Abstract It has been suggested that all of the Earth’s subduction zones are capable of producing an M w ∼9 earthquake. Such a possibility has also been suggested for the Himalayan collisional arc. We examined recent and historical earthquakes in the Himalayan region and suggest that, in the past few hundred years, only great earthquakes (M w ∼8) have occurred in the region, the largest being the 1950 Assam earthquake with M w 8.6. Several aseismic ridges segment the Himalayan arc and these appear to have limited the extent of rupture in earthquakes of the past 200 years. There is also evidence of distributed deformation and out-of-sequence thrusting, where only part of the detachment participates in the accommodation of the convergence. We concluded that, although it is unlikely that an M w ∼9 earthquake could occur in the Himalaya, from the hazard point of view a great earthquake of M w ∼8 anywhere along the Himalayan arc could kill about one million people in the Himalaya and adjoining Indo-Ganga plains. As an example, a scenario developed for a hypothetical M w 8 earthquake occurring in the middle of the night at Mandi, close to the epicentre of the 1905 Kangra earthquake, predicts an estimated loss of one million human lives.