A modular multilevel converter (MMC) synthesizes output voltages by combining voltages from various submodules (SMs) connected in series; thus, it is ideal for high-voltage transmission and distribution ranging from tens to several hundreds of kilovolts. However, MMCs incur high costs and losses, thus necessitating reliable operation. Research has been conducted to assess and enhance MMC reliability. Prior studies have presumed an exponential distribution for the reliability of SMs, including redundancy, thus implying that the failure rate does not change over time. This assumption signifies a constant system availability for the converter station, regardless of time, thus presenting a limitation in adequately comprehending the breakdown impact caused by a failure of the converter station. Additionally, as no standard exists for determining redundancy in medium-voltage direct current (MVDC) systems, a methodology is required to establish a redundancy criterion during the design process. Particularly for MVDC systems, the impact of power outages must be evaluated by considering the cost of expected outages based on system availability, as the load is interconnected. As a solution, we propose a method to assess the tradeoff between system availability and cost for an MVDC MMC, employing the Weibull distribution-based SM failure rate. The L-BFGS algorithm calculates the optimal value of the availabilitycost function, and the number of redundant modules minimizing total operating cost is evaluated. Our case study determines the appropriate number of redundancy modules based on outage costs. The proposed methodology enables redundancy design that considers both reliability and operational costs, thus providing a practical solution to the problem of determining the number of redundant modules needed in MMCs.