The silicon cycle within Gulf Stream warm-core ring 82-B was examined before and after spring stratification using silicon isotope tracers. In April, prior to stratification, the ring was isothermal and isohaline to 400 m. Ambient silicic acid concentrations were ca 3 μmol l −1 and biogenic silica concentrations ranged from 80 to 100 nmol l −1 throughout this depth range. Specific production rates tended to be greatest at the surface and were vertically uniform below 20 m. Net silica production occurred from the surface to 300 m on 25 April, about three times deeper than has been observed before in the sea. The zone of net production shoaled to <150 m by 4 May. These depths were 2 to >4 times the 1% light depth, indicating that the diatom assemblage was producing substantial amounts of silica at depths where little or no photosynthesis was taking place. By June, a seasonal pycnocline had developed and silicic acid concentrations were depleted to ⩽0.2 μmol l −1 in the upper 20–30 m. Vertically integrated particulate silica concentrations in the upper 80 m had increased four-fold compared to those for the upper 110 m in April/May. Siliceous biomass was concentrated in the upper 40 m, with a subsurface maximum of ca 1000 nmol Si l −1 within the seasonal pycnocline. The upper 80 m remained a zone of net silica production, but net dissolution was occurring at all depths >100 m. The average specific production rate of biogenic silica (0.0098 h −1) was five times less than observed during April/May (0.041 h −1), suggesting slower diatom division rates in the stratified water column. The decrease in specific production rates following stratification was nearly offset by the increased particulate silica levels such that the vertically integrated production rate of biogenic silica for the upper 80 m was essentially unchanged from that observed for the upper 110 m during April/May. The percentage of silica production that could be supported by in situ dissolution increased from ⩽30% for the upper 110 m during April/May to 68% for the upper 80 m in June. This transition from a “new production” cycle to a “regenerated production” cycle is the first observation of a seasonal trend in the silicon cycle in the surface ocean. Furthermore, while net silica production occurred throughout the upper 110–300 m in April, net silica production in June was confined to the relatively nutrient-rich waters below seasonal pycnocline, wih a near balance between silica production and dissolution in the oligotrophic watersabove (average production:dissolution = 1.03). These changes in the production:dissolution ratio and the increased reliance on new nutrients with depth imply that the cycling of particulate silica in June fit a two-layer model of stratified oceanic systems where nearly complete and rapid recycling of materials occurs within a nutrient-poor surface layer with a zone of net production below.