In this thesis, we propose neural dust, an entirely new method of wireless power and data telemetry using ultrasound, which can address fundamental issues associated with using EM to interrogate miniaturized implants. Key concepts and fundamental system design trade-offs and ultimate size, power, and bandwidth scaling limits of such system are analyzed from first principles. We demonstrate both theoretically and experimentally that neural dust scales extremely well, down to 100's, if not 10's of $\mu$m. We highlight first wireless recordings from nerve and muscle in an animal model using neural dust prototype. The thesis concludes with strategies for multi-neural dust interrogation and future directions of neural dust beyond neuromodulation.