The localization and propagation of rifting between Arabia and Somalia are investigated by assessing the deformation geometry and kinematics at different scales between the eastern Gulf of Aden and the Gulf of Tadjoura, using bathymetric, magnetic, seismological, and structural evidence. Large‐scale, southwestward propagation of the Aden ridge, markedly oblique to the Arabia‐Somalia relative motion vector, began about 30 Myr ago between the Error and Sharbithat ridges. It was an episodic process, with stages of rapid propagation, mostly at rates >10 cm/yr, interrupted by million year pauses on transverse discontinuities coinciding with rheological boundaries between different crustal provinces of the Arabia‐Somalia plate. The longest pause was at the Shukra‐El Sheik discontinuity (≈45°E), where the ridge tip stalled for ≈13 Myr, between ≈17 and ≈4 Ma. West of that discontinuity, rifting and spreading took place at an azimuth (≈N25°±10°E) and rate (1.2±0.3 cm/yr) different from those of the global Arabia‐Somalia motion vector (≈N39°, ≈1.73 cm/yr), implying an additional component of movement (N65°±10°E, 0.7±0.2 cm/yr) due to rotation of the Danakil microplate. At Shukra‐El Sheik, the typical oceanic ridge gives way to a narrow, WSW trending axial trough, resembling a large fissure across a shallow shelf. This trough is composed of about eight rift segments, which result from normal faulting and fissuring along N110°–N130°E trends. All the segments step to the left southwestward, mostly through oblique transfer zones with en échelon normal faults. Only two segments show clear, significant overlap. There is one clear transform, the Maskali fault, between the Obock and Tadjoura segments. The latter segment, which encroaches onland, is composed of two parallel subrifts (Iboli, Ambabbo) that propagated northwestward and formed in succession. The most recent, southwestern subrift (Ambabbo) represents the current tip of the Aden ridge. We propose a mechanical model in which the large‐scale propagation of the ridge followed a WSW trending zone of maximum tensile stress, while the small‐scale propagation of its NW trending segments was dictated by the orientation of that stress. Oblique propagation was a consequence of passive lithospheric necking of the Arabia‐Somalia plate along its narrow section, in map view, between Socotra and the kink of the Red Sea‐Ethiopian rift, above the Afar plume. Individual ridge segments oriented roughly perpendicular to plate motion, like lithospheric cracks, were forced to jump southward because of confinement within the necking zone. Self‐sustaining, plate‐scale necking may explain why the Aden ridge did not connect with the Red Sea through Bab El Mandeb but continued straight into Afar.