China's State Council Endorses AI Procurement: A Turning Point for Domestic AI

Honestly, when I saw the news yesterday afternoon, I almost spilled my coffee.

On April 21, China’s State Council issued the “Opinion on Promoting the Expansion and Quality Improvement of the Service Industry.” One sentence made me read it three times: “Deeply implement the ‘AI+’ initiative, accelerate the development and use of intelligent programming tools, and support the procurement of large model and AI agent services.”

This marks the first time at the State Council level that “large models” and “AI agents” have been explicitly included in procurement support lists.

What does this mean? My personal take: China’s AI industry has officially moved from the “encourage exploration” phase to the “government procurement” era.

Why This Matters

Here’s some context. Over the past two years, local governments have introduced various AI support policies. But most remained at the level of “encouraging innovation” or “building industrial clusters.” Companies seeking actual financial support had to apply for research grants or compete for talent subsidies—few policies directly generated orders.

This time is different.

The phrase “support procurement” means the government will start spending money to buy AI services. Not funding you to develop, but directly purchasing what you’ve built. For startups, this is huge—finally, a major client willing to pay.

The document specifically mentions “intelligent programming tools” and “AI agent services.” These happen to be among the most active areas in Chinese AI right now. Several friends working on AI agents told me they couldn’t sleep last night, they were so excited.

What This Means for Domestic AI

I see this from three angles:

First, demand-side activation.

Previously, government IT procurement focused on traditional software development and system integration. Now that AI services are officially on the procurement list, government agencies can legally procure AI products during digital transformation. This represents a massive market expansion for AI companies—intelligentization demand in government, healthcare, education, and finance is being unlocked.

Second, opportunities for domestic AI.

The document doesn’t state it explicitly, but government procurement has an unwritten rule: prioritize domestic products. This sends a clear signal to Chinese LLM companies like DeepSeek, Zhipu, Baichuan, and MiniMax, as well as enterprise agent developers. You don’t have to compete with OpenAI and Anthropic for overseas markets—plenty of opportunities exist domestically.

Third, accelerated industry standards.

Government procurement isn’t random; standards are required. What defines “qualified AI agent services”? How do we set performance metrics, security requirements, and service specifications? The process of establishing these standards will drive industry maturation. I expect to see a series of AI procurement standards emerge over the next six months.

My Concerns

Overall, I’m optimistic about this policy, but I have some concerns.

Government procurement can breed “relationship deals”—whoever has the best connections or lowest price wins. I hope technical thresholds are genuinely established, preventing AI procurement from becoming another “IT construction movement” where money is spent but nothing useful gets built.

Also, AI agent services remain in early stages; many products are still exploring. Can governments actually buy genuinely useful AI services? I’m cautiously skeptical. I recommend procurement departments conduct thorough POCs (proof of concepts) rather than getting swayed by fancy presentations.

Final Thoughts

This policy marks a watershed moment for Chinese AI. The transition from “wild growth” to “orderly procurement” brings both opportunities and challenges.

For AI companies, it’s no longer just about technology—they need to figure out how to serve government clients. For governments, the challenge is buying AI services that actually work, rather than being led by vendors.

I’ll keep following this development. Let’s wait and see the data.