The total amount of N derived from symbiotic N2 fixation of legumes in mixed grasslands is hypothesized to be controlled by soil nitrogen (N) and phosphorus (P) availability, which might affect both (i) the legume biomass production and (ii) the N2 fixation rate per unit of legume biomass. We evaluated symbiotic N2 fixation in 17 grasslands on four continents that are subjected to the same full factorial N and P addition treatment. All sites are part of the global Nutrient Network project (NutNet). At each site, four treatments were studied: N addition (+N), P addition (+P), as well as N and P addition (+NP) and a control treatment without element addition (control). Application of N and NP reduced the biomass production of legumes when compared to the control. In contrast, +P had no significant impact on the aboveground biomass production of legumes when compared to the control, but aboveground biomass production of legumes was consistently higher in +P than in the +N and +NP treatments. The negative impact of N addition on legume biomass production is likely related to lower competitiveness of legumes compared to grasses and non-fixing forbs at high N supply. However, nutrient addition had no effect on the N2 fixation rate per unit biomass. However, due to the higher legume biomass production, the annual N2 fixed per grassland area was significantly increased in the +P treatment (7.55 kg N ha-1 yr-1) compared to +N (2.01 kg N ha-1 yr-1) and +NP (2.20 kg N ha-1 yr-1). Our results reveal that the effects of nutrient addition on symbiotic N2 fixation are driven by changes in grassland composition (abundance of legumes) rather than changes in the N2 fixation rate. Therefore, a reduction in legume biomass, and thus reduced N2 fixation per area grassland can be expected in the long-term due to increasing soil N enrichment by anthropogenic activities.