Neurostimulation is a common medical treatment modality used to treat neurological disorders. It applies electrical pulses to revert and prevent undesired neural behavior or to create desired neural behavior. The required specificity of said stimulation treatments is oftentimes achieved by implanting the neurostimulator close to the target area. Implantation of active medical device poses a series of challenges and requirements in size and power consumption. Wireless power transfer (WPT) technologies have been used to reduce the size or eliminate the use of batteries, which are the main limiting component in terms of size and longevity of the implants. In conventional WPT systems, the energy must be focused from an external transmitter to a receiver in the implanted neurostimulator. This one-to-one link cannot be used in some applications in which a network of neurostimulators is required to deliver treatment in multiple locations distant from each other. This is the case for some chronic headache treatments in which both the supraorbital and occipital nerves must be stimulated. This thesis proposes a new system level topology for a neurostimulator network for the treatment of chronic headaches, in which all implants are powered through a single WPT link. It was theorized that the bone tissue in the skull can be used as an acoustic conductor for ultrasonic waves, similarly to bone-anchored hearing aid systems that already use bone conduction to conduct acoustic waves in the audible range. Finite element simulations showed that the skull can conduct ultrasonic energy in two frequency bands, which are 0.1-0.6 MHz and 1-2 MHz. A 2 MHz operating frequency was chosen for safety reasons, since the 1-2 MHz band does not leak unwanted energy into neighbouring soft tissues like the brain. At this frequency, the WPT link undergoes an attenuation of about 20 dB. Each individual neurostimulator employs the ultra high frequency (UHF) stimulation technique to improve in power efficiency. Both the UHF stimulation and ultrasonic wireless power transfer work in the MHz frequency range, so the ultrasound signal can be “directly” used for the electrical stimulation of the tissue by means of a simple ultrasound transducer and some power conditioning. Previous designs using this concept are not easily scalable and cannot perform charge balancing to ensure safe stimulation of tissue. In addition, the use of a WPT powering method cannot guarantee a stable and uniform supply for the implanted neurostimulator. As a result, there is a need for a multichannel system that can deliver UHF stimulation pulses in a safe manner, while ensuring efficacy of the stimulation regardless of the power levels received through the WPT link. This thesis proposes a novel output stage for implantable neurostimulators that ensures the efficacy and safety of the stimulation. Since the efficacy is directly related to the total charge delivered to the tissue, a novel charge metering circuitry is designed to control said amount of charge. Safety is obtained by minimizing the residual voltage after a stimulation cycle, which can otherwise damage the tissue and the electrodes. This is done by means of a novel charge balancing scheme that monitors the voltage across the electrodes and stops the stimulation when it crosses 0 V. Circuit simulations successfully validate the design. The proposed neurostimulator was also implemented on a PCB board and tested. The charge delivery resolution of the charge metering circuit is 510 pC. A residual voltage of 0.5-29.5 mV was achieved with the charge balancing circuitry, using an electrical model of the tissue. In vitro measurements in a phosphate buffered saline solution show a 80 mV residual voltage. Hence, this work successfully presents a new neurostimulator output stage topology that ensures efficacy and safety in the presence of an unreliable WPT link, while being compatible with multichannel operation and an IC implementation.