We systematically investigated near-axis subsidence on the ridge flanks of intermediate and fast spreading mid-ocean ridges using bathymetric data from well-surveyed portions of the Southeast Indian Ridge (spreading at 72--76 mm/yr), the northern East Pacific Rise (91--96 mm/yr), and the southern East Pacific Rise (144 mm/yr). In all three regions, the mean subsidence rate of young (<1--1.4 Ma) seafloor is less than 220 m/m.y.1/2. The distribution of individual estimates shows a distinct peak at 180--220 m/m.y.1/2 with few profiles having subsidence rates greater than 275 m/m.y.1/2. The observed subsidence rates are significantly lower than both the worldwide average (~350 m/m.y.1/2) and subsidence rates observed for older lithosphere at the same ridge segments. Intense hydrothermal circulation at the ridge axis can result in low subsidence rates on the adjacent ridge flanks provided the vigorous flow is confined to the immediate vicinity (<~5 km) of the axis. According to our model, the extremely vigorous hydrothermal circulation ceases off-axis, and conductive heat flow becomes a primary mechanism of vertical heat transport on the ridge flanks. The very low geothermal gradient within the cooled portion of the uppermost lithosphere retards conductive cooling, and the cooled area needs to be heated from below before a geothermal gradient can be established which permits significant heat to be conducted out of the lithosphere. As a result, subsidence in very young (~0.1--1 Ma) lithosphere is suppressed. A simple one-dimensional thermal model with a Nusselt number parameterization was used to estimate the effect of hydrothermal circulation. An upper layer with a high Nusselt number and the half-space that it overlies are initially at a temperature of 1300 ¿C and the surface is maintained at 0 ¿C. After 0.1 m.y. of cooling (about 3.5 to 7 km from the spreading axis), the Nusselt number of the top layer is set to 1 so that normal conduction is simulated in the cooled layer. We used an explicit finite difference method to solve for the temporal changes in temperature with depth. This model produces subsidence rates in the range that we observe for Nu in the range of 15--30. Isotherms resulting from the modeling imply rapid lithospheric thickening very near the axis, which is incompatible with most current models for the formation of the bathymetric axial high observed at fast spreading mid-ocean ridges. ¿ 2001 American Geophysical Union