Vertical-axis wind turbines (VAWT) are becoming popular solutions for electric power generation. Among them, Savonius-type VAWTs have low rotational speed, reduced noise and ability to self-start, but their disadvantage is the low energy conversion efficiency. This paper investigates a novel VAWT concept and whether it can provide a higher energy conversion efficiency. The novel concept is a Savonius-type VAWT with flexible blades that change their shape passively due to aerodynamic and blade inertial forces during rotation. An initial design was optimized by a customized shape optimization workflow based on genetic algorithms. The blade design evaluation was performed by CFD model with two-way fluid-structure interaction. The optimization objective was to determine the optimal blade shape, thickness distribution and location of the blade support arms by maximizing the power coefficient and keeping structural stresses below the design limit. The results show that 8% improvement of the power coefficient is possible while keeping the structural stress within the design limit. A detailed performance analysis in the paper shows that there is potential for further optimization considering annual energy production for a selected location. For that purpose, the proposed optimization workflow can serve as a tool for development of future designs. (C) 2018 Elsevier Ltd. All rights reserved.