Microwave remote sensing over land has lagged behind remote sensing over oceans. This is due to the larger land emissivity values and their changes on daily to seasonal timescales. The lack of surface emissivity knowledge has hindered full exploitation of the capabilities of operational satellite sensors such as AMSU and SSM/I in remote sensing and data assimilation. Microwave land surface models that can be used to predict the emissivity have been developed and show promise, but independent global measurements of emissivity are needed for comparison. We have developed an optimal estimation retrieval, demonstrated with AMSU data, which retrieves global emissivity at frequencies from 23 to 183 GHz. A simultaneous retrieval of the temperature and water vapor profiles as well as cloud liquid water is performed. The simultaneous retrieval allows the masking effects of water vapor and some clouds to be reduced. The method does not currently use infrared data as a surface temperature constraint in clear sky regions. Initial comparisons with the NOAA NESDIS Microwave Emissivity Model are encouraging. The retrieval provides an independent estimate of emissivity, which is especially useful over difficult surface types such as snow and ice. These early results point the way to the creation of dynamic global emissivity fields which have applications to satellite microwave data assimilation, remote sensing of soil moisture, and future microwave sensors.