The productivity of rainforests growing on highly weathered tropical soils is expected to be limited by phosphorus availability1. Yet, controlled fertilization experiments have been unable to demonstrate a dominant role for phosphorus in controlling tropical forest net primary productivity. Recent syntheses have demonstrated that responses to nitrogen addition are as large as to phosphorus2, and adaptations to low phosphorus availability appear to enable net primary productivity to be maintained across major soil phosphorus gradients3. Thus, the extent to which phosphorus availability limits tropical forest productivity is highly uncertain. The majority of the Amazonia, however, is characterized by soils that are more depleted in phosphorus than those in which most tropical fertilization experiments have taken place2. Thus, we established a phosphorus, nitrogen and base cation addition experiment in an old growth Amazon rainforest, with a low soil phosphorus content that is representative of approximately 60% of the Amazon basin. Here we show that net primary productivity increased exclusively with phosphorus addition. After 2 years, strong responses were observed in fine root (+29%) and canopy productivity (+19%), but not stem growth. The direct evidence of phosphorus limitation of net primary productivity suggests that phosphorus availability may restrict Amazon forest responses to CO2 fertilization4, with major implications for future carbon sequestration and forest resilience to climate change.