Mega hydropower projects in tropical forests pose a major emergent threat to terrestrial and freshwater biodiversity worldwide. Despite the unprecedented number of existing, under-construction and planned hydroelectric dams in lowland tropical forests, long-term effects on biodiversity have yet to be evaluated. We examine how medium and large-bodied assemblages of terrestrial and arboreal vertebrates (including 35 mammal, bird and tortoise species) responded to the drastic 26-year post-isolation history of archipelagic alteration in landscape structure and habitat quality in a major hydroelectric reservoir of Central Amazonia. The Balbina Hydroelectric Dam inundated 3,129 km2 of primary forests, simultaneously isolating 3,546 land-bridge islands. We conducted intensive biodiversity surveys at 37 of those islands and three adjacent continuous forests using a combination of four survey techniques, and detected strong forest habitat area effects in explaining patterns of vertebrate extinction. Beyond clear area effects, edge-mediated surface fire disturbance was the most important additional driver of species loss, particularly in islands smaller than 10 ha. Based on species-area models, we predict that only 0.7% of all islands now harbor a species-rich vertebrate assemblage consisting of ≥80% of all species. We highlight the colossal erosion in vertebrate diversity driven by a man-made dam and show that the biodiversity impacts of mega dams in lowland tropical forest regions have been severely overlooked. The geopolitical strategy to deploy many more large hydropower infrastructure projects in regions like lowland Amazonia should be urgently reassessed, and we strongly advise that long-term biodiversity impacts should be explicitly included in pre-approval environmental impact assessments.