In new microbial-biogeochemical models, microbial carbon use efficiency (CUE) is often assumed to decline with increasing temperature. Under this assumption, soil carbon losses under warming are small because microbial biomass declines. Yet there is also empirical evidence that CUE may adapt (i.e., become less sensitive) to warming, thereby mitigating negative effects on microbial biomass. To analyze potential mechanisms of CUE adaptation, I used two theoretical models to implement a tradeoff between microbial uptake rate and CUE. This rate-yield tradeoff is based on thermodynamic principles and suggests that microbes with greater investment in resource acquisition should have lower CUE. Microbial communities or individuals could adapt to warming by reducing investment in enzymes and uptake machinery. Consistent with this idea, a simple analytical model predicted that adaptation can offset 50% of the warming-induced decline in CUE. To assess the ecosystem implications of the rate-yield tradeoff, I quantified CUE adaptation in a spatially-structured simulation model with 100 microbial taxa and 12 soil carbon substrates. This model predicted much lower CUE adaptation, likely due to additional physiological and ecological constraints on microbes. In particular, specific resource acquisition traits are needed to maintain stoichiometric balance, and taxa with high CUE and low enzyme investment rely on low-yield, high-enzyme neighbors to catalyze substrate degradation. In contrast to published microbial models, simulations with greater CUE adaptation also showed greater carbon storage under warming. This pattern occurred because microbial communities with stronger CUE adaptation produced fewer degradative enzymes, despite increases in biomass. Thus, the rate-yield tradeoff prevents CUE adaptation from driving ecosystem carbon loss under climate warming.