Recovering data symbols in a wireless communications system consists of two main estimation steps: channel estimation based on transmitted pilot symbols, and estimation of data symbols using the acquired channel information. The amount of energy allocated to each of pilot and data transmission determines the performance of each estimation step, which further impacts the overall system performance. In this paper, we consider a linear minimum mean squared error (LMMSE) receiver that uses the LMMSE estimator for both channel information acquisition and data symbol recovery in the context of a massive MIMO system. We derive the mean squared error (MSE) of the estimated symbols as a function of the energy allocation. Exploiting the large dimensionality of the problem, we leverage tools from random matrix theory to express the MSE only in terms of the deterministic parameters of the system. We further utilize the deterministic expression to find the optimal energy allocation. The theoretical results are matched with simulations showing high level of congruence.