The AI Plus initiative – China’s blueprint for AI diffusion

This note was originally published for subscribers of Trivium Tech Daily on August 27. It has been republished here with additional commentary and minor edits.

On August 26, the State Council published a high-level directive on implementing the AI Plus policy initiative. 

The directive is a big deal: It represents China’s most comprehensive blueprint yet for how it plans to develop and deploy AI domestically – and compete on AI internationally.

Beyond charting the economic roadmap for AI, it also offers some insight into whether China is actively pursuing Artificial General Intelligence (AGI).

We have been waiting for officials to put some policy flesh on the AI Plus initiative’s bones: Beijing first announced the initiative in the 2024 Government Work Report (GWR), and mentioned it again in this year’s version.

• But, until now, officials had not provided a comprehensive account of what the initiative actually covers.
• As it turns out, the answer is “pretty much everything.”

In this brief, provide a snapshot of each aspect of AI Plus as described in the directive, highlighting the interesting parts.

• First, though, we explain where the initiative fits in China’s overall economic and technological development strategy.

The 30,000-foot view: AI and economic development

China is entering a period of profound demographic contraction that will define its economic trajectory for decades to come.

• The country’s population peaked at roughly 1.41 billion in 2021 and is expected to shrink by around 100-200 million by 2050.
• UN forecasts suggest that by 2100, China could lose approximately 40% of its current population, shrinking to below 800 million.

The number of working-age adults will also fall over the next 10 years:

• Today, around 22% of Chinese citizens are over the age of 60.
• That is forecast to increase to 32.5% by 2035.

Babymakin’ isn’t coming to the rescue: Fertility rates are on a downward trajectory, despite Beijing abolishing the One Child Policy.

• China needs a fertility rate of 2.1 children per woman to maintain population replacement.
• That figure stood at a dismal 1.0 in 2024.

This shift spells the end of the growth model that China has so successfully pursued over the past four decades – built on cheap labor and exports, rapid urbanization, and a housing-fueled investment boom.

At the same time, China is edging into the economic stagnation danger zone that has tripped up multiple developing countries, raising the risk of the country falling into the “middle-income trap.”

• The leadership aims to avoid this trap by achieving an average GDP growth of at least 4.5% through 2035 (Hong Kong VPN required to access).

This raises the question: With fewer resources to work with, how can China maintain its growth trajectory to 2035 and beyond?

Xi Jinping’s solution lies in a concept the state is calling “New Quality Productive Forces” (NQPFs) – a strategy to boost total factor productivity.

Quick primer: “Total factor productivity” measures how efficiently a country (or company) squeezes output from its existing resources. It does this by quantifying the growth that cannot be accounted for by more resource input.

• For example, if a factory’s output rises after the workforce expands from 100 to 200 workers, that growth is explained by the addition of more labor, not by an improvement in total factor productivity.
• But if the workforce stays constant at 100 workers, working the same number of hours, and the capital stock remains fixed, yet output still increases, then the additional growth must come from improved efficiency – in other words, higher total factor productivity.

Total factor productivity can be improved in many ways, such as through:

• Technological innovation – advanced machinery, digital tools, and process automation
• Managerial and organizational improvements – more efficient logistics, workflows, and business practices
• Human capital development – stronger worker skills, training, and education
• Institutional and regulatory efficiency – a better legal system and more open markets
• Knowledge spillovers and network effects – the diffusion of ideas across firms, industries, and clusters

In summary, through the NQPFs initiative, Beijing aims for China to achieve more with less by holistically boosting efficiency through innovation, education, and institutional reforms.

• And since Beijing is going to have fewer resources in the coming decades, boosting efficiency is a matter of urgent necessity.

NQPFs will feature in the 15th Five-Year Plan (15th FYP), which will define China’s national policy priorities for 2026-2030.

• On April 30, Xi Jinping stated that “During the 15th Five-Year Plan period, developing New Quality Productive Forces…must be given a more prominent strategic position.”
• That language has since filtered down the state hierarchy, with officials frequently linking NQPF to the 15th FYP.

That likely means boosting total factor productivity will be a strategic priority for China for at least the next half-decade.

Enter AI, which has the potential to be the greatest general-purpose efficiency boosting technology of our age.

