An encoder-decoder layer stacked RNN with GRU basic cells is used to predict Bitcoin value. RNNs are very good at learning sequences, and indeed the Bitcoin prediction is reasonably accurate even with a simple model based on 2 layers and 12 GRU cells. Of course, this prediction code is not an encouragement to invest in Bitcoin but only to discuss a deep learning methodology. Moreover, more experiments are needed in order to verify whether or not we have a situation of data overfitting.