Federated learning is a recent approach to machine learning where clients locally train a global model based on their private data and then upload their model parameters to a cloud server for merging into a global model. While it offers great opportunities for privacy-preserving AI solutions e.g. for medical applications, several challenges remain, limiting its efficiency and effectiveness, namely communication cost and the model parameter divergence in the presence of non-IID training data. In this paper, we analyze the state of the art with respect to the aforementioned challenges. Considering the characteristics of modern wireless networks, we notice that there is a huge opportunity for leveraging wireless broadcasting in a hybrid system architecture comprising peer-to-peer subgroups and a hierarchical server infrastructure. We design a new protocol for federated learning where peers share gradient updates with other nearby peers via wireless broadcasting and globally exchange gradient updates via a hierarchical server network. This way, we benefit from the efficiency of wireless broadcasting to increase communication efficiency and decrease server involvement. Furthermore, the frequent exchange of gradient updates between peers allows us to better cope with non-IID training data. Our protocol serves as a pattern that can be fine-tuned using many recently published contributions in federated learning. The impact of our work is expected to become even stronger in future 5G+ networks because it benefits from high device density and mobility.