Atmospheric C sequestration in agricultural soils is viewed as one of the most promising negative emission technologies currently available. Nonetheless, it remains unclear how strongly soil organic carbon (SOC) stocks respond to agricultural practices, especially for subsoil. Here, we assess the SOC storage potential in croplands and how the presence of temporary grasslands (TG) in the crop rotation affects SOC stocks. We developed a new approach to correct for bias in bulk density (BD) induced by sampling conditions and land-use effects with a datadriven model to predict the BD of fine soil (<2 mm) for reference condition. Using 54 permanent grassland and cropland sites with various proportions of TG from a monitoring network in Switzerland, we showed that SOC stock differences down to 50-cm depth between cropland and permanent grasslands (maximum: 3.0 ± 0.8 kg C m-2) depend on the TG proportion in the crop rotation, regardless of clay content and pH. An increase of the TG proportion by 10% would induce a SOC gain of 0.40 ± 0.13 kg C m-2. The responses of topsoil (0–20 cm) and subsoil (20–50 cm) SOC stocks to TG proportion were linear and equivalent. The effect of TG on SOC storage would have been underestimated by 58% without accounting for subsoil stocks response and by 16% without BD corrections. The conversion of all croplands to permanent grasslands in the study region would potentially store a quantity of SOC equivalent to the anthropogenic greenhouse gas emissions generated by the same region during one year. Although the potential of agricultural soils as negative emission technology is relatively modest compared to former expectations, the findings demonstrate the potential to manage SOC and its associated ecosystem services at large scales and down to deep soil layers.