We quantified the impacts of four years of nominal (within the historic range of variability) and extreme (1-in-100-year recurrence frequency) droughts on aboveground productivity with the International Drought Experiment (IDE), a coordinated, distributed network of 74 grassland and shrubland sites located on six continents across the globe. We expected that aboveground productivity, a key measure of terrestrial ecosystem functioning, would be impacted progressively (i.e., decline over time) as duration and severity of droughts increased. An alternative prediction is that acclimation may occur, whereby – productivity is either maintained or may even recover when drought is prolonged over multiple years. On average across all IDE sites and irrespective of whether drought was nominal or extreme, aboveground productivity declined significantly in the first of year drought, but the magnitude of loss in productivity did not change in years 2 to 4. Thus, we found evidence overall for acclimation to prolonged drought, and a similar acclimation response was observed with prolonged, nominal drought. Yet, when drought was extreme, progressive losses in ecosystem productivity was observed, with the largest losses observed with increasing drought severity in the third and fourth years of drought. Furthermore, the largest losses in productivity (>70%) were observed when drought was extreme for more than two consecutive years. Our results provide evidence for a strong interaction between drought severity and duration. When drought is within the historic range of variability, grassland and shrubland ecosystems have the potential to acclimate, but when extreme, ecosystems that historically were resistant may experience profound losses in functioning over time. This finding has important implications for terrestrial ecosystem functioning in the future, given forecasts for more severe and longer duration droughts with climate change - ecosystems may shift from being resistant to prolonged drought to experiencing catastrophic losses in ecosystem functioning.