In grassland ecosystems, where soil water most frequently controls ecosystem processes, expected changes in precipitation and temperature may have dramatic effects on ecosystem dynamics. Previous observational studies have reported that aboveground net primary production (ANPP) in grasslands is very sensitive to changes in precipitation. Yet, we lack experimentally based evidence to support these observations. Further, most of the studies have focused solely on ANPP, neglecting belowground production (BNPP). This is an important gap in our knowledge, as BNPP represents 50% or more of total net primary production (NPP) in grasslands. Here, we present results from a 3-year water manipulation experiment (2008–2010) at two sites in the central grassland region of North America, USA. We were successful in changing the soil water content in our treatments, but these changes resulted in different, but significant responses in ANPP and BNPP at our two sites. At the shortgrass steppe, we found that neither NPP nor ANPP were sensitive to treatment precipitation, and although we found BNPP was sensitive to changes in treatment precipitation, the direction of the response varied between years. In contrast, ANPP was very sensitive to treatment precipitation on the mixed-grass prairie, whereas BNPP was insensitive. Based on our finding that two grassland ecosystems showed dramatically different above and belowground production responses to soil water manipulations, we cannot assume that predicted changes in climate will cause similar above- and belowground production responses. Second, our results demonstrated that sites within the same region may differ markedly in the sensitivity of ANPP to changes in growing season precipitation.