The future of the land carbon sink depends on the availability of nitrogen (N) and, specifically, on symbiotic N fixation , which can rapidly alleviate N limitation. The temperature response of symbiotic N fixation has been hypothesized to explain the global distribution of N-fixing trees and is a key part of some terrestrial biosphere models (TBMs) , yet there are few data to constrain the temperature response of symbiotic N fixation. Here we show that optimal temperatures for N fixation in four tree symbioses are in the range 29.0-36.9 °C, well above the 25.2 °C optimum currently used by TBMs. The shape of the response to temperature is also markedly different to the function used by TBMs (asymmetric rather than symmetric). We also show that N fixation acclimates to growing temperature (hence its range of optimal temperatures), particularly in our two tropical symbioses. Surprisingly, optimal temperatures were 5.2 °C higher for N fixation than for photosynthesis, suggesting that plant carbon and N gain are decoupled with respect to temperature. These findings may help explain why N-fixing tree abundance is highest where annual maximum temperatures are >35 °C (ref. ) and why N-fixing symbioses evolved during a warm period in the Earth's history . Everything else being equal, our findings indicate that climate warming will probably increase N fixation, even in tropical ecosystems, in direct contrast to past projections .