Scalability and flexibility are widely considered as two major design goals for 5G networks. Aiming at these goals, this article first identifies a promising architecture based on the heterogeneous cloud radio access network (H-CRAN), reviews the challenges in MIMO precoding for H-CRAN, and then proposes a scalable and flexible massive MIMO precoding scheme by exploiting the null-space of user signals. Specifically, the system can accomplish effective radio resource management and flexible spatial coordination by distinguishing the intended and victim users' CSI, and avoid the interference by precoding within the null-space for the CSI of victim users. Simulation results indicate that the proposed scheme is capable to effectively alleviate the interference to victim users and support high QoS as well as spectral efficiency.