Channel state information (CSI) plays an important role in the next generation of cellular systems with massive multiple-input multiple-output (MIMO) technology as the indicator of wireless channels. In hyper cellular networks (HCNs), the traffic base stations (TBSs) improve the energy efficiency by dynamical sleeping. However, the conventional pilot-based CSI acquisition method cannot be applied to sleeping cells. We proposed a novel CSI scheme strategy based on channel learning for this problem. Different from location-aided CSI acquisition scheme, the proposed method utilizes the CSI at the control base station (CBS) as input to avoid the error caused by positioning. We validate our scheme in a HCN generated by the geometry-based stochastic channel model (GSCM). The prediction accuracy of the proposed scheme is better than the K nearest neighbor (KNN) method and close to the location-aided CSI acquisition scheme, which requires the knowledge of user position.