The effect of delay spread (DS), the specular multipath components (SMCs), and the impact of dense multipath components (DMCs) on multiple-input/multiple-output (MIMO) system performance in a tunnel environment are studied by considering a Richter’s maximum-likelihood (RiMAX)-based estimator. The path loss (PL), angular spread, interuser spatial correlation, and capacity of a massive MIMO framework are analyzed at the 3.5-GHz frequency band through a measurement campaign inside a subway tunnel to implement the latest 5G standards. The radio channel consists of a uniform rectangular array of 32 transmitting elements and a uniform cylindrical array of 64 receiving elements with horizontal and vertical polarizations. The power ratio of the DMCs is found to be distance dependent, as the DMCs and SMCs behave differently in tunnels. Moreover, the study shows that the channel capacity, angular spread, and DS are reduced as the distance between the transmitting and receiving antenna arrays increases, whereas the PL exponent increases with the distance. This research highlights the importance of considering spatial characteristics in 5G massive MIMO system models and provides guidelines to enhance the networks’ accuracy in underground mobile communications.