Moment method is one of the most popular frequency domain simulation methods in computational electromagnetics (CEM). Since it solves an integral equation rigorously through boundary discretization, huge computer memory and computation time are needed to simulate the scattering phenomena for targets of large electrical sizes. Recently the development of the fast multipole method (FMM) and the multilevel fast multipole algorithm (MLFMA) has led to the reduction of the computational complexity of the matrix-vector multiplication operation in moment method from O(N') to O(N") or O(NlogN) [1,2]. The FLFMA-based code FISC  is now capable of computing the radar cross section (RCS) of a full-size fighter airplane at I GHz on a workstation in several hours of run time. While such an advance in computational capability has brought much excitement to the CEM community, real-world radar signature prediction often requires that the same calculation be repeated over many frequencies and aspect angles. For example, in automatic target identification applications, range profiles of the target are commonly used to pinpoint target features. To generate range profiles requires hundreds of frequency calculations that make the computation time still extraordinarily long even using FISC. In this paper, we present a frequency extrapolation scheme to speed up the signature prediction procedure using FISC by avoiding such exhaustive computations.