Beijing Unveils AI-Empowered Research Plan for 2028
Beijing has released an ambitious three-year implementation plan to integrate artificial intelligence across scientific research disciplines, aiming to transform the Chinese capital into a “globally influential scientific intelligence innovation hub” by 2028. The plan, issued on June 30 by the Beijing Municipal Committee’s Education, Science, Technology and Talent Work Leading Group, represents China’s first comprehensive local policy dedicated to AI for Science (AI4S).
Background and Strategic Context
The Beijing Implementation Plan for Accelerating AI-Empowered Scientific Research (2026-2028) builds on a preliminary AI4S policy released in 2025, which was China’s first local policy specifically targeting scientific intelligence. The new plan significantly expands on that foundation with 18 specific tasks across five strategic directions, concrete funding mechanisms, and measurable targets for 2028.
Beijing is uniquely positioned to lead this push. According to official figures, the city hosts over 800 academicians of the Chinese Academy of Sciences and Chinese Academy of Engineering, more than 1 million scientific researchers, 4 national laboratories, and 145 state key laboratories — roughly 30 percent of China’s total. R&D spending intensity has consistently exceeded 6 percent of GDP, and 17 of China’s 20 national science data centers are located in Beijing.
Five Strategic Pillars
The plan organizes its 18 key tasks around five strategic directions, each targeting a critical component of the AI-driven research ecosystem.
Autonomous Laboratory Systems
A centerpiece of the plan is the development of autonomous laboratories that integrate AI, embodied robots, and high-throughput instruments. These facilities are designed to operate the full research cycle autonomously — from hypothesis generation and experiment planning to data collection, simulation, validation, and discovery. The plan calls for the development of “AI Scientist” research agents and toolkits, as well as the establishment of construction standards and tiered evaluation mechanisms.
As Li Xinyu, Dean of the Beijing Academy of AI for Science, noted, “The traditional trial-and-error research model has long cycles and high costs. Autonomous laboratories with full-process autonomous operation capabilities are the key carrier for achieving dry-wet closed-loop and reshaping the research paradigm.”
High-Value Scenario Applications
The plan identifies six priority domains for AI-powered research: high-energy physics, materials science, healthcare and life sciences, quantum technology, and biological breeding. In healthcare, for example, the plan envisions AI-powered drug discovery agents capable of target identification, molecular design, efficacy prediction, and safety evaluation. In materials science, intelligent characterization and simulation tools aim to dramatically shorten R&D cycles.
Scientific Model Ecosystem
A three-tier scientific model system is proposed, spanning foundational theory, general-purpose scientific foundation models, and specialized domain-specific models. The plan calls for continued iteration on existing models including Panshi — China’s first general scientific foundation model — DPA, the world’s first large atomic model for microscopic systems, and MegaDFT, a density functional theory model.
Academician Xu Bo, Director of the Institute of Automation at the Chinese Academy of Sciences, emphasized that “scientific models embedded with physical and chemical laws can achieve ultra-large-scale, long-period simulations that traditional computing cannot accomplish.”
Scientific Data Infrastructure
To address what the plan explicitly terms “breaking through scientific research data silos,” Beijing will establish a dedicated Beijing Science Data Center, develop scientific data management measures, and create intelligent data processing toolchains and standardized, reusable datasets aggregated from research projects, journals, and major science facilities.
Innovation Ecosystem
The plan outlines measures to cultivate interdisciplinary AI-science talent through the National AI Academy, launch international exchange initiatives including a “One Conference, Two Competitions” platform, build dedicated AI-for-science computing clusters, and strengthen Beijing-Tianjin-Hebei regional collaboration.
Expert Reactions
Liu Weihua, Deputy Director of the Beijing Municipal Science and Technology Commission, stated that the plan “is based on Beijing’s integrated development of education, science, technology, and talent, building a full-chain promotion system. Next, we will promote the implementation of tasks in a list-based, project-based manner.”
Academician E Weinan, Chair of the Academic Committee of the Beijing Academy of AI for Science, said the plan “coordinates various innovation infrastructure, and will provide a Beijing demonstration for the construction of national scientific intelligence infrastructure.” Academician Zhang Jin, Executive Vice President of Peking University, added that the measures “can promote the continuous ‘self-evolution’ of the scientific research system and enhance overall innovation capability.”
Analysis and Implications
The plan positions Beijing in direct competition with global AI-for-science leaders such as DeepMind, Microsoft Research, and academic centers in the United States and Europe. China’s advantages lie in centralized policy coordination, large-scale government investment, and a vast scientific workforce.
As Embodied Global noted in its analysis, the plan’s emphasis on autonomous labs and embodied robots represents a distinct Chinese approach to AI4S infrastructure, moving lab automation from a vendor-driven procurement category into a policy-supported infrastructure class.
However, challenges remain. Developing interdisciplinary AI-science talent, ensuring effective data sharing across institutions, securing sufficient computing resources, and navigating geopolitical constraints on international collaboration will all be critical to the plan’s success.
What to Watch For
With the plan now in effect through 2028, key indicators to monitor include the establishment of the first benchmark-level autonomous laboratories, progress on the Beijing Science Data Center, the development of the next generation of scientific foundation models, and the extent to which Beijing’s approach is adopted by other Chinese provinces and municipalities. The plan’s success or failure will offer important lessons for the global AI4S community as nations worldwide race to harness artificial intelligence for scientific discovery.