China outlines national data infrastructure push

The policy, dated 31 December 2024 and published by the NDA in January 2025, defines national data infrastructure as a new type of infrastructure providing society-wide services for data collection, aggregation, transmission, processing, circulation, utilisation, operation and security. In plain English, this is the plumbing and rules needed to make data work reliably at national scale — not merely another data centre with a heroic name.

The framework combines hardware and software with model algorithms, technical standards, specifications and institutional mechanisms. It divides the job into four areas: data-circulation facilities, computing facilities, network facilities and security facilities. The timetable is equally explicit. China aims to run technical-route pilots and complete top-level design by 2026; support large-scale trusted data circulation and basic coverage of large and medium-sized cities by 2028; and broadly establish the system’s main structure and operating mechanisms by 2029.

The NDA’s 2026 China Data Industry Development Report, released at the China International Big Data Industry Expo on 28 August, puts China’s data-industry scale at RMB 6.78 trillion in 2025, up 15.7% year on year, with 482,000 data enterprises. The expo opened in Guiyang with more than 16,000 registered participants and 372 Chinese and foreign exhibitors, according to the NDA.

ChinaTechNews also reported figures attributed to an NDA survey or related official remarks, including a rise in external-data use for model training from 54% to 74.36%, a 142.58% increase in high-quality data volume, 199.48 exabytes of national AI training and reasoning data, and infrastructure spanning 15 industries and more than 50 cities. However, the underlying survey, report table, transcript or presentation was not available in the official NDA material reviewed. Those ChinaTechNews figures are therefore not included here as established facts.

AI gets the headlines, but AI is only as useful as the data it can access and trust. China’s plan targets unglamorous bottlenecks: common standards, secure exchange, computing capacity and rules for circulation. If it works, organisations should spend less time wrestling with incompatible systems and more time using data for services, research and industrial automation. If it becomes primarily a control layer, the same plumbing could make central oversight more efficient.

This is a serious infrastructure push, not proof that every headline about China’s AI data boom is true. The verified plan is detailed, and its 2026, 2028 and 2029 milestones give it something rare in tech policy: a schedule. Take the architecture seriously, but treat unsourced superlatives with a very large pinch of salt.

Sources: National Data Administration official notice and policy briefing: https://www.nda.gov.cn/sjj/zwgk/zcfb/0106/20250106095112713400492_pc.html; NDA expert interpretation: https://www.nda.gov.cn/sjj/zwgk/zjjd/0109/20250109162607676345998_pc.html; NDA China Data Industry Development Report release: https://www.nda.gov.cn/sjj/swdt/xwfb/0828/20260828233336307959824_pc.html; NDA expo report: https://www.nda.gov.cn/sjj/swdt/xwfb/0828/20260828213345032974683_pc.html; ChinaTechNews: https://www.chinatechnews.com/2026/08/30/128378-ai-drives-systemic-expansion-of-chinas-national-data-infrastructure-experts-say