A series of San Carlos olivine aggregates, sintered at high pressure and high temperature, with two different porosities (around 1 and 10%) and grain sizes (1–5 μm and 0–38 μm) were reacted at 300 °C and 500 bars in the presence of pure water. The reaction progress was monitored magnetically and the composition and distribution of the reaction products were analyzed at the end of each experiment. Brucite formation mainly occurred at the aggregate surface as a result of both congruent olivine dissolution and aqueous Mg and Si buffering by the reaction products, i.e. brucite and lizardite. The measured reaction progress did not exceed 2.6% after 290 d, which strongly contrasts with previous studies performed on San Carlos olivine powders (i.e., isolated grains in aqueous solution). Hence, limited water transport through the intergranular region of the aggregate drastically decreased the olivine surface area accessible to water and thus slowed down the whole serpentinization process. When extrapolated to peridotite relevant olivine grain sizes, our experimental results indicate that the water diffusion rate will become so slow that the first layer of primary minerals exposed to water within a mesh structure must fully react before the next mineral layer starts reacting (“layer by layer” mechanism). This type of reaction-transport mechanism is obviously not consistent with the micro-scale serpentine distribution in the mesh of oceanic peridotite samples, therefore additional water transport pathways are required. Cracks formed under extensional thermal stresses are good candidates since, in comparison to grain boundary or reaction-induced fractures, they are wide enough to drastically enhance water transport in oceanic peridotites and therefore account for the observed textures. The ‘layer by layer’ mechanism inferred here can only set a lower time bound for serpentinization completion. Assuming a mesh size of 1 mm and an initial grain size of 100 μm and considering a temperature ranging from 100 to 300 °C with permanent water saturation, completion should take place within 100–1000 yr. Surprisingly, this duration represents only 1 to 10% of the estimated timescale of the natural serpentinization process, emphasizing the central role played by water availability in the natural reaction process.