Ever since the late 1960s when Warren Hamilton proposed that the great Cordilleran batholiths of the western Americas are the roots of volcanic arcs like the Andes and were generated by longstanding eastward subduction, most geologists have followed suit, despite the evergrowing recognition that many Cordilleran batholiths are complex, composite bodies that developed with intervals of intense shortening and exhumation between and during periods of magmatism. The Peninsular Ranges batholith of Southern and Baja California provides a superb place to unravel the complexities because there is a lot of data and because it is longitudinally composed of two parts: an older western portion of weakly to moderately deformed, low-grade volcanic and epizonal plutonic rocks ranging in age from ~128–100 Ma; and a more easterly sector of deformed amphibolite grade rocks cut by compositionally zoned, mesozonal plutonic complexes of the La Posta suite, emplaced from 99–86 Ma. While plutons of the La Posta suite are generally considered to be the product of continued eastward subduction, they are enigmatic, because they and their wall rocks were rapidly exhumed from as deep as 23 km and eroded during, and just after, their emplacement, unlike plutons in magmatic arcs, which are generally emplaced in zones of subsidence. Here we resolve the enigma with a model where westward-dipping subduction led to arc magmatism of the western sector, the Santiago Peak–Alisitos composite arc, during the period ~128–100 Ma. Arc magmatism shut down when the arc collided with a west-facing Early Cretaceous passive margin at about 100 Ma. During the collision the buoyancy contrast between the continental crust of the eastern block and its attached oceanic lithosphere led to failure of the subducting slab. The break-off allowed subjacent asthenosphere to upwell, adiabatically melt, and rise into the upper plate to create the large zoned tonalite–granodiorite–granite complexes of the La Posta suite. While compositionally similar to arc plutons in many respects, the examples from the Southern California and Baja segments of the batholith have geochemistry that indicates they were derived from partial melting of asthenosphere at deeper levels in the mantle than typical arc magmas, and within the garnet stability field. This is consistent with asthenosphere upwelling through the torn lower-plate slab. We identify kindred rocks with similar geological relations in other Cordilleran batholiths of the Americas, such as the Sierra Nevada, which lead us to suggest that slab failure magmatism is common, both spatially and temporally.