Recently, a switch-based hybrid massive MIMO structure that aims to reduce the hardware complexity and power consumption has been proposed as a potential candidate for millimeter wave (mmWave) communications. Exploiting the sparse nature of the mmWave channel, compressive sensing (CS)-based channel estimators have been proposed. When applied to real mmWave channels, the CS-based channel estimators may encounter heavy computational burden due to the high dimensionality of the basis. Meanwhile, knowledge about the response of the antenna array, which is needed for constructing the basis of the CS estimators, may not be perfect due to array uncertainties such as phase mismatch among array elements. This can result in the loss of sparse representation and hence the degraded performance of the CS estimator. In this paper, we propose a novel matrix completion (MC)-based low-complexity channel estimator. The proposed scheme is compatible with switch-based hybrid structures, does not need to specify a basis, and can avoid the basis mismatch issue. Compared with the existing CS-based estimator, the proposed basis-free scheme is immune to array response mismatch and exhibits a significantly lower complexity.