Model predictive control has emerged as a promising approach to govern modular multilevel converters (MMCs), due to its flexibility to include multiple control objectives and simple design process. However, this control scheme presents relevant issues, such as high computational complexity and variable switching frequency. This work proposes a sequential phase-shifted model predictive control (PS-MPC) for MMCs. The key novelty of this proposal lies in the way the predictive control strategy is formulated to fully exploit a phase-shifted pulsewidth modulation technique through an appropriate choice of synchronized average models for each carrier. In this way, the proposed predictive controller obtains independent optimal modulating signals for each carrier in a sequential manner, by solving an optimization problem with reduced computational effort independent of the number of submodules. This sequential optimization allows one to formulate the optimal control problem to achieve multiple control objectives, similarly to the finite-control-set MPC (FCS-MPC). Nevertheless, the MMC governed with the proposed PS-MPC generates an output voltage with fix-spectrum and operates with an even power loss distribution among semiconductors in steady-state, outperforming the standard FCS-MPC strategy. Experimental results are provided to verify the proposed PS-MPC effectiveness when governing a three-phase MMC with four half-bridges per stack.