China’s next five-year plan: AI moves from slogan to system

Every five years, China develops a new plan to outline the country’s economic and developmental priorities, and implements it to build out expectations and steer the course. The 14th Five-Year Plan (2021–2025) is now closing, so we are entering the ‘handover’ phase, and drafting for the 15th plan that will run 2026–2030 has begun.

Each half decade marks an ever greater signaling moment, where China holds its cards ever so slightly less close to the chest and decides how to commit, or pivot, depending on how fair the winds (and the track record) feel, to the future.

What gets written now into these plans doesn’t stay on paper, it moves into action and ultimately shapes the prices of goods, the phones and cars you may purchase in the global marketplace, and the rules of global competition.

AI in center seat

And what matters now is not the page-turn, but that pivot: Beijing is moving artificial intelligence (AI) from a priority on paper to a production system in the real economy. In late August, the State Council released an “AI Plus” directive that frames AI as general-purpose infrastructure, to be embedded in scientific research, manufacturing, services, public welfare, and governance, rather than ring-fenced inside big tech alone (State Council “AI Plus,” Aug. 27, 2025). The timing, the ‘pivot’, is not an accident, and policymakers are using the last months of the current plan to lock in the narrative that the next five years will be about: technological diffusion, domestic capacity, and insulation from external shock.

The phrase that stitches this together is already familiar in Beijing policy circles, yet may be novel for more curious ears: “new quality productive forces.” This is an ideological label, for sure, but the substance is very much practical.

The National Development and Reform Commission (NDRC) defines it as advanced productivity that leaves the old growth model behind, to prioritize high tech, high efficiency, and high quality, all aligned with a new development philosophy. We can read it as an “AI green light”, a mandate for AI, robotics, green manufacturing, and upgraded supply chains to absorb more of the growth burden (NDRC). If the last Five-Year Plan was about building platforms and scale, this upcoming cycle will be all about embedding intelligence into physical industries and public services.

What shifts in the 15th plan

A plan is not a line-item budget, rather, it is a policy framework that tells actors at the ministries and provincial level what to prioritize. And ahead of publication, we can already discern several key shifts:

First, application over hype. The “AI Plus” directive does not glorify or glamourize new tech adoption, instead directing it in ordinary places: traffic management, construction codes, factory scheduling, hospital triage, school curricula, agricultural yields, and municipal services. The goal is productivity and resilience, not headline demos, and that design language signals a calculated bet on cumulative impact, many small wins across millions of, yes, mundane workflows, rather than one showy moonshot.

Second, hardware sobriety and domestic substitution. U.S. export controls have turned chips into a national vulnerability for China. In response, regulators and firms are accelerating a domestic stack. This month’s reporting that Alibaba and Baidu are training models on their own accelerators, partially substituting for Nvidia, fits the direction of travel. It is not yet a full replacement, and it is not painless, but it shows where the state wants the market to go (Reuters). This move has been quickly followed up with an outright ban on Chinese firms purchasing Nvidia chips (FT). We can expect the plan to ‘cushion the impact’ by pairing incentives for indigenous tech with standards for energy-efficient data centers and an overall push for open-source model ecosystems that can run on lower-tier hardware. In effect, a move to quickly wean off external dependency.

Third, productivity politics. China’s leadership has framed growth in 2026–2030 as a test not just of stimulating demand, but of governance. “New quality productive forces” links, to a reasonable extent, party legitimacy to whether AI and automation can lift total productivity. That is why the plan will keep AI close to state priorities, think smart manufacturing in the rust belt, service quality in provincial capitals, logistics efficiency along inland corridors, and precision upgrading for thousands of mid-sized firms. This is not Silicon Valley’s move-fast culture. It is not aiming to be ‘cool’, but it is a cool, bureaucratic diffusion, laid out with targets that embed AI across the bedrock of economic society.

Where AI meets the real economy

The plan is likely to concentrate on three practical fronts.

Manufacturing upgrades. China will push computer-vision quality control, predictive maintenance, digital twins, and scheduling optimization into the factories that make everything from machine tools to textiles. Much of this may seem boring on the surface, but it makes all the small moves count: defect rates, energy use, throughput, and delivery times could all improve in ways that compound efficiencies. Think less chatbots, more sensors and scheduling. Provinces will be evaluated on adoption rates, vendor ecosystems, and export performance in “smart” product categories.

Public services and the city stack. AI has obvious value in healthcare triage, chronic disease management, and radiology workflows. In education, it can do a great deal to close gaps in remote regions. In cities, the target is smoother traffic, safer worksites, and carbon-aware grids. These are not shiny proofs of concept; they are both rural and municipal KPIs. Expect the plan to tie AI adoption to public-sector procurement reform and to mandate data interconnection that has often been hampered by local blockages.

Energy and climate constraints. Data centers have widely come under scrutiny for the energy involved in drawing power and water, and AI model training has a non-trivial carbon footprint, to say the least. The next plan will likely require greener power for computing clusters, a closer coupling between AI build-outs and renewable generation, and critically, standards for measurement and labeling. The “AI Plus” release already paired adoption with a content-labeling rulebook, which is a sign that environmental metrics could be codified alongside safety and traceability in the coming cycle (State Council).

