Facilitated by microelectromechanical systems (MEMS) technology, MEMS speakers or microspeakers have been rapidly developed during the past decade to meet the requirements of the flourishing audio market. With advantages of a small footprint, low cost, and easy assembly, MEMS speakers are drawing extensive attention for potential applications in hearing instruments, portable electronics, and the Internet of Things (IoT). MEMS speakers based on different transduction mechanisms, including piezoelectric, electrodynamic, electrostatic, and thermoacoustic actuation, have been developed and significant progresses have been made in commercialization in the last few years. In this article, the principle and modeling of each MEMS speaker type is briefly introduced first. Then, the development of MEMS speakers is reviewed with key specifications of state-of-the-art MEMS speakers summarized. The advantages and challenges of all four types of MEMS speakers are compared and discussed. New approaches to improve sound pressure levels (SPLs) of MEMS speakers are also proposed. Finally, the remaining challenges and outlook of MEMS speakers are given.