Recent models that attempt to explain the formation of an axial topographic high at fast spreading ridges invoke the presence of a narrow, low-density, low-viscosity melt conduit to buoyantly support the high. Mantle upwelling beneath mid-ocean ridges must be highly focused to create the narrow melt zone required by these buoyancy models. We show that an alternative model in which extensional stresses in the upper crust are relieved by diking predicts uplift at the ridge and can match the shape and magnitude of the axial high. In this model, the extensional stresses deepen toward the ridge axis creating dynamic moments within the crust that are balanced by the axial topography. Along-axis variations in the height and width of the axial high may be caused by changes in the depth and width of the axial magma chamber or flexural rigidity of the lithosphere. This model is similar to dynamic models of median valley formation at slow spreading ridges in that the axial topography is uncompensated and supported by extensional stresses in the lithosphere and asthenosphere. In addition to being able to explain the axial topography observed at fast spreading ridges, this model is consistent with either broad (passive) or narrow (dynamic) upwelling at mid-ocean ridges. The change from median valley to axial high is caused by differences in strength of the axial lithosphere and rate of thickening of the plate. ¿ 1998 American Geophysical Union