The Rio Solimões/Amazonas (Amazon River) and its major tributaries have been analyzed for U-series nuclides. 238U-234U -230Th-226Ra disequilibria have been measured in the dissolved (<0.2 μm) and suspended loads (>0.2 μm) as well as bed sands. U-series disequilibria are closely related to major and trace element compositions and therefore reflect elemental fractionation during chemical weathering. Moreover, while the dissolved load records present-day weathering, suspended particles integrate the erosion history over much longer time scales (>100 ka). Lowland rivers are characterized by long time scales of chemical erosion (≥100 ka) resulting in a high weathering intensity. Moreover, exchange between suspended particles and the dissolved load may explain the U-series signature for these rivers. By combining U-series and Pb isotopes in suspended particles, we show that erosion in the Rio Madeira basin occurred as a multi-step process, whereby the pristine continental crust was eroded several hundreds of Ma ago to produce sediments that have then been integrated in the Cordillera by crustal shortening and are currently eroded. In contrast, recent erosion of a pristine crust is more likely for the Rio Solimões/Amazonas (<10 ka). The suspended particles of the rivers draining the Andes (Solimões/ Amazonas, Madeira) suggest time scales of weathering ranging between 4 and 20 ka. This indicates that suspended particles transported by those rivers are not stored for long periods in the Andean foreland basin and the tropical plain. The sediments delivered to the ocean have resided only a few ka in the Amazon basin (6.3 ± 1 ka for the Rio Amazonas at Óbidos). Nevertheless, a large fraction of the sediments coming out from the Andes are trapped in the foreland basin and may never reach the ocean. Erosion in the Andes is not operating in steady state. U-series systematics shows unambiguously that rivers are exporting a lot more sediments than predicted by steady-state erosion and that is a consequence of soil destruction greater than production. By relating this observation to the short time scales of weathering inferred for the Andes (a few ka), it appears that the erosion regime has been recently perturbed, resulting in high denudation rates. A possible explanation would be the increase in precipitation less than 5 ka proposed by recent paleoclimatic studies. Our results indicate that erosion responds rapidly to high-frequency climatic fluctuations.