Beijing’s municipal government said on 7 September that the Beijing Economic-Technological Development Area had issued an AI4Chip action plan for 2026–2028, aimed at applying artificial intelligence in the integrated-circuit sector. The plan, described as “AI4Chip 20 measures,” contains five action areas and 20 core tasks. By 2028 it aims to cultivate three to five AI4Chip ecosystem companies with international influence and create more than 10 applications suitable for wider industry promotion. These are future policy targets, not current company counts or completed application totals. The announcement says the area wants to apply AI across a chain that runs from chip design through manufacturing, packaging and testing, equipment and materials.

The plan’s action areas cover AI-native foundations, AI-assisted chip design, manufacturing and packaging-testing, equipment and materials, and the industry ecosystem. On design, it highlights AI chip architectures, automation across the design process, lifecycle management and coordination between design and manufacturing. For production, it points to AI-enabled fabrication lines, smart packaging and testing systems, and factory automation. The equipment and materials section proposes digital twins, edge-computing tools, virtual simulation and AI agents for research and development. Ecosystem measures include dedicated computing resources, industry datasets, specialized models, standards, workforce development and public services. This list describes policy scope; it does not supply measured productivity gains or cost reductions.

The district says it plans to use central and local public funding, computing and model vouchers, and project calls to support the work. It reports more than 400 integrated-circuit-related companies in the area and an industry scale above 130 billion yuan; those figures are background on the district, not results of the AI4Chip plan. Implementation could face constraints involving data access, model accuracy, hardware-software compatibility, security and integration into engineering workflows. For example, design-automation suggestions would still need manufacturability and reliability testing. The announcement does not specify funding allocated to individual firms, which products will use the applications or how the targets will be measured. The numerical goals should therefore be read as management objectives for future assessment, not completed industry output.