Abstract: Optimal foraging theory predicts that diving predators should maximise their energy intake while minimizing the costs of their foraging activities at depth. This hypothesis was tested in two breath-hold divers that foraged in the pelagic waters off Kerguelen Island, southern Indian Ocean. Macaroni penguins (Eudyptes chrysolophus) feed predominately on crustaceans at moderate depths (ca. 50 m), while king penguins (Aptenodytes patagonicus) feed on mesopelagic fish at deep depths (ca. 100–150 m). We instrumented 22 penguins with time-depth recorders during the breeding season of the 2014 southern summer, just before they went to sea. Birds were recaptured after one to five foraging trips and the dive data were analysed on the level of single dive cycles and entire dive bouts. Both species adjusted their diving behaviour and modulated their foraging activity in accordance with their experience in their previous dives. Foraging activity during a dive (in terms of number of wiggles per dive) was greater if birds targeted the same depth as the previous dive. The penguins spent less time at the surface between two prey capture attempt dives than when a prey capture attempt dive was preceded by a non-prey capture attempt dive. Dive bout analyses showed that foraging activity increased with bout duration, while transit time during dives decreased with bout duration. Our findings are in general agreement with predictions based on optimal foraging models. The results suggest that penguins anticipated the depth of their next dive based on their encounter of a prey patch in the previous dive and optimized their chances of feeding success by spending minimal time at the surface and in transit to the patch depth. Once penguins detected a prey patch, these behavioural adjustments allowed them to rapidly return to that patch, increasing their chances of feeding again before the prey disperses.