It is increasingly commonly suggested that grasslands are a perpetual sink for carbon, and that just maintaining grasslands will yield a net carbon sink. I examine the evidence for this from repeated soil surveys, long term grassland experiments and simple mass balance calculations. I conclude that it is untenable that grasslands act as a perpetual carbon sink, and the most likely explanation for observed grassland carbon sinks over short periods is legacy effects of land use and land management prior to the beginning of flux measurement periods. Simply having grassland does not result is a carbon sink, but judicious management or previously poorly managed grasslands can increase the sink capacity. Given that grasslands are a large store of carbon, and that it is easier and faster for soils to lose carbon that it is for them to gain carbon, it is an important management target to maintain these stocks.