A recurrent neural network for computing inverse matrices in real-time is proposed. The proposed recurrent neural network consists of n independent subnetworks where n is the order of the matrix. The proposed recurrent neural network is proven to be asymptotically stable and capable of computing large-scale nonsingular inverse matrices in real-time. An op-amp based analog neural network is discussed. The operating characteristics of the op-amp based analog neural network is also demonstrated via an illustrative example.