Due to widespread nitrogen limitation, reactive nitrogen input is expected to stimulate vegetation carbon fixation, thus potentially offsetting the decomposition-induced soil carbon losses. However, this projection is largely based on short-term measurements. The crucial question is whether such responses persist over time. Combining a decade-long nitrogen-addition experiment in an alpine grassland with biochemical and plant metabolomics analyses, we showed that a decade of nitrogen addition elicited only a transient increase in net ecosystem productivity (NEP; balance between ecosystem carbon uptake and release); the stimulating effect diminished over time. This phenomenon was likely associated with sustained phosphorus limitation, as indicated by increased plant nitrogen:phosphorus ratio and reduced metabolites involved in phosphorus-related metabolic pathways. The temporal decline in NEP response was reversed with phosphorus supplementation, providing experimental evidence of aggravated phosphorus limitation under long-term nitrogen loadings. Further analysis demonstrated that plant metabolic traits surpassed the classical traits in mediating productivity response to long-term nitrogen:phosphorus imbalance; the metabolites in the Calvin-Benson cycle and pentose phosphate pathway of the dominant species were important predictors of variations in productivity under nutrient additions. Taken together, our findings imply that prior short-term measurements may overestimate the nitrogen-triggered ecosystem carbon sink and provide metabolic insights into the nitrogen effects on primary productivity.