The collapse of turbulence in the nocturnal boundary layer is studied by means of a simple bulk model that describes the basic physical interactions in the surface energy balance. It is shown that for a given mechanical forcing, the amount of turbulent heat that can be transported downward is limited to a certain maximum. In the case of weak winds and clear skies, this maximum can be significantly smaller than the net radiative loss minus soil heat transport. In the case when the surface has low heat capacity, this imbalance generates rapid surface cooling that further suppresses the turbulent heat transport, so that eventually turbulence largely ceases (positive feedback mechanism). The model predicts the minimum wind speed for sustainable turbulence for the so-called crossing level. At this level, some decameters above the surface, the wind is relatively stationary compared to lower and higher levels. The critical speed is predicted in the range of about 5–7 m s-1, depending on radiative forcing and surface properties, and is in agreement with observations at Cabauw. The critical value appears not very sensitive to model details or to the exact values of the input parameters. Finally, results are interpreted in terms of external forcings, such as geostrophic wind. As it is generally larger than the speed at crossing height, a 5 m s-1 geostrophic wind may be considered as the typical limit below which sustainable, continuous turbulence under clear-sky conditions is unlikely to exist. Below this threshold emergence of the very stable nocturnal boundary layer is anticipated.