In this study, we present a new approach to design a Least Mean Squares (LMS) predictor. This approach exploits the concept of deep neural networks and their supremacy in terms of performance and accuracy. The new LMS predictor is implemented as a deep neural network using multiple non linear LMS filters. The network consists of multiple layers with nonlinear activation functions, where each neuron in the hidden layers corresponds to a certain FIR filter output which goes through nonlinearity. The output of the last layer is the prediction. We hypothesize that this approach will outperform the traditional adaptive filters.