The minerals constituting the Earth's upper mantle are nominally anhydrous silicates (NAMs). However they do contain hydrogen as a trace element, decorating point defects in their crystalline structure. Experimental petrology and mineralogy have quantified the maximum concentration under several compositional and thermodynamic conditions, but systematic studies on the hydrogen concentration in minerals from mantle-derived rocks have only recently been carried out. Here, we have compiled the distribution of hydrogen in upper mantle peridotite xenoliths, from which several conclusions can be drawn. NAMs from peridotite xenoliths contain a few ppm wt. H2O in their structure. From the current database, the hydrogen concentrations in olivine regularly increase with increasing depth. The amount of hydrogen in NAMs from peridotite xenoliths from subduction contexts is not higher than in other geological context for similar temperature and pressure conditions. The highest hydrogen concentrations is found in peridotitic olivines from cratonic mantle, and are likely due to the depth of origin. The increasing hydrogen concentration in olivine with increasing depth is likely controlled by the increase of H partitioning into olivine at the expense of orthopyroxene as imposed by a decrease in Al content in opx with depth. However, the sparse data could also indicate that the bulk hydrogen concentration slightly increases with depth > 150 km. In this case, it would suggest, locally (Udachnaya for example), a possible increase in water fugacity due to fluid saturation. Even if the most abundant mineral in mantle rocks is olivine, the bulk hydrogen concentration in peridotites is controlled by the amount of hydrogen stored in pyroxenes. However, hydrogen concentration in olivine remains crucial for consequences on physical properties such as rheology and electrical conductivity. Kinetics of hydrogen transport is reviewed and hydrous melt/fluid percolation appears necessary to homogenize the hydrogen distribution at km-scale. Sampling of natural rock specimens is currently biased (e.g., in favor of optically attractive samples) and needs improvement, which will be achieved by increasing sample diversity (all type of grain sizes, lithologies, geological settings) and size of rock samples, as well as advances in analytical techniques. Acquisition of high quality data will be achieved by studying the co-existing minerals in mantle specimens, exploring each sample by linear and mapping measurements, and using appropriate FTIR calibrations for polarized and unpolarized radiation.