The thickness of the crust created at ocean spreading centres depends on the spreading rate and melt production in the mantle. It is ∼5–8 km for a crust formed at slow and fast spreading centres and 2–4 km at ultra-slow spreading centres away from hotspots and mantle anomalies. The crust is generally thin at the fracture zones and thick beneath hotspots and large igneous provinces. Here we present results for the crust generated at the fast Wharton spreading centre 55–58 Ma ago using seismic reflection and refraction data. We find that the crust over a 200 km segment of the Wharton Basin is only 3.5–4.5 km thick, the thinnest crust ever observed in a fast spreading environment. A thin crust could be produced by the presence of depleted and/or cold mantle. Numerical simulations and recent laboratory experiments studying the impact of a hot plume under a lithosphere show that a curtain of weak cold downwellings of depleted mantle material is likely to develop around the edges of the hot plume pond. Because of a strongly temperature-dependent viscosity of lithospheric material, the hotter, therefore less viscous, bottom of the lithosphere can be mobilized by an impinging plume. If sampled by a spreading centre, the locally cold and depleted mantle should result in low production of melt. We suggest that the observed thin crust in the Wharton Basin is likely to have been formed by the interaction between the Kerguelen mantle plume and the Wharton spreading centre ∼55 Ma ago.