The authors compare wind speed retrieved from the Geosat altimeter, from two Special Sensor Microwave/Imager (SSM/I) microwave radiometers. The SSM/I F08 and SSM/I F10, and from the European Space Agency ERS-1 scatterometer. As ground truth, ship reports were used that provide a continuous time series of consistent measurements at large scale during the whole period covered by the three satellites, and TOGA TAO data in the tropical Pacific Ocean that are more accurate though more limited in geographical extent than ship wind speeds. It is evidenced that the Geosat wind speed retrieved using the Witter and Chelton algorithm is underestimated at high wind speed. The authors find that the SSM/I wind speeds retrieved by the Wentz algorithm are underestimated by more than 1 m s−1 with respect to the ship wind speeds in large regions at high latitudes, this effect being larger with SSM/I F10 than with SSM/I F08. The authors compare the ERS-1 wind speeds retrieved from the Cersat preliminary algorithm and from the ESA CMOD4 algorithm; while the former gives wind speeds consistent with the ship measurements, the latter is shown to overestimate low wind speed and to highly underestimate high wind speed. A comparison of the ERS-1 and SSM/I F10 gridded data shows a 0.5–1 m s−1 overestimate of the SSM/I wind speed in the western tropical Pacific and in the intertropical convergence zone and south tropical convergence zone strengthening that the SSM/I wind speeds are disturbed in regions of high atmospheric water content.