Earth System Models (ESMs) used to project future biosphere-climate feedbacks rely on predictions of terrestrial carbon dynamics. Furthermore, soil nutrient availability strongly modulates land surface carbon dynamics, including plant sequestration of atmospheric CO2. Plant growth under future environmental changes (e.g., nitrogen and phosphorus deposition) depends on how well plants compete with microbial and abiotic competitors. Here, we surveyed recent developments of nutrient competition representations in ESMs that participated in the CMIP5 project. We found that nutrient competition is over-simplified despite its ecological significance. Existing ESMs either assume that soil-decomposing microbes (1) outcompete plants or (2) are evenly competitive, both of which are inconsistent with theoretical understanding and field observations. We compiled and synthesized global data of forest carbon productivity in response to nitrogen and phosphorus fertilization experiments. Using this synthesis, we show that existing ESMs with the first and second competition schemes lead to underestimation and overestimation, respectively, of fertilization effects on plant growth. We reduced these systematic biases by applying a new competition scheme in CLM4.5 and the essentially equivalent ACME land model (ALMv0) based on the Equilibrium Chemistry Approximation, which is based on classical equilibrium chemical kinetics theory. This approach dynamically updates nutrient competitiveness among multiple consumers (e.g., plants, decomposing microbes, nitrifier, denitrifier, mineral surfaces) as a function of soil nutrient status. There has been a long-term debate regarding how to implement theoretically realistic and computationally efficient nutrient competition schemes in ESMs. Our approach reconciles the complex nature of ecosystem nutrient competition with a computationally tractable approach applicable to ESMs. More importantly, our results imply that previous estimates of plant biomass production in response to nutrient deposition might be systematically biased.