Extended exposure to acid rain has vastly limited soil microbial activity with the consequences for soil carbon (C) storage, but less is known about the microbial responses within soil aggregates that to some extent determine soil C stabilization. Here, we investigated the main microbial group compositions and the relevant potential enzyme activities within different soil aggregates sizes (microaggregates (<250 μm), small macroaggregates (250–2000 μm), and microaggregates (>2000 μm)) in a subtropical forest with decade-long simulated acid rain (SAR) treatments. Four SAR treatments were set by irrigating plots with water of different pH values (i.e., 3.0, 3.5, 4.0, and 4.5 as a control). Results showed that the SAR treatment significantly inhibited microbial activities, specifically decreasing both bacterial and fungal abundances, leading to declines in C-degrading potential enzyme activities. Conversely, potential enzyme activities related to phosphorus (P) and nitrogen (N) mineralization as well as the enzyme stoichiometry for P/N ratio significantly increased under the SAR treatment. The SAR treatment showed no significant differences in microbial abundance across the three soil aggregate sizes. However, it had a more pronounced effect on potential enzyme activities in their optimal aggregate sizes, such as hydrolytic enzymes like β-glucosidase in macroaggregates and oxidases like phenol oxidase and peroxidase in microaggregates. Overall, C-degrading potential enzyme activities were more strongly decreased in the microaggregates than in macroaggregates, and the distribution in aggregates was significantly altered, transforming from large to small sizes under the SAR treatment, which together may boost SOC stabilization and accumulation. Additionally, our findings indicate that prolonged acid rain also caused soil nutrient limitation and imbalance, particularly for P, in subtropical forests. This study highlights the importance of soil aggregate size in regulating microbial responses to acid rain, which should be integrated into ecosystem models to predict soil biogeochemical cycles under future climate conditions.