This dataset comprises global upper thermospheric cross-track neutral wind measurements obtained from accelerometer data of the CHAMP satellite during its almost ten year’s lifetime from 2001 to 2009. One key scientific instrument on-board CHAMP was a sensitive triaxial accelerometer. It was located at the spacecraft's centre of mass and sampled effectively accelerations due to non-gravitational forces with an accuracy of ~3×10^-9 ms^-2 (Doornbos et al., 2010). The along-track air drag measurements resulted in thermospheric mass density estimations, while the instrument was sensitive enough to deduce also the horizontal neutral wind component from the cross-track accelerations. The CHAllenging Minisatellite Payload (CHAMP) spacecraft circled the Earth from July 2000 to September 2010 on a near-polar orbit (inclination 87.3°). Each orbit period took about 93 minutes at an altitude of initially 455 km, and decaying to about 320 km in 2009. Due to CHAMP's precession, the satellite achieved full coverage of all local times within about 131 days in each case. This work was part of a study in 2007-2009 (Doornbos et al., 2009) funded by the European Space Agency’s General Studies Program which aimed at a more precise estimation of the non-gravitational forces, considering the precise satellite geometry and its optical and mechanical surface properties. To obtain the actual air drag forces, the modelled accelerations due to radiation pressure forces from the sun, the Earth's albedo, and the Earth's infrared radiation had to be computed and removed from the calibrated and edited accelerometer data to get the observed aerodynamic acceleration vector. The modelling of the radiation pressure forces comprised several nontrivial components like the modelling of eclipse and semi-shadow conditions for solar radiation pressure, values for the reflectivity and infrared emissivity of Earth surface elements, and models of the geometry and optical properties of the satellite surfaces (Doornbos et al., 2010). The detailed description of supersonic flow of the neutral gas particles across the satellite's surface and its reflection requires a model of the gas–surface interaction, which specifies the angular distribution and energy flux of the reflected particles. One has to make assumptions and educated guesses, because information on the gas–surface interaction, as well as in situ observations of aerodynamic model parameters like air temperature and neutral gas species' concentrations should be measured by independent instruments on the accelerometer-carrying satellite. Here, we relied on the empirical atmosphere model NRLMSISE-00 (Picone et al., 2002) and the rarefied aerodynamic equations for flat panels, derived by Sentman (1961). These equations take into account the random thermal motion of the incident particles and assume a completely diffuse distribution of the reflected particle flux. The energy flux accommodation coefficient alpha (Moe et al., 2004), which determines whether the particles retain their mean kinetic energy (alpha = 0) or acquire the temperature of the spacecraft surface wall (alpha = 1), was found to be optimally chosen with alpha = 0.8 for this data set. This thermospheric cross-track neutral wind data set consists of a series of annual CDF data files for both CHAMP wind measurements (subfolder: CH_PN_R03_denswind_iter2_Sentman_alpha08) and CHAMP orbital data (subfolder: CH_orbit_GEO_RSO). The CDF data files are documented in the header. The complete dataset contains more than 25 million data points with a temporal cadence of 10 sec. In addition to the data, we are providing supplementary Figures to Aruliah et al. (2019, subfolder: 2019-001_Foerster-Doornbos_Figures). They are complementary, in particular, to Figs. 1-4 of this paper, but additionally show the original data as “cloud” of data points in the background of the statistical averages. Each figure plot (png-format) has an accompanying txt-file of the same name (except the extension) with ASCII tables of the hourly statistical averages and their standard deviations. The data were used in various previous publications mainly with respect to high-latitude upper thermosphere studies (Förster et al., 2008, 2011) and investigations of the interhemispheric coupling processes of the magnetosphere, ionosphere, and thermosphere (Förster et al., 2017). Actually, this data publication serves as supplement to Aruliah et al. (2019).