China MIIT launches AI service provider ecosystem drive

Beijing is quietly doing something most other governments have barely started to talk about: industrialising the “last mile” of AI deployment. MIIT’s new action plan is not about building one more foundation model, it is about creating a national pool of thousands of certified AI application service providers that sit between hyperscale models and real world enterprises. Targets of 2,000 providers by 2026 and 3,000 by 2027, plus explicit requirements around safety, security and compliance, amount to a state backed channel strategy for AI adoption across manufacturing, services and the public sector. ([miit.gov.cn](https://www.miit.gov.cn/zwgk/zcwj/wjfb/tz/art/2026/art_6fbc038bf15c445ab53b2a94a3f9d4e4.html))

Strategically, this could narrow the gap between China’s strength in deployment and its relative lag in cutting edge frontier models. By subsidising compute access, standardising “FDE” style field deployment roles, and using first purchase and risk compensation mechanisms to buy large model, agent and token services, MIIT is trying to turn diffuse experimentation into repeatable products. That is the kind of boring plumbing work that often decides which ecosystems really scale.

For global competitors, the message is that the race to AGI will not be won only in research labs. It will also be shaped by which jurisdictions build dense networks of mid tier integrators who know how to make models earn their keep in factories, hospitals and municipal offices. China is now explicitly trying to industrialise that layer.