On August 31, the General Office of the Ministry of Industry and Information Technology (MIIT) issued a notice on launching a special action to cultivate artificial intelligence application service providers. The initiative outlines that a national resource pool will be established, with the goal of strengthening providers' technical innovation, integrated delivery capabilities, and security compliance know-how.
The stated objectives span two phases. By the end of 2026, the national pool is expected to exceed 2,000 providers, forming a multi-tiered structure that ensures reasonable composition, clear division of labor, and collaborative innovation. By the end of 2027, the pool should comprise at least 3,000 providers, supporting a fully collaborative, end-to-end, and comprehensive AI application service ecosystem.
According to the notice, an AI application service provider is defined as an enterprise or institution that offers consulting, planning, delivery, implementation, operation management, and security governance services for AI-driven solutions. Pan Helin, a member of the MIIT's Information and Communication Economy Expert Committee, explained that these providers essentially empower traditional industries by delivering practical AI solutions.
The action plan is rooted in the State Council's "AI+" initiative announced in August 2025, which calls for the cultivation of application service providers and the development of new models like "Model as a Service" and "Agent as a Service." Pan noted that the special action aims to build a well-structured, collaborative multi-level provider ecosystem that matches the rapid expansion of AI applications, enabling large-scale implementation while improving service quality and supply.
Four core tasks underpin the initiative: establishing the provider resource pool, enhancing service supply capabilities, promoting large-scale adoption, and reinforcing institutional support. For the resource pool, regional authorities are required to survey local providers and build profiles. By the end of 2027, provinces hosting national AI industry innovation application pilot zones must have at least 100 enterprises in their local pools. The MIIT will consolidate these regional lists into a unified national resource pool and publish it periodically.
To enhance service supply, providers will be encouraged to lead "AI application service clusters" that combine computing power, data, and algorithm expertise, collaborating with upstream and downstream partners to create integrated solutions. Support measures include facilitating connections with national computing network hubs, interconnection nodes, and China's computing power platform. Providers can leverage policy tools such as "computing vouchers" to reduce infrastructure costs, while universities and vocational colleges are encouraged to co-build training bases with leading providers to produce versatile talent skilled in business operations, model comprehension, security, and delivery.
Providers are also urged to establish field deployment engineer (FDE) teams that work directly on-site with customers to ensure smooth implementation. Zhao Gang, president of the SAI Industrial Research Institute, noted that the action's core is to foster FDE-led providers that bridge AI technology and sector-specific knowledge.
On the adoption front, providers will be guided to package modular, standardized solution products targeting high-frequency, essential, and reusable business needs, creating a suite of "small, fast, precise, and reliable" AI offerings. Industry alliances, associations, and open-source communities will organize targeted supply-demand matching events to promote providers. Real-world scenarios will be opened to facilitate pilot implementations, enabling technology validation and demonstration of best practices.
A notable initiative involves encouraging "first purchase, first use" and risk compensation mechanisms to increase procurement of large models, agents, and token-based services, enhancing the effectiveness of AI usage across industries. In regions with the right conditions, comprehensive "going global" service systems will be developed, leveraging multilateral platforms like the Belt and Road Initiative, BRICS, and China-ASEAN cooperation channels to promote AI project exports.
Pan Helin elaborated that "first purchase, first use" refers to a preferential procurement system where government entities and public institutions, using fiscal funds, prioritize purchasing innovative products that enter the market for the first time without prior commercial track records—an approach that accelerates AI application innovation. Risk compensation, he added, uses fiscal subsidies and insurance mechanisms to share the potential risks buyers face when testing new technologies, effectively lowering "trial-and-error costs" and encouraging enterprises to embrace emerging tools.
Zhao Gang concurred, pointing out that large models and agent technologies are still nascent and may exhibit issues like hallucinations or security vulnerabilities. The "first purchase, first use" policy encourages government and large enterprise users to be early adopters, giving providers market opportunities, application scenarios, and validation cases. When risks materialize during trials, government-backed compensation helps mitigate losses, fostering a willingness to experiment.
These policy measures are designed to create demonstration effects and accelerate the diffusion of AI adoption across the economy.
Disclaimer: This article is for informational purposes only and does not constitute investment advice. Please verify all data before use. Any actions taken based on this content are at your own risk.