The addition of iron to high-nutrient, low-chlorophyll regions induces phytoplankton blooms that take up carbon. Carbon export from the surface layer and, in particular, the ability of the ocean and sediments to sequester carbon for many years remains, however, poorly quantified. Here we report data from the CROZEX experiment in the Southern Ocean, which was conducted to test the hypothesis that the observed north-south gradient in phytoplankton concentrations in the vicinity of the Crozet Islands is induced by natural iron fertilization that results in enhanced organic carbon flux to the deep ocean. We report annual particulate carbon fluxes out of the surface layer, at three kilometres below the ocean surface and to the ocean floor. We find that carbon fluxes from a highly productive, naturally iron-fertilized region of the sub-Antarctic Southern Ocean are two to three times larger than the carbon fluxes from an adjacent high-nutrient, low-chlorophyll area not fertilized by iron. Our findings support the hypothesis that increased iron supply to the glacial sub-Antarctic may have directly enhanced carbon export to the deep ocean. The CROZEX sequestration efficiency (the amount of carbon sequestered below the depth of winter mixing for a given iron supply) of 8,600 mol mol(-1) was 18 times greater than that of a phytoplankton bloom induced artificially by adding iron, but 77 times smaller than that of another bloom initiated, like CROZEX, by a natural supply of iron. Large losses of purposefully added iron can explain the lower efficiency of the induced bloom(6). The discrepancy between the blooms naturally supplied with iron may result in part from an underestimate of horizontal iron supply.