Miscanthus is a low input energy crop suitable for low fertility marginal arable land and thought to provide carbon sequestration in soil. We analysed a long-term field experiment (14-year) to determine whether differences in genotype, growth habit, and root distribution affected soil carbon spatially under different Miscanthus genotypes. Soil cores were taken centrally and radially to a depth of 1 metre, and divided into six vertical segments. Total root length (TRL), root dry matter (RDM) and δ13C signature of soil organic carbon (SOC) were measured directly, and root length density (RLD), fractions of Miscanthus-derived soil organic C (SOCM), and residual soil carbon (SOCORIG) were calculated. Genotype was found to exhibit a statistically significant influence on spatial allocation of SOC. Grouping varieties into ‘tuft-forming’ (T) and ‘non-tuft-forming’ (NT) phenotypes revealed that respective groups accumulated similar amounts of RDM over 14 years (11.4 ±3.3 vs. 11.9 ±4.8 Mg ha-1, respectively). However, phenotype T allocated more carbon to roots in the subsoil than NT (33% vs. 25%). Miscanthus genotypes sequestered between 4.2 and 7.1 g C4-SOC kg-1 soil over the same period, which was more than the average loss of C3-derived SOC (3.25 g kg-1). Carbon stocks in the ‘A horizon’ under Miscanthus increased by about 5 Mg ha-1 above the baseline, while the net increase in the subsoil was marginal. Amounts of Miscanthus root C in the subsoil were small (1.2 - 1.8 Mg C ha-1) but could be important for sustainable sequestration as root density (RLD) explained a high percentage of SOCM (R2 = 0.66).