The deployment of solar photovoltaic (PV) technology has consistently outpaced expectations over the last decade. However, long-term prospects for PV remain deeply uncertain, as recent global scenarios span two orders of magnitude in installed PV capacity by 2050. Here we systematically compile an ensemble of 1,550 scenarios from peer-reviewed and influential grey literature, including IPCC and non-IPCC scenarios, and apply a statistical learning framework to link scenario characteristics with foreseen PV outcomes. We show that a large portion of the uncertainty in the global scenarios is associated with general features like the type of organization, energy model, and policy assumptions, without referring to specific techno-economic assumptions. IPCC scenarios consistently project lower PV adoption pathways and higher capital costs than non-IPCC scenarios. We thus recommend increasing the diversity of models and scenario methods included in IPCC assessments to represent the multiple perspectives present in the PV scenario literature.