Cornell researchers have developed a novel computing device that stores information electrically but reads it through mechanical vibration, aiming to reduce energy consumption in AI and scientific computing. - The device uses a ferroelectric nanoelectromechanical system (FeMEMS) with a vibrating beam to read stored analog values.
- It can distinguish roughly 200 electromechanical states, enabling precise analog computation. - The approach is designed for neuromorphic and analog in-memory computing, reducing energy wasted on moving data between memory and processing.
- The prototype chip was fabricated at Cornell and supported by DARPA and NSF. This innovation could lead to more energy-efficient AI hardware and adaptive microsystems, reviving older computing concepts with modern materials.