China launches industrial internet and AI action plan to upgrade factories

Beijing’s new action plan is a reminder that, outside Silicon Valley, the most immediate use of powerful models isn’t chatbots—it’s manufacturing. The policy pushes “cloud-edge-end” industrial AI, high-throughput low-latency networks, and standardized industrial datasets, with explicit targets: over 50,000 enterprises upgraded to new industrial networks and high-quality data sets built across 20 key industries by 2028. ([finance.people.com.cn](https://finance.people.com.cn/BIG5/n1/2026/0119/c1004-40647913.html?utm_source=openai))

For the AGI race, this is less about flashy benchmarks and more about scale of deployment. Embedding AI across China’s enormous manufacturing base will generate petabytes of real-world operational data and millions of hours of agent interaction with complex systems. That data is exactly what you want to train more robust planning, control and multimodal reasoning systems—ingredients for “physical AI” that can one day run factories, robots and supply chains with minimal human oversight. At the same time, the plan’s emphasis on model “pools,” industrial intelligent agents and secure data circulation hints at a domestically oriented ecosystem that reduces dependence on foreign models and chips.

In geopolitical terms, this deepens China’s bet that leading in applied industrial AI can compensate, at least partially, for constraints on access to the very latest training GPUs. If it works, the country could end up with world-leading vertical AI capabilities in manufacturing even if it lags slightly on frontier general-purpose models.