The possibility of creep cavity formation at subboundaries in austenitic stainless steels is analysed. It is demonstrated that such nucleation is thermodynamically feasible. A minimum stress must be exceeded in order to create cavities. The nucleation is assumed to take place where subboundaries on one side of a sliding grain boundary meet subgrain corners on the other side (double ledge models). Alternative cavitation positions can be found where particles meet subboundaries. The nucleation model can quantitatively predict the observed nucleation rate. The model gives a nucleation rate that is proportional to the creep rate in agreement with many experiments.