• And Beijing knows it.

Indeed, the State Council’s AI Plus directive opens by declaring that AI-powered economic transformation will:

• “Reshape the paradigm of human production and life, promote a revolutionary leap in productivity and profound changes in production relations, and accelerate the formation of a new intelligent economy”

An accompanying explainer from China’s macro planning agency (NDRC) doubles down, stating:

• “[The AI Plus plan] aims to elevate AI from a technical tool into a core engine for restructuring production factors, reshaping industrial paradigms, and reorganizing value chains.”
• “AI Plus can reshape the path to improving total factor productivity.”

In short, Beijing’s goal is for AI to serve as a new economic engine, keeping growth on track even as the country faces a shrinking workforce, potentially less capital, and rising geopolitical headwinds.

A deeper look at the AI Plus blueprint

The State Council’s AI Plus directive addresses AI development and adoption on multiple fronts.

• We break them down below.

1. AI as a tool for scientific discovery and innovation

The blueprint’s first stated goal is to equip scientists with AI tools to accelerate breakthroughs and unearth new discoveries. To that end, it calls for:

• Accelerating the development and application of scientific large models
• Upgrading major scientific and technological infrastructure with AI capabilities
• Creating open and shared high-quality, interdisciplinary scientific and technical datasets

Why it matters: This goal takes top billing because it is a force multiplier for all of the other goals outlined in the blueprint.

• As more breakthroughs are commercialized and adopted at scale, they promise to lift productivity and accelerate technological upgrading.

This is also interesting: The section on scientific discovery encourages philosophers and social scientists to investigate how humans and machines can collaborate more effectively – highlighting state support for research into cognition, judgment, and ethics.

The point: If we want to align AI systems with human values and train AI to do what humans want, we need to study humans.

2. AI as a tool for industrial transformation

Upgrading China’s industrial base to incorporate intelligent technologies is another efficiency-boosting strategy. As such, the blueprint highlights the need to support:

• “Intelligent-native [businesses] whose underlying architecture and operational logic are based on AI” – in other words, the state wants to see companies incorporate AI into strategic planning, organizational structures, and business processes
• Using AI in all industrial processes, from design and testing to production and customer service
• Using AI for forecasting and decision-making in industrial use cases

The directive also prioritizes agriculture and services for near-term AI deployment.

• For agriculture, the directive calls for AI-driven breeding and cultivation, the scaling of intelligent machinery, drones, and robots, as well as the use of AI in production management and risk mitigation.
• For the service sector, it pushes for new AI-powered business models, human-machine collaboration, and AI-powered devices, particularly in areas like finance, law, transportation, and logistics.

Why it matters: Strengthening China’s food security and boosting services consumption are also likely to be priorities in the 15th FYP.

3. AI as consumption booster

Speaking of consumption: For the last several years, policymakers have pursued a series of initiatives designed to boost domestic demand and juice household spending.

• That’s because, with the property market in crisis and the export outlook increasingly shaky, domestic consumption now has to carry more of the growth burden.
• But Chinese households just don’t spend as much as policymakers would like, especially on services.

Ultimately, policymakers are betting on AI-driven scientific breakthroughs and AI-enabled industrial production to power growth, but the efficiency gains of those efforts won’t likely become obvious for years.

• In the more immediate term, the state needs consumers to pick up more of the slack.

Thus, the AI Plus directive pushes for the development of AI-powered consumer technologies to stimulate demand for both services and products. The blueprint focuses on:

• Supporting the development of AI-upgraded services like entertainment, e-commerce, housekeeping, property management, travel, eldercare, and childcare
• Driving new and exciting consumer experiences through AI
• Developing new consumer products, like intelligent connected vehicles, AI-powered smartphones, robotics, smart home IoT, and wearables
• Integrating AI into new ecosystems, like the metaverse, low-altitude flight, and brain-computer interfaces, to galvanize new types of spending

Why it matters: Foreign tech firms take note – these are the areas where the state will welcome foreign investment.

4. AI as a support for humans, not just a replacement

AI should also enhance the quality of life and boost the efficiency of human production. The blueprint calls for:

• Using AI to develop smarter ways of working, allowing workers to collaborate more closely with machines and leverage intelligent AI agents.
• Incorporating AI into every aspect of learning and education to improve education quality, and “shifting the focus of education from knowledge transfer to enhancing human competence.”
• Ensuring humans live longer and better lives through the wide application of AI in healthcare, including by developing intelligent medical devices.

