This paper evaluates the suitability of airborne gravimetry for geoid determination. Spectral analysis of gravity data is used to derive the high-frequency geoid spectrum and to compare it to the error spectrum of the geoid, as derived from an airborne gravity system. Gravity values in various areas with typical topographic features are used for the spectral analysis. The achievable accuracy of airborne gravimetry in geoid determination is studied through an assessment of inertial navigation system (INS) and global positioning system (GPS) errors in the spectral ranges of interest. The results of these computations are presented in a number of graphs which allow the following conclusions. To determine geoid undulations with an accuracy of 1 cm (rms), the minimum wavelength to be resolved is 14 km in flat terrain and 5 km in mountainous terrain. Using a flight speed of 300 km/h at an altitude of 2.5 km or less above ground, current hardware sensitivity is sufficient to resolve the geoid spectrum for such wavelengths. The total GPS/INS induced geoid undulation error (rms) in this case is less than 1 cm for wavelengths between 5 km and 100 km and less than 10 cm for wavelengths between 5 km and 500 km. The cumulative rms geoid error is less than 30 cm for wavelengths up to 1000 km. The accuracy for relative geoid determination is 0.1 ppm for wavelengths up to 100 km and 0.2 ppm for 500 km. The results were verified by analyzing the data collected during an airborne test. The results indicate that airborne gravimetry has the potential to determine geoid undulation with centimeter-level accuracy and to fill the gaps in the global gravity map in an efficient manner and with an accuracy superior to current terrestrial methods. ¿ American Geophysical Union 1996