Plant root-derived carbon (C) inputs (Iroot) are the primary source of C in mineral bulk soil. However, a fraction of Iroot may lose directly (Iloss, e.g., via rhizosphere microbial respiration, leaching and fauna feeding) without contributing to bulk soil C pool. This loss has never been quantified, particularly at global scale, inhibiting reliable estimation of soil C dynamics. Here we integrate three observational global datasets including radiocarbon content, allocation of photosynthetically assimilated C, and root biomass distribution in 2,034 soil profiles to quantify Iroot and its contribution to the bulk soil C pool. We show that global average Iroot in the 0-200 cm soil profile is 3.5 Mg ha-1 yr-1, ~80% of which (i.e., Iloss) is lost rather than entering bulk soil. If ignoring Iloss, bulk soil C turnover will be incorrectly estimated to be four times faster. This can explain why Earth system models (in which all Iroot enters bulk soil C pools) predict much faster soil C turnover than radiocarbon-constrained estimates. Iroot decreases exponentially with soil depth, and the top 20 cm soil contains >60% of total Iroot. Actual C input to bulk soil (i.e., Iroot – Iloss) shows a similar depth distribution to Iroot. We also map Iloss and its depth distribution across the globe. Our results demonstrate the global significance of direct C losses which limit the contribution of Iroot to bulk soil C storage; and provide spatially explicit data to facilitate reliable soil C predictions via separating direct C losses from total root-derived C inputs.