The processes that control water distribution in nominally anhydrous minerals from peridotites are twofold. Melt depletion will remove water while metasomatism can potentially add water to these minerals. These processes can lead to a wide range of outcomes in water contents, which in turn could play a role in mantle rheology and long-term cratonic root stability. To examine these complexities, water concentrations in minerals from well-characterized peridotites from the Udachnaya kimberlite in the central Siberian craton were analyzed by FTIR. The peridotites span a complete top to bottom cross-section of typical cratonic lithospheric mantle (2–7 GPa and 700–1400 C). Diffusion modeling of water content profiles across olivine grains shows that water loss during decompression is limited to the 100 lm rims of olivines; the cores preserved their mantle water contents. Water contents range from 6 to 323 ppm wt H2O in olivine, 28–301 ppm H2O in orthopyroxene (opx), 100– 272 ppm H2O in clinopyroxene (cpx) and 0–23 ppm H2O in garnet. Melting modeling cannot reproduce the high water contents of cratonic mantle peridotites and any potential partial melting trend must have been erased by later events. The water contents of minerals, however, are correlated with modal abundances of clinopyroxene and garnet, bulk rock FeO, TiO2 and SiO2 as well as with light and middle rare earth elements in clinopyroxene and garnet. These relationships are best interpreted as interaction of residual, melt-depleted peridotites with silicate melt, which produced modal and cryptic metasomatism. Importantly, the water enrichment in the Siberian cratonic mantle took place prior to kimberlite magmatism and eruption. Water addition by metasomatism occurred from pressures >4 GPa all the way to the base of the cratonic root below central Siberia, but was limited to shallower levels (<5 GPa) in the Kaapvaal cratonic lithosphere. The difference in olivine water contents at the deepest levels of the Kaapvaal (<5 ppm H2O) and Siberian (6–323 ppm H2O) cratonic roots may be linked to oxygen fugacity and resulting fluid speciation or, alternatively, to reaction with different metasomatic agents. Calculated viscosities for the deepest Udachnaya samples are similar to those inferred for the asthenosphere. If these xenoliths are representative of the deep cratonic lithosphere, water is not as important a parameter as previously thought in the strength of cratonic lithosphere, otherwise the cratonic root beneath Udachnaya would have been delaminated. Alternatively, the metasomatic xenoliths may not be representative of the Siberian cratonic root and kimberlites preferentially sample cratonic mantle lithosphere material located near, and metasomatized by, melt conduits, which served as channels for upward migration of water-rich melts and fluids including kimberlites. In that case, the cratonic root overall still may have relatively low water contents, which in addition to its less metasomatized (more refractory) and thereby buoyant nature, still play a role in making it strong enough to resist delamination by the surrounding asthenosphere.