Earth's continental crust is dominantly made of buoyant, felsic igneous material (granitoids), that were ultimately extracted from the mantle as a result of Earth's differentiation. Since felsic melts are not in chemical equilibrium with the mantle, they can originate either from melting of older crustal lithologies, or from differentiation of a primitive mantle melt; only the latter case will contribute to crustal growth. To understand the mechanisms of continental crust growth and differentiation through time, it is therefore necessary to unravel the respective contribution of these two different mechanisms in the genesis of granitoid suites. In modern Earth, granitoids are chiefly generated in convergent plate boundaries (subduction and collision). This paper examines the granitic suites in a late-collision environment, the Variscan French Massif Central (FMC), and compares them with the suites found in an oceanic arc. We therefore describe, and compare, two end-members sites of granite generation.In the FMC, several main types of granites are described. Muscovite and Cordierite bearing Peraluminous Granites (resp. MPG and CPG) contain large amounts of inherited zircons, and their chemistry demonstrates that their sources were older crustal material (resp. Metasediments and metaigneous). On the other hand, Potassic Calc-alkaline Granites (KCG), associated to potassic diorites (vaugnerites) do not contain inherited zircons, and ultimately derive from the vaugnerites. The vaugnerites in turns form by partial melting of a mantle contaminated by the regional crust. Therefore, although they are isotopically similar to the crust, the KCG are net contributors to crustal growth. Thus we conclude that although late-orogenic settings are dominated by crustal melting and recycling, they may be sites of net crustal growth, even though this is not visible from isotopes only. In contrast, arc granitoids are purely or almost purely mantle derived. However, the preservation potential of arcs is much smaller than the preservation of late-orogenic domains, such that at the scale of a whole orogenic belt, late-orogenic magmatism is probably as important as arc magmatism.Finally, we speculate that the situation may have been similar in the Archaean, or even more skewed towards late-orogenic sites (or similar environments, dominated by melting of a altered mafic protocrust), owing to the hotter mantle and less stable subductions during that period.