—Due to its size and high altitude, the growth of the Tibetan Plateau remains an enigma. Based on a synthesis of anterior collisions, paleoaltimetric data, geochemistry of ultrapotassic lava and their rare mantle enclaves, combined with a reinterpretation of tomographic data, we suppose that Tibet’s growth took take place in two main stages. Initially, the accretion of Gondwana terranes to the margin of South Asia, especially during the Early Triassic–Cretaceous period, resulted in the first episode of plateau growth, which affected an area of about 2/3 of the current plateau. We suppose that during the Late Cretaceous, the Tibetan crust reached a thickness of about 50–55 km, which is equivalent to an altitude of about 2500 to 3000 m, with local landforms that could have exceeded 4000 m. Another important consequence of these successive accretions was a strong metasomatism and a softening of the upper part of the Tibetan cover. The P wave low-velocity anomaly currently observed under the central part of Tibet would correspond not to a temperature anomaly but to a composition anomaly. From 50 Ma onwards, the convergence between India and Asia, estimated at about 1000 km on the Tibetan side, led to a shortening of the plateau by about 40%. We suppose that this additional shortening, which has led to the current thickness of the Earth’s crust of about 70 km and an average altitude of 4800 m, has been compensated by the reactivation of the continental slabs along the previous sutures and by the homogeneous shortening of the crust.