The development of underground spaces in megacities is crucial for alleviating urban land pressure, optimizing transportation, and enhancing resilience. However, under the dual pressures of global climate change and rapid urbanization, the traditional concept of engineering resilience struggles to address challenges such as the complexity of underground space systems, multi-hazard coupling, and intelligent data processing. Consequently, a shift toward an AI-enabled intelligent resilience concept, focusing on resilient prevention and mitigation of disasters, has become an imperative direction for future development. This study first categorizes the types of disasters faced by underground space systems in megacities and elaborates on the concept of intelligent resilience for such systems. It then identifies key scientific and technological issues involved in developing intelligent resilience. Subsequently, the study analyzes the current state and existing problems in intelligent resilience development across several dimensions: spatiotemporal distribution of disaster types and intensities; theories for failure mechanism analysis and resilience assessment; synergistic material-structural systems; comprehensive sensing technologies and high-fidelity databases; and AI-driven autonomous decision-making and intelligent evolution. Furthermore, a multi-dimensional collaborative management and control framework based on the integration of physical, information, and simulation domains with dynamic real-time data interaction is constructed, which clarifies the implementation pathways for the intelligent resilience systems of underground space in megacities. Finally, the study recommends strengthening comprehensive safeguards across physical, information, and simulation domains to construct an integrated support system, thereby promoting the orderly and efficient advancement of intelligent resilience in megacity underground space systems.