We study the effect of environmental risk on cooperation in the Voluntary Contribution Mechanism. Our baseline is the standard setting in which the personal return from the public good is deterministic, homogeneous, and publicly known. Our experimental treatments alter this classic design by making the marginal per capita return from the public good probabilistic. In the homogeneous risk treatment, the random draw is made for the whole group, while in the heterogeneous risk treatment this happens independently for each group member. Our main result is that risk does not harm cooperation either in the one-shot or in the finitely repeated version of the game. This suggests that the standard experimental methodology provides a robust and conservative measure of human cooperation.