Millimeter Wave (mmWave) massive Multiple In- put Multiple Output (MIMO) systems realizing directive com- munication over large bandwidths via Hybrid analog and digital BeamForming (HBF) require reliable estimation of the wideband wireless channel. However, the hardware limitations with HBF architectures in conjunction with the short coherence time inherit in mmWave communication render this estimation a challenging task. In this paper, we develop a novel wideband channel estima- tion framework for mmWave massive MIMO systems with HBF reception. The proposed framework jointly exploits the low rank property and the available angular information to provide more accurate channel recovery, especially for short beam training in- tervals. We introduce a novel analog combining architecture that includes a random spatial sampling structure placed before the input of the analog received signals to the digital component of the HBF receiver . This architecture supports the proposed matrix- completion-based estimation approach in providing the sampling set of measurements for recovering the unknown channel matrix.