The interannual variations of the carbon dioxide system and air‐sea CO2 fluxes are analyzed in the southwestern Indian Ocean from both in situ (Océan Indien Service d'Observations cruises in 1998 and 2000) and simulated oceanic CO2 fugacity dissolved inorganic carbon, total alkalinity, and nutrients. During austral summer of 1998, the ocean was warmer from 20°S to 60°S. In the Subtropical Zone (20°S–37°S), the warming, associated with the subtropical dipole pattern, creates an oceanic CO2 source around 2 mmol m−2 d−1 in January 1998 where all previous observations, included in 2000, indicated that this region was a small sink in summer. In the Sub‐Antarctic and Polar Front Zones (37°S–50°S), the interannual signal is not clearly detected because of the complex coupling between the horizontal and vertical dynamics and the biological activity that creates very high mesoscale variability in each summer. For the austral region, south of the Polar Front, we observe large variability in the CO2 sources and sinks between summer 1998 and 2000. In the Seasonal Ice Zone (south of 58°S) the interannual variation of the CO2 flux was mainly controlled by a warmer ocean, from 1.1°C, in summer 1998 (CO2 source of 3.2 mmol m−2 d−1) and an enhanced primary production occurring during summer 2000 (CO2 sink of −3.8 mmol m−2 d−1). In the Permanent Open Ocean Zone (POOZ, latitude 50°S–57°S), despite the half‐degree warmer sea surface in 1998, we observe lower (−15 μatm (1 atm = 105 × 1.01325 N m−2) on average) during this period, and consequently, we estimate a double of the oceanic CO2 sink in this region in 1998 than 2000, from −3.2 mmol m−2 d−1 in 1998 to −1.3 mmol m−2 d−1 in 2000. The strong oceanic CO2 sink in 1998 is associated with a dramatic decrease of nutrients (silicates less than 2 μmol kg−1). The enhanced biological activity in summer 1998, probably diatoms bloom, is also apparent in Sea‐viewing Wide Field‐of‐view Sensor (SeaWiFS) chlorophyll a estimates. Using a one‐dimensional biogeochemical model applied in the POOZ and SeaWIFS time series, we predict a stronger oceanic sink during summer season than the winter, which is contrary to previous studies. The model suggests that biological activity controls most of the sink anomaly in summer 1998 (for about 70%) and dominates the warming effect.