Politburo Meeting Urges Deep ‘AI+’ Implementation as Core AI Industry Tops RMB1.2 Trillion in 2025

A meeting of the Political Bureau of the CPC Central Committee called for deep implementation of the “AI+” initiative, cultivating new forms of the intelligent economy and improving the AI governance system. It is estimated that China’s core AI industry scale exceeded RMB1.2 trillion in 2025.

The 2026 World Artificial Intelligence Conference and High-Level Meeting on Global AI Governance was held in Shanghai, with the total exhibition area exceeding 100,000 square metres for the first time. Over 1,100 enterprises participated, showcasing more than 3,000 exhibits, among which over 300 products made their global debut. The two major tracks of intelligent computing and embodied AI each gathered over 200 companies. Dozens of leading supply-chain anchor enterprises, central and state-owned enterprises, and top foreign firms exhibited together with their ecosystem partners, highlighting cutting-edge practices of AI empowering various industries.

Manufacturing is the main battlefield for AI applications. Currently, the adoption rate of AI in China’s large industrial enterprises has surpassed 30%, and humanoid robots have started entering factories and workshops. China Mobile displayed a 1:5 scale flexible manufacturing island workstation. Through digital-twin design, synchronous iteration of production, and intelligent scheduling, it meets the high-flexibility, high-efficiency requirements of small-batch, multi-process production. The workstation is suitable for multi-variety, small-batch, low-takt production lines, as well as new vehicle trial production and export model production, and can be extended to assembly scenarios in discrete manufacturing industries such as rail transit and aerospace. China Unicom’s intelligent manufacturing line replicated the full production line process from phone quality inspection to packaging, while simultaneously building a high-precision digital twin training ground. Collaborative training, parameter optimization, and risk verification are completed in virtual space, and the optimal solution is then deployed to the real production line. The solution has been implemented at a leading enterprise, significantly shortening the production line iteration cycle and helping the company reduce line trial-and-error costs by 30%.

Artificial intelligence is accelerating its integration into daily life. Shanghai has created 24 AI-themed scenarios, including “Model Speed Space”, “Model Power Community”, an agent town, ultra-HD audio-visual, and “AI+Beauty”, and launched experience projects such as robot flash mobs, digital human interactions, and AI performing arts. ZTE Nubia publicly unveiled the world’s first AI Agent smartphone, positioned as a mass-production flagship that reshapes human-machine interaction, enabling users to have AI execute complex tasks across apps with a single voice command. Lenovo showcased achievements including an enterprise super agent and a city super agent. Its AI PC-embedded personal super agent “Tianxi AI” and the Lenovo Baiying AI host’s built-in Baiying Claw are ready to use out of the box. Dewu App, in partnership with Qianjue Robot, created the world’s first AI authentication robot. Taking sneakers as an example, the robotic arm picks up the product, high-definition cameras collect data, and image recognition technology transmits data in real time to an “AI brain” for pixel-level comparison, generating an authentication report within five seconds. In addition, 169 large models in China have completed regulatory filings, providing foundational innovation capability support for various new business formats.

AI infrastructure is accelerating its evolution. Moore Threads introduced the MTT C256 super node, which adopts an integrated computing and switching design to increase GPU deployment density in intelligent computing centres, creating a computing power base for hyperscale, stable long-term training and inference clusters at the ten-thousand-card or even hundred-thousand-card level. Tsingmicro’s REX81 Supernode, a domestically developed 4K super node system, reduces inter-node interconnection costs by 90% through an innovative computing-power mesh interconnect architecture. Taichu (Hangzhou) Integrated Circuit Co. , Ltd. ’s hyper-converged intelligent and scientific computing system, Yuanji HyperIntelliX, is compatible with both trillion-parameter large-model training and AI for Science scenarios. It employs a new-generation heterogeneous many-core AI chip and can seamlessly scale to a hundred-thousand-card computing scale.

Wang Ruomeng, Deputy Director of the Innovation and High-Tech Development Department of the National Development and Reform Commission, stated that during the 15th Five-Year Plan period, China’s AI development will strengthen key technology breakthroughs in models, computing power, and data, increase basic research, focus on areas with high economic contribution, strategic value, and social benefits, open up high-value scenarios, build benchmark applications, and continuously promote deep hardware-software adaptation, co-prosperity of open-source and closed-source models, and integration of industry, academia, research, and application, thereby enhancing ecological synergy.