A provincial government in southeast China has stopped using slogans about artificial intelligence and started writing numbers.
Fujian's new plan says how many industry models it wants, how many smart campuses it will back, and which departments must publish use-case lists each year.
It is the sort of document that quietly determines what gets bought, and from whom.
On 7 November 2026, the Fujian Provincial People's Government General Office issued the Fujian Action Plan for Promoting "AI Plus" (人工智能+) (2026–2028), document number 闽政办〔2026〕34号. It was published on the provincial government portal on 12 November. The plan translates a national directive into fifteen tasks across six areas, and it attaches numbers to many of them.
Key takeaways
- Fujian issued its AI Plus (人工智能+) Action Plan (2026–2028) on 7 November 2026, implementing the State Council's 2025 opinion on the national AI Plus initiative.
- The plan sets fifteen tasks across six areas and attaches targets to many: 60-plus industry vertical models, 100-plus typical application scenarios, 50 quality AI products and 100-plus high-quality datasets.
- The document is already in force: it takes effect on the date of issuance and runs to 31 December 2028. It is not a consultation paper or a draft.
- Leading departments must open up scenarios and publish at least ten benchmark applications a year, supported by compute vouchers (算力券) and a provincial AI industry fund.
- The plan shows how a broad national slogan becomes local procurement: counts, named departments and an expiry date.
Four stages, and where this sits
Policy documents in China pass through recognizable stages, and confusing them changes what a reader thinks has happened.
A consultation draft (征求意见稿) is circulated to gather comments. A draft (草案) is a more settled text that may still change. Formal adoption or issuance (正式通过/印发) produces an authoritative document. Effective implementation (生效实施) is when it actually governs conduct. Some documents take effect only after a separate date; this one says it takes effect from the day it is issued.
Fujian's action plan belongs in the fourth category. It was formally issued by the provincial government office, published on the government portal, and states that it runs until the end of 2028. It is not open for comment and it is not a proposal.
The national document behind it — the State Council's Opinion on Deeply Implementing the "AI Plus" Initiative (国发〔2025〕11号), issued in August 2025 — is also an issued policy, not a draft. Fujian's plan is the implementation layer beneath it. That distinction matters: the national text sets direction, while the provincial text names quantities, departments and deadlines.
What the numbers actually commit to
The plan covers science and technology, industry, consumption, public services, governance and international cooperation. Its targets are unusually specific.
In research, it promises ten AI research and innovation platforms and ten key AI technology projects. In industry, it calls for 50 quality AI products and more than 100 typical application scenarios, ten smart farms or ranches a year, and at least ten agricultural agents. In education, it aims for more than 100 smart-campus pilots, 100 typical teaching cases and 100 quality online courses.
On data and models, the plan wants 60-plus industry vertical models and 100-plus high-quality datasets in fields such as marine resources and low-carbon themes. Transport targets include 100 smart toll stations and upgrades to 83 service areas. Government-service and ecological-governance sections each call for more than ten application scenarios.
A separate clause requires leading departments to open industry scenarios and produce at least ten benchmark applications a year that can be copied elsewhere. That clause is the bridge between a target and a purchase order: a number with a responsible department attached tends to become a procurement plan.
From slogan to shopping list
The most useful way to read this document is as a map of demand. "AI Plus" as a national phrase is broad enough to mean almost anything. A provincial plan that names 60 vertical models and 100 datasets tells suppliers what a local buyer may soon be looking for.
The plan also names the supporting ingredients: compute vouchers, a provincial AI industry fund, national pilot bases for testing, and a requirement that cities and counties reuse provincial computing and model resources rather than build their own. That last point matters for vendors. A province that consolidates computing into a shared platform is signalling that it wants a few large deployments, not hundreds of small ones.
For companies selling into the region, the practical unit is no longer "an AI strategy" but a scenario that fits one of the numbered categories — and a department listed as responsible for it.
Honest limitations
This article is based on the official text published by the Fujian provincial government, together with a CCTV local report and an English-language account by China Daily. These are authoritative for what the document says; they are not evidence that the targets will be met.
Plans of this kind typically publish counts and deadlines but not budgets. I have found no line item in the sources that prices the 60 models or the 100 scenarios, so this article makes no claim about spending. It also cannot tell you which contracts have been signed.
The targets are stated as minimums — "more than" or "at least" — and the English title uses "60-plus" and "100-plus" to preserve that. A target is a goal, not an accomplished result, and nothing here should be read as Fujian having already built these systems.
The document is a provincial implementation of a national opinion. It does not change national law, and it sits below rules on data security, personal information and generative-AI services that operate independently. Chinese policy names, document numbers and concepts are kept in their original form where translation could blur their meaning.
What readers can do now
- If you sell AI services, read the department column, not just the targets. A numbered goal with a named department is a more reliable buying signal than a headline figure.
- If you work in Fujian, check whether your sector appears in the fifteen tasks. The plan rewards scenarios that can be replicated, so a pilot designed for reuse has a better chance of funding than a one-off demo.
- If you follow Chinese policy, keep the four stages straight. A plan that is issued and in force is a different commitment from one still circulating for comments.
The striking thing about Fujian's document is not its ambition but its shape. It turns a national idea into a list that an office can tick off. Whether that produces useful AI or merely busy procurement is a question the plan itself cannot answer.
