Low-latency communication is currently attracting significant attention due to the emergence of mission-critical Internet of Things (IoT) applications and content-centric services. A deep understanding of the delay performance is essential for efficient wireless system design and end-to-end latency guarantees. In this paper, we investigate the network-layer performance of physical layer multi-antenna multicasting, i.e., when the same data is simultaneously conveyed to multiple users. We provide a statistical characterization of the service process in terms of its Mellin transform and derive probabilistic delay bounds using tools from stochastic network calculus. Furthermore, using extreme value theory, we characterize the service process for very large number of users and derive scaling laws as the number of antennas and/or users is taken to infinity. Our results can be used for system dimensioning to guarantee the delay requirements in wireless multicast networks.