Why it matters: Smarter, healthier, more efficient workers will be less of a drag on productivity.

5. AI as a tool for government efficiency

The efficiency of any state – and particularly a top-down, centrally directed state like China’s – depends heavily on the capacity of the bureaucrats doing the directing. The AI Plus directive thus aims to harness AI to improve governance, calling for:

• Developing smart upgrades to city planning, construction, and governance to boost “intelligent urban operations”
• Developing AI-powered governance services to identify needs, plan responses, and handle processes end-to-end
• Deploying AI in public-resource bidding, procurement, and oversight
• Using AI for intelligent, cross-domain planning in land, air, sea, and space
• Applying AI to atmospheric, water, soil, marine, and biodiversity monitoring, as well as carbon-market oversight

This also caught our eye: The plan dedicates a paragraph to AI in law enforcement and national security, instructing officials to:

• “Enhance our ability to use AI to maintain and shape national security”
• “Promote the establishment of a multi-faceted and integrated public security governance system for people, digital humans, and intelligent robots”

What that means: We’ll get back to you on that one.

The plan also identifies AI as a tool for cyberspace governance, specifically for “strengthening capabilities such as accurate information identification, proactive situation analysis, and real-time risk management.”

• That likely refers to using AI to enhance online censorship, which China frames as a domestic security concern.

6. AI and international competitiveness

China aims to be a leading player in shaping the global development of AI. As such, the directive states that China will:

• Be collaborative, “treating AI as an international public good that benefits humanity, and fostering an open ecosystem for AI capacity building based on equality, mutual trust, diversity, and win-win outcomes”
• Make Chinese open-source tools widely available
• Be a champion of the Global South by “strengthening their AI capacity building, empowering all countries to participate equally in the intelligent development process, and bridging the global intelligence divide”
• Support the United Nations as the leading body for global AI governance

Why it matters: This strategy aligns with the ideas outlined in China’s Global AI Governance Action Plan and Global AI Governance Initiative, released over the past two years.

7. Strengthening the foundations of AI development

The plan also outlines priorities for strengthening the technical foundations and components that underpin AI development – namely, models and algorithms, high-quality data, computing power, technical talent, and the open-source ecosystem.

A. In foundational models, China will:

• “Strengthen research into fundamental theories of artificial intelligence [and] innovation in model infrastructure”
• “Accelerate research into more efficient model training and inference methods”
• Establish new frameworks for assessing model capabilities

What that means: China will attempt an end-run around US chip controls by building AI training systems that don’t require as much compute or energy.

B. To increase the availability of high-quality AI datasets, China will:

• “Improve data property rights and copyright systems adapted to the development of AI”
• “Promote legal and compliant open access to copyrighted content generated by publicly funded projects”
• “Explore models for cost compensation and revenue sharing based on value contribution”

Why it matters: The incentives for governments, private companies, and individuals to share and sell their data are currently poor, leading to data fragmentation and siloing – particularly for high-value datasets in government, science, and industry.

• By establishing legal mechanisms that ensure everyone can protect their economic rights to data and receive compensation for its use, the state is aiming to stimulate the data market.

C. To ensure China can develop its own AI compute over the long term, China will:

• Support innovation in AI chips and the development of supporting software ecosystems
• Continue working to interconnect China’s various computing power hubs so that national compute resources can be pooled
• Attempt to increase energy efficiency and improve data center utilization rates
• Encourage the development of scalable cloud services

What’s interesting: While it’s common knowledge that China is struggling to manufacture AI chips on par with those produced by Nvidia, Nvidia’s arguably bigger moat isn’t its chip hardware, but its chip software.

• Nvidia’s CUDA development environment is what enables AI companies to easily train models using Nvidia chips.
• While Huawei is trying to develop an alternative, its software is significantly lagging.
• With officials prioritizing the development of chip software, Huawei, or any viable competitors, will likely receive significant state support in this endeavor.