Tech sovereignty is an economic strategy, not just a security claim

AI is the emblematic tool, but the structural aim is risk management. The last three years taught Beijing hard lessons about choke points: advanced graphics processing units (GPUs}, electronic design automation (EDA) tools, lithography (the process used to manufacture integrated circuits on chips), firmware, and high-end memory. It’s important to note this will not happen overnight, the substitution across the stack will be partial and uneven, and the plan will seek to take any step that reduces single-point exposure.

Severing the link to external dependency sits at the point of current exposure, and incentives for domestic fabs and accelerators are likely one path. Another is changing the kind of AI China promotes, favoring architectures and training methodologies that can live with less exotic hardware, a point that calls back the historic reputation of ‘dependable, no-frills’ Soviet space tech, built for resilience and results. A third area is supply-chain diplomacy, pairing capital flows with partner countries that host assembly, testing, and component niches. A well-regulated form of outsourcing, if you will.

The message to global firms is simple: the next five years are designed to de-risk inputs and to expand exportable AI-enabled products. For governments, it means China will be both a market and a rival in AI-heavy sectors from industrial robots, connected vehicles, to applications that may only be discernible in coming years.

What this means outside of China

For exporters and investors, the story is the risks of diffusion, and the associated opportunity. If China succeeds in scaling AI into midsize manufacturers, the export mix changes. Expect a wave of ‘smart enough’ equipment and turnkey factory packages aimed at the Global South. Buyers in Southeast Asia, Africa, and MENA will compare time to value, to speed of access, not ideology. The white goods, consumer phase of the Belt and Road Initiative.

That shift won’t stop at the factory gate, it will determine what shows up in global markets, from more affordable AI-embedded electronics to domestic appliances and personal transport choices.

For competitors, the risk is the productivity dividend in traditional industries. If a million small factories squeeze out two to five percent more product with lower defects and faster delivery, market share will shift. Trade partners will respond with their own industrial policies and standards for safety, IP, and related environmental impacts. It’s likely we could see a self-enforcing standards race.

For climate, however, there is a certain tension. Certainly, AI can help optimize grids, smooth demand, and cut overall resource inputs, but it can also increase the digital/data footprint. The next plan will try to claim the first while containing the second, with frequently reported efficiency gains. Internationally, watch whether Chinese firms bundle AI with solar, storage, and grid equipment in development financing, to bridge the gap from ‘technology policy’ and entwine it with outright foreign policy and power projection.

The plan is a signal, not a guarantee

Five-year plans seek to set a direction to steer in, while physical delivery resides in budget realities, local implementation success, and commercial end viability. There are obvious implementation risks: hardware gaps, ongoing debt overhang in local governments, pushback from private-sector caution, and, of course, demographic drag. There is also a global context: export controls tighten and loosen, and supply chains must reroute in response.

Notwithstanding the challenges, the intent is clear. China wants AI to be an ordinary tool of production, not an elite product launch. It wants models it can afford and applications that help to raise the floor, critically, across provinces and sectors. A digital closing of the gap akin to the great reduction in nationwide poverty achieved between the 1980s and 2015. The 15th Five-Year Plan will read like a catalog of that ‘harmonizing’ intent, and the AI Plus order is the preview.

How to read the next five years without getting lost

We can treat the coming plan as a map of where China wants capital and talent to flow, and there are three kinds of signals to watch for as the text lands and provincial plans follow:

1. Budgeted infrastructure, not just targets. If computing clusters, power upgrades, and logistics corridors show up with funds, the adoption story is real.
2. Procurement rules, especially in health, education, transport, and utilities. That is how diffusion will be implemented and enforced at that ‘basic’ level.
3. Hardware pragmatism, including acceptance that ‘near-frontier’ models are not required in order to obtain the most value. If domestic accelerators plus efficient software are “good enough,” adoption will scale.

If you are looking for the geopolitical angle, it is hiding in plain sight. The next plan treats AI as a lever for resilience at home and influence abroad. That is industrial policy by design, and it is how China intends to compete in a fractured global economy.

Read this. Notice that. Do something.

Read this: A concise rundown of the “AI Plus” directive from Aug. 27, 2025, and why it matters for diffusion across economic sectors and Chinese society (IAPP summary). For the policy frame behind it, take a look at the NDRC’s definition of “new quality productive forces” (NDRC). For a more sober preview of the 15th Five-Year Plan priorities, start here (China Briefing overview).

Notice that: Chinese platforms are beginning to train on domestic chips, a practical response to export controls. It is partial, but directionally important and building steam (Reuters). Standards and procurement will be as important as models, as will, of course, the time taken to adapt, a weakening force on U.S. leverage.

Do something: If you track China exposure, build a simple list of three key items: provincial AI adoption targets, public-sector procurement rules, and announcements of domestic computing projects. Pair these with an external comparator like the Stanford AI Index to separate the narrative blurb from actual, measurable progress (Stanford HAI AI Index).

And remember, the visible effects won’t just be in policy documents, they will be seen in product availability, pricing, and also the standards labeling on our everyday goods.

Previously on GYST: Ethiopia’s dam gamble.

Next up: Pocket power: how China is rewriting the AI rules