Stacking of floating structures, known as submodules or cells, provides the modular multilevel converter with theoretically unlimited voltage scalability. However, such a convenience comes at a price of increased control complexity, especially in the domain of internal energy control. In other words, energies of the submodule clusters must be controlled to their setpoint values; otherwise, stable and high-performance operation of the converter cannot be ensured. So far, several approaches toward balancing of the modular multilevel converter internal energy, in both vertical and horizontal directions, have been proposed. Nevertheless, differences among them have never been analytically supported. In this article, three seemingly different energy balancing strategies are thoroughly explained, providing a framework for the theoretical comparison of different dynamic responses provided by them. All the results are verified on the large-scale hardware-in-the-loop platform, serving as a digital twin of a grid-connected 3.3-kVac/5-kVdc 1.25-MW converter being driven by the industrial ABB PEC800 controller.