When driving an electric vehicle, the interior sound experience is characterized by low overall levels. The frequency spectrum is dominated by tonal, high-frequency elements. The main source of this sound signature is the electric drive unit (EDU) powering the vehicle. The challenge this poses to NVH development is twofold. On the one hand, tonality becomes the main criterion for evaluating the interior noise quality of electric vehicles. This means generating and applying metrics and methods that can robustly analyze and evaluate tonality is a must for electric vehicle NVH engineering. And on the other hand, the focus of powertrain NVH development shifts from achieving the lowest overall level possible to limiting the acoustic prominence of the electric drive unit’s tonal and high-frequency noise components. To reconcile these challenges, AVL has developed a robust and flexible NVH development methodology for electric drive units. In this paper, AVL presents how this methodology can be applied throughout the development process. Starting from vehicle benchmarking to generate reasonable NVH targets, these targets are detailed and broken down to system and component level where they support the conception and initial design of the electric drive unit and its components. In the concept and the design phase multiple subsystem models are combined to a Digital Twin accompanying the development of the physical electric drive unit. This Digital Twin ensures the continuous optimization of the system by using simulation and artificial intelligence techniques until the final vehicle integration. This allows robust virtual studies of different attributes like tolerances and temperatures and their influence on NVH behavior. From the first prototype onwards simulation and testing go hand in hand, refining the individual systems with each hardware generation resulting in an NVH-optimal electric drive unit. The method guarantees an efficient target achievement process and helps to avoid costly mistakes.