The pantropical orb web spider family Nephilidae is known for the most extreme sexual size dimorphism among terrestrial animals. Numerous studies have made Nephilidae, particularly Nephila, a model lineage in evolutionary research. However, a poorly understood phylogeny of this lineage, relying only on morphology, has prevented thorough evolutionary syntheses of nephilid biology. We here use three nuclear and five mitochondrial genes for 28 out of 40 nephilid species to provide a more robust nephilid phylogeny and infer clade ages in a fossil-calibrated Bayesian framework. We complement the molecular analyses with total evidence analysis including morphology. All analyses find strong support for nephilid monophyly and exclusivity and the monophyly of the genera Herennia and Clitaetra. The inferred phylogenetic structure within Nephilidae is novel and conflicts with morphological phylogeny and traditional taxonomy. Nephilengys species fall into two clades, one with Australasian species (true Nephilengys) as sister to Herennia, and another with Afrotropical species (Nephilingis Kuntner new genus) as sister to a clade containing Clitaetra plus most currently described Nephila. Surprisingly, Nephila is also diphyletic, with true Nephila containing N. pilipes+N. constricta, and the second clade with all other species sister to Clitaetra; this "Nephila" clade is further split into an Australasian clade that also contains the South American N. sexpunctata and the Eurasian N. clavata, and an African clade that also contains the Panamerican N. clavipes. An approximately unbiased test constraining the monophyly of Nephilengys, Nephila, and Nephilinae (Nephila, Nephilengys, Herennia), respectively, rejected Nephilengys monophyly, but not that of Nephila and Nephilinae. Further data are therefore necessary to robustly test these two new, but inconclusive findings, and also to further test the precise placement of Nephilidae within the Araneoidea. For divergence date estimation we set the minimum bound for the stems of Nephilidae at 40 Ma and of Nephila at 16 Ma to accommodate Palaeonephila from Baltic amber and Dominican Nephila species, respectively. We also calibrated and dated the phylogeny under three different interpretations of the enigmatic 165 Ma fossil Nephila jurassica, which we suspected based on morphology to be misplaced. We found that by treating N. jurassica as stem Nephila or nephilid the inferred clade ages were vastly older, and the mitochondrial substitution rates much slower than expected from other empirical spider data. This suggests that N. jurassica is not a Nephila nor a nephilid, but possibly a stem orbicularian. The estimated nephilid ancestral age (40-60 Ma) rejects a Gondwanan origin of the family as most of the southern continents were already split at that time. The origin of the family is equally likely to be African, Asian, or Australasian, with a global biogeographic history dominated by dispersal events. A reinterpretation of web architecture evolution suggests that a partially arboricolous, asymmetric orb web with a retreat, as exemplified by both groups of "Nephilengys", is plesiomorphic in Nephilidae, that this architecture was modified into specialized arboricolous webs in Herennia and independently in Clitaetra, and that the web became aerial, gigantic, and golden independently in both "Nephila" groups. The new topology questions previously hypothesized gradual evolution of female size from small to large, and rather suggests a more mosaic evolutionary pattern with independent female size increases from medium to giant in both "Nephila" clades, and two reversals back to medium and small; combined with male size evolution, this pattern will help detect gross evolutionary events leading to extreme sexual size dimorphism, and its morphological and behavioral correlates.