We propose a semiclassical version of Shor's quantum algorithm to factorize integer numbers, based on spin-1/2 SU(2) generalized coherent states. Surprisingly, we find evidences that the algorithm's success probability is not too severely modified by our semiclassical approximation. This suggests that it is worth pursuing practical implementations of the algorithm on semiclassical devices.