Beijing Outlines 2026–2028 Strategy to Advance AI for Science Hub

Beijing municipal authorities have issued an action plan for 2026 through 2028 to accelerate the integration of artificial intelligence into scientific research. The strategy outlines 18 specific tasks across five major areas, including building autonomous laboratory networks, upgrading scientific instruments, creating domain-specific foundation models, and expanding data infrastructure. By leveraging its concentration of research institutes and national data centers, municipal officials aim to establish Beijing as a global innovation hub for AI for Science.

Beijing’s targeted policy seeks to modernize traditional trial-and-error research paradigms through automated labs, domain models, and computing stacks, serving as a benchmark for national scientific AI adoption.

Beijing municipal authorities have unveiled the "Beijing Implementation Plan for Accelerating AI-Empowered Scientific Research (2026–2028)," establishing a three-year roadmap to transform the city into a global center for "AI for Science" (AI4S). According to state media reports, the plan outlines 18 key tasks across five core dimensions: establishing autonomous laboratory networks, promoting landmark applications in high-value scenarios, bolstering scientific model architectures, solidifying scientific data infrastructure, and building a supportive innovation ecosystem. Planned measures include upgrading major research instruments with intelligent capabilities, developing scientific research agents, and integrating AI with traditional simulation tools such as Density Functional Theory (DFT). Liu Weihua, deputy director of the Beijing Municipal Science and Tech Commission and ZGC Management Committee, noted that the blueprint establishes an end-to-end framework spanning models, data, AI agents, autonomous laboratories, application scenarios, and ecosystem integration. To support physical implementation, Beijing plans to deploy financial tools including special local government bonds, innovation vouchers, and first-set equipment policies. Li Xinyu, head of the Beijing Institute for AI for Science, highlighted autonomous laboratories—which operate self-contained "dry-wet" feedback loops—as central to replacing conventional, trial-and-error experimental methods. The policy also details efforts to build a multi-tiered model framework comprising basic theory, general models, and domain-specific tools. Xu Bo, director of the Institute of Automation at the Chinese Academy of Sciences, emphasized that science models embedded with physical and chemical principles enable large-scale and long-cycle simulations beyond traditional computational limits. Beijing already hosts significant infrastructure in this domain, including 17 of China's 20 national scientific data centers and platforms like the Bohr Scientific Research Space Station, which reportedly serves over 4.5 million users across hundreds of institutions. The city has previously introduced specialized tools, such as the Rock Science Foundation Model, the DPA Atomic Model, the MegaDFT calculation model, the FlagOS software stack, and the Uni-Lab-OS operating system. Key application areas already targeted include drug discovery, high-energy physics, disease diagnosis, quantum technology, and materials science. To sustain long-term growth, the plan incorporates talent development measures, including interdisciplinary education initiatives and the establishment of a National AI Academy to train researchers.