For the North Sea, a semienclosed shelf sea in the northeastern North Atlantic, the seasonal and annual CO2 air‐sea fluxes (ASF) had been estimated for 2001 and 2002 in earlier work. The underlying observations, ΔpCO2, salinity, and temperature had been combined with 6‐hourly wind data derived from ERA40 reanalysis. In order to assess the impact of different wind data products on the computation of CO2 ASF, we compared ERA40 wind data with coastDat data derived from the nonhydrostatic regional climate model COSMO‐CLM. From the four observational months September, November, February, and May all but the May data show higher wind speeds for coastDat than for ERA40, especially off the Norwegian, UK, and continental coasts. Largest differences occur in the northern offshore areas. The comparison with observed wind data supports this feature generally: At Helgoland, an island in the German Bight, and at the Belgium pile “Westhinder” the ERA40 data underestimate both, the coastDat data and the observations. Wind observations for two Norwegian North Sea platforms were available: At the northern station “Troll” off the Norwegian coast the coastDat data overestimate the observations in winter. At “Ekofisk” in the central North Sea the ERA40 data fit the observations well, while the coastDat data slightly overestimate the observational data in all months but in May. The corresponding CO2 ASF estimates show strongest deviations off the Norwegian coast. Using different bulk formulas for determining the net annual ASF resulted in differences due to different wind products of up to 34%.