The Xunhua, Guide and Tongren intermontane basin system in the NE Tibetan Plateau, situated near the Xining basin to the N and the Linxia basin to the E, is bounded by thrust fault-controlled ranges. These include to the N, the Riyue Shan, Laji Shan and Jishi Shan ranges, and to the S the northern West Qinling Shan (NWQ). An integrated study of the structural geology, sedimentology and provenance of the Cenozoic Xunhua and Guide basins provides a detailed record of the growth of the NE Tibetan Plateau since the early Eocene. The Xining Group (ca. 52–21 Ma) is interpreted as consisting of unified foreland basin deposits which were controlled by the bounding thrust belt of the NWQ. The Xunhua, Guide and Xining subbasins were interconnected prior to later uplift and damming by the Laji Shan and Jishi Shan ranges. Their sediment source, the NWQ, is constrained by strong unidirectional paleocurrent trends towards the N, a northward fining lithology, distinct and recognizable clast types and detrital zircon ages. Collectively, formation of this mountain–basin system indicates that the Tibetan Plateau expanded into the NWQ at a time roughly coinciding with Eocene to earliest Miocene continental collision between India and Eurasia. The Guide Group (ca. 21–1.8 Ma) is inferred to have been deposited in the separate Xunhua, Guide and Tongren broken foreland basins. Each basin was filled by locally sourced alluvial fans, braided streams and deltaic-lacustrine systems. Structural, paleogeographic, paleocurrent and provenance data indicate that thrust faulting in the NWQ stepped northward to the Laji Shan from ca. 21 to 16 Ma. This northward shift was accompanied by E–W shortening related to nearly N–S-striking thrust faulting in Jishi Shan after 11–13 Ma. A lower Pleistocene conglomerate (1.8–1.7 Ma) was deposited by a through-flowing river system in the overfilled and connected Guide and Xunhua basins following the termination of thrust activity. All of the basin–mountain zones developed along the Tibetan Plateau's NE margin since Indian–Tibetan continental collision may have been driven by collision-induced basal drag of old slab remnants in the manner of N-dipping and flat-slab subduction, and their subsequent sinking into the deep mantle.