Similar to the conventional orthogonal frequency-division multiplexing (OFDM) system, an OFDM multiple access (OFDMA) system will have a carrier frequency offset (CFO) problem. Since CFOs of all users are different, CFO compensation in the OFDMA uplink system is much more involved. A simple, yet efficient, method is the zero-forcing (ZF) compensation method. However, it involves an inverse of an N x N CFO-induced ICI matrix, where N is the number of subcarriers. Thus, the complexity can become very high when N is large, a case commonly seen in OFDMA systems. In this work, we propose a low-complexity ZF method to overcome the problem. The main idea is to use Newton's method to solve matrix inversion iteratively. We explore the structure of the CFO-induced ICI matrix and develop a method that can implement Newton's method with fast Fourier transforms (FFTs). As a result, the required computational complexity is significantly reduced from O(N3) to ο(2N log2 N). Simulations show that, with only three iterations, the proposed method can have similar performance to the direct ZF method.