In the next 5 y, more than 10 thousand kilometers of roads will be built or improved in the Amazon. Well-designed projects can increase employment opportunities, reduce transport costs, and support regional development. However, roads will also drive deforestation, threatening biodiversity and ecosystem services, jeopardizing the welfare of indigenous people, and moving the biome toward irreversible shifts in vegetation. Data to support good decisions are remarkably scarce. Typical feasibility studies, where they exist, inadequately address environmental and social impacts and do not facilitate comparison across projects. This study contributes to informed decision-making by quantifying the environmental, social and economic effects of 75 planned projects. It demonstrates that fewer projects in carefully chosen locations would dramatically improve outcomes of all types.