Power inductors are indispensable in power electronics systems, where they play a critical role in ensuring efficient energy conversion. However, these components contribute to significant inefficiencies and occupy valuable space. Traditional optimization techniques, such as genetic algorithms (GA) and covariance matrix adaptation evolution strategy (CMA-ES), offer promising designs to address these issues but are hindered by high computational costs. This study proposes a novel optimization approach based on the actor-critic (AC) method, a reinforcement learning technique. The AC method not only reduces computational time substantially but also achieves stable convergence toward optimal inductor shapes. Through direct comparisons with GA and CMA-ES, we demonstrate the AC method’s potential to set a new standard for efficient and compact power inductor design.