Deforestation to establish monocrops in the tropics is causing massive reductions in ecosystem C stocks. Amazonia is a principal target of this process, owing to the expansion of the agribusiness frontier throughout the transition with the Cerrado biome, the zone known as the "Arc of Deforestation". In this vast contact region between the two largest South American biomes, conversion of primary forest to soybean and pasture systems has led to the deforestation of nearly five million hectares since 1980. Despite this, we lack precise understanding of the effects of land use on ecosystem C stocks and pools in this region. Addressing this knowledge gap is crucial to improve predictions and fit models for different land use scenarios in Amazonia. To reduce uncertainty regarding the magnitude of the impacts of deforestation on the C cycle, we evaluated ecosystem C stocks in contrasting land-use systems across a topographically, climatically, and edaphically near-homogeneous landscape in southern Amazonia. We investigated the soil, litter, fine root and aboveground biomass (AGB) C stocks of soybean plantations and compared them to those of remnant native forests and rubber plantations; the latter is considered a priori as a cropping system with low impact on the C cycle. We found that the conversion of native forest to soybean plantation caused a 130.5 Mg C ha-1 loss, about threefold higher than the loss when forest is converted to rubber plantations, 48.5 Mg C ha-1. While 30-year old rubber plantations had recovered 84% of forest carbon stocks, all plantation types induced sustained losses of at least onethird of the original soil carbon. Fine root allocation changed sharply in the two crops following conversion, indicating an alteration in plant nutrient dynamics. Our results show that perennial and annual monocrops have very different impacts on the C cycle, which need to be accounted for in carbon-climate models as well as in public policies regulating land use in Amazonia. Our results show that while silviculture has the potential to restore most of the above-ground C stocks of previously forested areas, neither silviculture nor conventional agriculture may ever restore Amazon soil C stocks once they become vulnerable and oxidized after deforestation. If such conversion-driven soil carbon losses were scaled across the Amazon they would induce a cumulative loss of more than 5 Pg in soil carbon by 2050.