The Sunshine Zhengyan GPT is an LLM that showcases the immense potential of China’s vertical-specialization approach, combining proprietary data with sector-specific applications. Developed by Sunshine Insurance Group, this LLM was trained to enhance customer service, sales support and intelligent claims processing. Since its implementation, the model has improved information extraction accuracy by 15% compared to the pre-trained language model BERT.21 Meanwhile, the injury claims bot launched in March 2024 is now used in over 80% of cases handled by the firm, with an accuracy rate of nearly 96% in document classification.22CASE STUDY 2 Sunshine Zhengyan GPT – transforming insurance with AI-driven claims and service excellence 2.4 Enabling sustainable energy solutions to power AI As AI advances, its energy demands are surging. Foundation models require substantial computing power and can consume up to 33 times more energy in order to complete a task compared to task- specific software.23 In response to these demands, China is prioritizing sustainable energy solutions to power AI while minimizing its environmental impact. One significant advantage for China in advancing its AI infrastructure is its ability to rapidly supply green energy. The country has established the world’s largest and most comprehensive renewable energy industry chain.24 China is projected to account for nearly 60% of all renewable energy capacity installed globally between now and 2030 – thereby positioning the country to host almost half of the world’s renewable power capacity by the end of the decade.25 In addition to increasing renewable energy capacity, China is actively redistributing computing resources internally, from densely populated eastern regions to western regions where renewable energy is abundant. The East Data, West Computing strategic initiative has received over $6.1 billion in government funding over the past two years alone.26 For their part, green data centres in China are rapidly evolving, driven by government initiatives, technological innovation and industry commitments to sustainability. To date, 246 green data centres have been established in the country, with an average power-usage effectiveness of approximately 1.2. 27 Cooling systems, which account for 30-40% of a data centre’s energy consumption,28 present a significant opportunity for efficiency gains. By optimizing these systems alongside renewable energy and intelligent management, China can create greener, more sustainable data infrastructure.2.3 Advancing industry-specific AI models In the rapidly advancing field of large language models (LLMs), China is witnessing a strong trend towards the development of industry- specific or domain-specialized models. Through close collaboration with industry partners, these models integrate specific data, knowledge and task requirements, thereby enabling deeper customization and optimization. This vertical specialization could unlock significant advantages for domain-specific use cases within industries that are characterized by high levels of digitalization, robust data foundations and intricate knowledge systems. Concerns about data security and the potential leak of proprietary information, however, have made some organizations cautious about data sharing, which limits the generalizability and accuracy of industry-specific or domain-specific models. To maximize the impact of domain-specific LLMs, research shows that cross-industry collaboration and coordinated efforts between upstream and downstream partners are essential. China is prioritizing sustainable energy solutions to power AI while minimizing its environmental impact. Blueprint to Action: China’s Path to AI-Powered Industry Transformation 14

Blueprint to Action Chinas Path to AI-Powered Industry Transformation 2025