In this paper, we investigate the prediction of the cell composition of bacteria with respect to their medium. By modeling the bacterium as an interconnection of subsystems, the problem is written as a nonsmooth convex optimization problem equivalent to a Linear Programming feasibility problem. We then obtain a new method, called Resource Balance Analysis (RBA), predicting the distribution of the available resources in the medium among the various cellular subsystems. Beyond its predictive capability, the proposed approach grasps some fundamental aspects of the bacterium physiology by including a refined model. This method reveals the existence of an intrinsic bottleneck in the system resource distribution of the bacterium, leading to the existence of a structural limitation of its growth rate which can be predicted. RBA is also able to predict the configuration of the metabolic network for a given medium at steadystate regimen which nicely fits the available experimental results for the gram-positive model bacterium Bacillus subtilis.