Long-term exposure of humans to air pollution enhances the risk of cardiovascular and respiratory diseases. A novel Global Exposure Mortality Model (GEMM) has been derived from many cohort studies, providing much-improved coverage of the exposure to fine particulate matter (PM2.5). We applied the GEMM to assess excess mortality attributable to ambient air pollution on a global scale and compare to other risk factors. Methods and results We used a data-informed atmospheric model to calculate worldwide exposure to PM2.5 and ozone pollution, which was combined with the GEMM to estimate disease-specific excess mortality and loss of life expectancy (LLE) in 2015. Using this model, we investigated the effects of different pollution sources, distinguishing between natural (wildfires, aeolian dust) and anthropogenic emissions, including fossil fuel use. Global excess mortality from all ambient air pollution is estimated at 8.8 (7.11–10.41) million/year, with an LLE of 2.9 (2.3–3.5) years, being a factor of two higher than earlier estimates, and exceeding that of tobacco smoking. The global mean mortality rate of about 120 per 100 000 people/year is much exceeded in East Asia (196 per 100 000/year) and Europe (133 per 100 000/year). Without fossil fuel emissions, the global mean life expectancy would increase by 1.1 (0.9–1.2) years and 1.7 (1.4–2.0) years by removing all potentially controllable anthropogenic emissions. Because aeolian dust and wildfire emission control is impracticable, significant LLE is unavoidable. Conclusion Ambient air pollution is one of the main global health risks, causing significant excess mortality and LLE, especially through cardiovascular diseases. It causes an LLE that rivals that of tobacco smoking. The global mean LLE from air pollution strongly exceeds that by violence (all forms together), i.e. by an order of magnitude (LLE being 2.9 and 0.3 years, respectively).