern China and the adjacent converted shrimp ponds. The relative abundance of Verrucomicrobia, Cyanobacteria, and Firmicutes was significantly increased in the shrimp pond sediment compared with that in the mangrove sediment, whereas the archaea, represented by Crenarchaeota and Euryarchaeota, exhibited the opposite patterns. Moreover, the sediment microbial communities in the shrimp ponds exhibited more obvious biogeographical distributions than those in the mangrove habitat. Null model analysis revealed that variable selection (via deterministic processes) governed the microbial community assembly in the mangrove sediments, while dispersal limitation (via stochastic processes) shaped the microbial community structure in the shrimp pond sediments. Our findings suggest that converting mangrove forest habitats to shrimp ponds significantly alters the sediment microbial networks, rendering them unstable (by disrupting the network topology parameters such as modularity, total cohesion, and negative cohesion). This may in turn lead to alterations in various ecosystem functions in response to habitat conversion, highlighting the need for better preserving mangroves through appropriate climate change mitigation measures.