We present the interpretation of a new set of closely spaced marine magnetic profiles that complements previous data in the northeastern and southwestern parts of the South China Sea (Nan Hai). This interpretation shows that seafloor spreading was asymmetric and confirms that it included at least one ridge jump. Discontinuities in the seafloor fabric, characterized by large differences in basement depth and roughness, appear to be related to variations in spreading rate. Between anomalies 11 and 7 (32 to 27 Ma), spreading at an intermediate, average full rate of ≈50 mm/yr created relatively smooth basement, now thickly blanketed by sediments. The ridge then jumped to the south and created rough basement, now much shallower and covered with thinner sediments than in the north. This episode lasted from anomaly 6b to anomaly 5c (27 to ≈16 Ma) and the average spreading rate was slower, ≈35 mm/yr. After 27 Ma, spreading appears to have developed first in the eastern part of the basin and to have propagated towards the southwest in two major steps, at the time of anomalies 6b-7, and at the time of anomaly 6. Each step correlates with a variation of the ridge orientation, from nearly E-W to NE-SW, and with a variation in the spreading rate. Spreading appears to have stopped synchronously along the ridge, at about 15.5 Ma. From computed fits of magnetic isochrons, we calculate 10 poles of finite rotation between the times of magnetic anomalies 11 and 5c. The poles permit reconstruction of the Oligo-Miocene movements of Southeast Asian blocks north and south of the South China Sea. Using such reconstructions, we test quantitatively a simple scenario for the opening of the sea in which seafloor spreading results from the extrusion of Indochina relative to South China, in response to the penetration of India into Asia. This alone yields between 500 and 600 km of left-lateral motion on the Red River-Ailao Shan shear zone, with crustal shortening in the San Jiang region and crustal extension in Tonkin. The offset derived from the fit of magnetic isochrons on the South China Sea floor is compatible with the offset of geological markers north and south of the Red River Zone. The first phases of extension of the continental margins of the basin are probably related to motion on the Wang Chao and Three Pagodas Faults, in addition to the Red River Fault. That Indochina rotated at least 12° relative to South China implies that large-scale “domino” models are inadequate to describe the Cenozoic tectonics of Southeast Asia. The cessation of spreading after 16 Ma appears to be roughly synchronous with the final increments of left-lateral shear and normal uplift in the Ailao Shan (18 Ma), as well as with incipient collisions between the Australian and the Eurasian plates. Hence no other causes than the activation of new fault zones within the India-Asia collision zone, north and east of the Red River Fault, and perhaps increased resistance to extrusion along the SE edge of Sundaland, appear to be required to terminate seafloor spreading in the largest marginal basin of the western Pacific and to change the sense of motion on the largest strike-slip fault of SE Asia.