Despite that no change is observed in the δ13C of sediments and upper mantle rocks over the Earth's history, the carbon isotope cycle is probably not at steady state. The δ13C of -5? of the upper mantle is different from that of the subducted carbon, which is close to -1?. Indeed, most of the subducted carbon is made of carbonates: 85% of subducted sedimentary carbon plus carbonates formed by low temperature hydrothermal alteration of oceanic crust. Since no long term evolution is observed for the mantle and sedimentary rocks, the net subduction flux of heavy carbon requires to be compensated by a deep mantle process. End-member models for the carbon isotope cycle at the Gyr scale suggest that it could be a flux from a deep primitive reservoir or the segregation of subducted oceanic crust at the core-mantle boundary.