Most current models of soil C dynamics predict that climate warming will accelerate soil C mineralization, resulting in a long-term CO2 release and positive feedback to global warming. However, ecosystem warming experiments show that CO2 loss from warmed soils declines to control levels within a few years. Here, we explore the temperature dependence of enzymatic conversion of polymerized soil organic C (SOC) into assimilable compounds, which is presumed the rate-limiting step of SOC mineralization. Combining literature review, modelling and enzyme assays, we studied the effect of temperature on activity of enzymes considering their thermal inactivation and catalytic activity. We defined the catalytic power of enzymes (Epower) as the cumulative amount of degraded substrate by one unit of enzyme until its complete inactivation. We show a universal pattern of enzyme’s thermodynamic properties: activation energy of catalytic activity (EAcat) < activation energy of thermal inactivation (EAinact). By investing in stable enzymes (high EAinact) having high catalytic activity (low EAcat), microorganisms may maximize the Epower of their enzymes. The counterpart of such EAs’ hierarchical pattern is the higher relative temperature sensitivity of enzyme inactivation than catalysis, resulting in a reductionin Epower under warming. Our findings could explain the decrease with temperature in soil enzyme pools, microbial biomass (MB) and carbon use efficiency (CUE) reported in some warming experiments and studies monitoring the seasonal variation in soil enzymes. They also suggest that a decrease in soil enzyme pools due to their faster inactivation under warming contributes to the observed attenuation of warming effect on soil C mineralization. This testable theory predicts that the ultimate response of SOC degradation to warming can be positive or negative depending on the relative temperature response of Epower and microbial production of enzymes.