Massive multiple-input multiple-output (MIMO) systems need to support massive connectivity for the application of the Internet of things (IoT). The overhead of channel state information (CSI) acquisition becomes a bottleneck in the system performance due to the increasing number of users. An intermittent estimation scheme is proposed to ease the burden of channel estimation and maximize the sum capacity. In the scheme, we exploit the temporal correlation of MIMO channels and analyze the influence of the age of CSI on the downlink transmission rate using linear precoders. We show the CSI updating interval should follow a quasi-periodic distribution and reach a trade-off between the accuracy of CSI estimation and the overhead of CSI acquisition by optimizing the CSI updating frequency of each user. Numerical results show that the proposed intermittent scheme provides significant capacity gains over the conventional continuous estimation scheme.