In recent years, statistics have indicated that the number of patients with malignant brain tumors has increased sharply. However, most surgeons still perform surgical training using the traditional autopsy and prosthesis model, which encounters many problems, such as insufficient corpse resources, low efficiency, and high cost. With the advent of the 5G era, a wide range of Industrial Internet of Things (IIOT) applications have been developed. Virtual Reality (VR) and Augmented Reality (AR) technologies that emerged with 5G are developing rapidly for intelligent medical training. To address the challenges encountered during neurosurgery training, and combining with cloud computing, in this paper, a highly immersive AR-based brain tumor neurosurgery remote collaborative virtual surgery training system is developed, in which a VR simulator is embedded. The system enables real-time remote surgery training interaction through 5G transmission. Six experts and 18 novices were invited to participate in the experiment to verify the system. Subsequently, the two simulators were evaluated using face and construction validation methods. The results obtained by training the novices 50 times were further analyzed using the Learning Curve-Cumulative Sum (LC-CUSUM) evaluation method to validate the effectiveness of the two simulators. The results of the face and content validation demonstrated that the AR simulator in the system was superior to the VR simulator in terms of vision and scene authenticity, and had a better effect on the improvement of surgical skills. Moreover, the surgical training scheme proposed in this paper is effective, and the remote collaborative training effect of the system is ideal.