Polar wind analyses for numerical weather prediction (NWP) have long been hampered by a lack of wind observations in polar regions. Independent rawinsondes from Arctic field experiments indicate considerable biases in the polar wind fields of NCEP/NCAR and the ECMWF reanalyses. In collaboration with the Cooperative Institute for Meteorological Satellite Studies (CIMSS), an initial study to investigate the impact of a new satellite-derived polar wind data set on ECMWF analyses and forecasts has been undertaken. These observations are derived at CIMSS by tracking structures in subsequent swaths from the MODerate resolution Imaging Spectrometer (MODIS) instrument on the TERRA satellite. They provide unprecedented observational coverage over the polar regions. A first 30-day series of experiments with the MODIS winds has been performed using ECMWF's 3-dimensional variational (3DVAR) assimilation system. These experiments show that the fit of other observations to the model first guess is not altered when MODIS winds are assimilated, indicating a good consistency between MODIS winds and the rest of the observing network in the Arctic. A slight degradation is observed over the Antarctic coastal areas, possibly related to a poor vertical and horizontal resolution of the forecast fields used in the height assignment for MODIS winds, and therefore requiring a stringent quality control. Overall, the mean polar wind analysis is considerably altered when MODIS winds are assimilated. There is a significant positive impact on forecasts of geopotential heights over Arctica spreading out to the whole Northern Hemisphere, and to a lesser extent over Antarctica, too. MODIS AMVs are now received in Near Real Time at ECMWF. At the time of writing, new assimilation experiments using ECMWF's 4DVAR system are underway to evaluate the potential of these data during a Northern Hemisphere summer period. The paper will present the latest experiments together with the preparation undertaken at ECMWF towards an operational assimilation of polar MODIS winds.