The northeastern extension of the Altyn Tagh fault, China, beyond the apparently truncated Qilian Shan has been controversial since the recognition of this large left-slip fault. We propose that the Altyn Tagh fault may have been contiguous with the east-northeast to northeast-trending, currently inactive (but apparently active during the Cenozoic) Alxa–East Mongolia fault; the latter truncated the Beishan orogen to the northwest and the Inner Mongolia orogen to the southeast. The Beishan orogen can be correlated with the Inner Mongolia orogen on the basis of the recognition of three offset petrotectonic units and two offset sutures. This correlation requires 400 ± 50 km of left-lateral offset along the Alxa–East Mongolia fault, identical to the slip along the northern segment of the Altyn Tagh fault inferred by P. Molnar and P. Tapponnier. We suggest a two-stage model for fault evolution: during the first stage, 400 km of displacement separated the Inner Mongolia orogen from the originally contiguous Beishan orogen along a continuous Altyn Tagh–Alxa–East Mongolia fault. During the second stage, the Alxa–East Mongolia fault became inactive, and additional offset along the Altyn Tagh fault has been mainly accommodated by shortening of the Qilian Shan and the Qaidam basin. We infer that the first stage of movement along the Altyn Tagh–Alxa–East Mongolia fault system may have begun around Oligocene time; the region transformed to the second stage of fault movement ca. 13–16 Ma, probably because of thinning of the lithospheric mantle beneath northern Tibet. The proposed model suggests that the Derbur fault in northernmost China and beyond is an extension of the Alxa–East Mongolia fault, and has a history of strike-slip movement. The nearly 400 km of offset on the Alxa–East Mongolia fault may have been accommodated mainly by subduction in the Sea of Okhotsk.