Abstract Decay of the Greenland Ice Sheet (GIS) figures prominently in climate change predictions, but direct carbon cycle feedbacks are poorly constrained. Here, we show that melting of the GIS yields a previously unknown flux of CO 2 to the atmosphere, and we use a reactive-transport model to examine if, how, and to what extent this flux will change as the ice sheet decays in a warmer world. Water emerges from beneath the Isunnguata and Russell Glaciers in West Greenland with CO 2 partial pressures (pCO 2 ) 3 - 10× supersaturated with respect to atmospheric equilibrium. During downstream transport in the Akuliarusiarsuup Kuua River, mineral weathering sequesters 75% of the excess CO 2 as HCO 3 − – a carbon sink on human timescales – and the remaining 25% evades to the atmosphere. Scaled to all rivers draining the GIS, the evasion flux of 0.11 ± 0.03 Tg C/year compares to fluxes reported for other rivers draining Precambrian shield crystalline rocks and having similar dissolved carbonate systematics. This flux is insufficient to influence modern-day atmospheric CO 2 levels, and we find that higher meltwater discharge alone will cause only moderate future increases. However, more substantial increases could occur if meltwaters intersect basal ice known to have elevated pCO 2 values. Worst-case model scenarios yield evasion fluxes of 100 ± 20–170 ± 40 Tg C/year by 2100. These atmospheric CO 2 inputs surpass those for Arctic Lakes and would augment by up to ~ 25% those predicted for permafrost thaw. Our findings suggest that positive feedbacks linking greenhouse gas emissions, Arctic climate change, and global warming may be stronger than previously realized.