Abstract. Recent years have seen a number of severe droughts in different regions around the world, causing agricultural and economic losses, famines and migration. Despite their devastating consequences, the Standardised Precipitation Index (SPI) of these events lies within the range of internal climate variability, which we estimate from simulations from the 5th phase of the Coupled Model Intercomparison Project (CMIP5). In terms of drought magnitude, regional trends of SPI over the last decades remain mostly inconclusive in observations and CMIP5 simulations, although Soil Moisture Anomalies (SMAs) in CMIP5 simulations hint at increased drought in a few regions (e.g. the Mediterranean, Central America/Mexico, the Amazon, North-East Brazil and South Africa). Also for the future, projections of meteorological (SPI) and agricultural (SMA) drought in CMIP5 display large uncertainties over all time frames, generally impeding trend detection. Analogue analyses of the frequencies rather than magnitudes of future drought display, however, more robust signal-to-noise ratios with detectable trends towards more frequent drought until the end of the 21st century in the Mediterranean, South Africa and Central America/Mexico. Other present-day hot spots are projected to become less drought-prone, or to display unsignificant changes in drought occurrence. A separation of different sources of uncertainty in drought projections reveals that for the near term, internal climate variability is the dominant source, while the formulation of Global Climate Models (GCMs) generally becomes the dominant source of uncertainty by the end of the 21st century, especially for agricultural (soil moisture) drought. In comparison, the uncertainty in Green-House Gas (GHG) concentrations scenarios is negligible for most regions. These findings stand in contrast to respective analyses for a heat wave indicator, for which GHG concentrations scenarios constitute the main source of uncertainty. Our results highlight the inherent difficulty of drought quantification and the uncertainty of drought projections. However, high uncertainty should not be equated with low drought risk, since potential scenarios include large drought increases in key agricultural and ecosystem regions.