The technology of Resilience 1.0 focused on establishing critical capabilities such as early warning and infrastructure hardening. Resilience 2.0 represents a technological leapfrog into a future defined by AI, autonomy and hyper-integration, creating a truly intelligent and adaptive utility network. The core of this leap is leveraging predictive analytics for pre-emptive maintenance and dynamic infrastructure reconfiguration. The AI of Resilience 2.0 will go beyond just forecasting a storm; it will predict the specific consequences of that storm at a component level. An AI model could analyse real-time sensor data from a transformer, combine it with a hyper-local heatwave forecast and flag a 90% probability of failure within 48 hours, automatically dispatching a maintenance crew before the failure occurs. In the event of an unavoidable outage, the system will move towards a “self-healing” state. An AI- powered grid controller will autonomously reroute electricity or water flows around the damaged section in milliseconds, isolating the fault and minimizing the scope and duration of the service disruption for customers. This will be complemented by the large-scale deployment of autonomous systems for rapid damage assessment and maintenance support, especially in remote and inaccessible areas. While drones are already used for reconnaissance and data collection, the next generation will feature swarms of automated drones and ground-based robotics capable of delivering lightweight components and, in time, performing initial repairs. This will drastically reduce restoration times and improve safety for human crews who would follow up to complete the work. The carbon emissions associated with AI applications are drawing significant attention, particularly the environmental impact of large language models (LLMs). According to recent analysis by Google, the median Gemini Apps text prompt uses less energy than watching nine seconds of television (0.24 Wh) and consumes roughly the equivalent of five drops of water (0.26 mL) – impacts that are small relative to many everyday activities.40 While such per -use metrics illustrate efficiency gains, it is equally important to address the broader, systemic challenge: the carbon footprint of the global data centres that power these AI applications. The total energy consumption of these facilities represents a significant environmental concern that requires a different scale of solutions. Addressing this macro-level challenge relies on broad, multi-faceted commitments. On one front, 3.2 Technological leapfrogging: AI-powered integration Resilience 2.0 represents a technological leapfrog into a future defined by AI, autonomy and hyper-integration, creating a truly intelligent and adaptive utility network.In addition, it is critical to strengthen market- orientated values and development. For example, the recent implementation of Policy No. 13639 on deepening the market-orientated reform of new energy on-grid electricity prices outlines a pathway for China’s utilities sector – particularly the power industry – towards greater marketization and enhanced resilience. By adopting an approach that “deregulates pricing while regulating mechanisms”, these policies shift the new energy industry from subsidy dependency to competition-driven efficiency, using market mechanisms to explicitly value system flexibility and reliability. This incentivizes all stakeholders – such as in generation, grid, load and storage – to collectively participate in building resilience. This evolving governance architecture will create a powerful feedback loop, turning resilience from an abstract priority into a measurable, reportable and auditable mandatory function that directly influences strategy, operations and market valuation. Building Climate-Resilient Utilities: Lessons from China and Future Pathways 19

Building Climate Resilient Utilities 2025