Large AI Model in the Air: Ubiquitous Intelligent Networks for Low Altitude Economy

Low Altitude Economy (LAE), empowered by aerial platforms such as unmanned aerial vehicles (UAVs), is rapidly evolving as a key enabler for next-generation smart services like urban mobility, environmental monitoring, and emergency response. To support these applications, aerial platforms should perform real-time, efficient navigation and sensing services in dynamic, dense, and uncertain low-altitude environments via large artificial intelligence (AI) models. However, enabling such intelligence requires overcoming significant challenges of limited onboard resources and dynamic connectivity for the aerial platforms. This paper provides a comprehensive study on addressing these challenges to enable large AI model-based services in LAE networks. We propose a cloud-edge-end architecture that facilitates collaborative model training and inference among aerial and ground nodes. We also explore three key enabling technologies, i.e., model collaboration, dynamic networking, and resource management. We validate the proposed framework through a case study on radio map generation for LAE networks. Moreover, we discuss future research directions that pave the way for scalable, intelligent, and resilient large AI model-driven LAE networks.