D. To ensure a steady supply of AI talent, China will:

• Incorporate AI into all levels of education
• Establish more AI-related majors
• Provide AI training for teachers
• Push companies that struggle to attract AI talent – typically industrial firms – to offer incentives like equity options to attract and retain qualified talent

E. Lastly, China will throw its weight behind the expansion of the open source ecosystem by:

• Supporting the development of AI open-source communities, models, tools, datasets, and projects
• Developing open-source projects and development tools with global reach and influence

This is also interesting: To incentivize students and researchers to participate in open-source projects, China will encourage universities to recognize open-source contributions as degree credits and reward contributions by faculty.

Why it matters: The success of DeepSeek’s open-source approach has convinced Beijing that open-sourcing technologies has the power to spur international adoption and increase the competitiveness of Chinese technologies.

• This thinking will result in increased state support for a broader range of open-source projects in strategic technologies, including non-AI tools such as chip architectures and operating systems.

8. Guarding against AI risks

The blueprint makes it clear that China intends to regulate AI systems and strengthen AI security. That means China will:

• “Guard against risks such as black box models, hallucinations, and algorithmic discrimination”
• Improve AI laws, regulations, and ethical standards, and advance legislation “related to the healthy development of AI”
• Tweak China’s algorithm registration system to ensure the state can monitor AI applications
• Develop emergency response systems for AI risks

Why it matters: How China strikes the balance between security and development will be definitive in determining whether the AI ecosystem flourishes or flounders.

Interestingly, the directive frames wasteful and unreliable AI investment as a risk to China’s AI ambitions.

• When state-backed funds and other investors pour money into weak or unpromising firms, it diverts capital from genuinely innovative players. And without stable financing, even the most promising AI companies face pressure to prioritize short-term survival over long-term innovation.

To mitigate these risks, China will:

• “Improve systems for assessing and evaluating state-owned capital investments in the field of AI”
• “Increase financial and fiscal support for the AI ​​sector, develop and strengthen long-term, patient, and strategic capital, improve risk-sharing and investment exit mechanisms, and fully leverage the role of fiscal funding, government procurement, and other policies”

Why it matters: If the state succeeds here, promising AI companies will have stable access to state funding, with smoother paths to IPO.

Is China pursuing AGI?

There is one glaring omission in the blueprint: It does not mention artificial general intelligence (AGI), superintelligence, or any other such concept.

Why that matters: There is considerable hand-wringing and tooth-gnashing in DC over whether China is actively pursuing AGI.

• Some pundits warn that China may secretly see beating the US in the race to AGI as its top technological priority – and that the US should respond accordingly.

However, taken at face value, the directive suggests that for China, the real “AI race” isn’t about beating the US to AGI supremacy.

• It is the race against the ticking clock of China’s demographic and economic challenges.
• Geopolitics and international competitiveness, while very important, are only part of the picture.

That’s not to say that China won’t release a policy on AGI – it very well might.

• We say that in part because Chinese tech policymakers tend to be hype-driven to some degree, and will swiftly issue policies that follow global tech trends.

Case in point: When Zuckerberg declared that the metaverse represented the bleeding edge of the technological frontier, China’s policy apparatus adopted the term within the year, releasing several local and national-level policies to spur metaverse development.

• However, those policies focused on supporting industrial applications for virtual reality (VR) environments, not on “beating the US in VR.”

The same thing could happen here: If the global hype machine continues to focus on AGI as the ultimate frontier of AI’s development trajectory, China may get a major case of FOMO.

• Plus, if the US is loud and proud in its pursuit of AGI, China will certainly not want to be caught on the back foot.

We also aren’t arguing that the state is not supporting research into AGI, or backing companies and innovators that do so.

• DeepSeek has clearly stated its intentions to pursue AGI.
• Multiple state-backed scientific institutions are also running projects related to frontier models.
• It’s highly likely that China is pursuing frontier AI research for military applications, and generally speaking, AI is already contributing to military modernization.

But the fact remains that, thus far, headline national policies currently prioritize AI diffusion – the productive uptake of AI across every sector of society – over the uncertain pursuit of AGI dominance.

• For AI to serve the economic growth function China needs it to serve, it must be adopted as quickly and widely as possible, and to maximum effect.

Will the AI Plus policy succeed?

All of the above raises a few final questions:

• Can China diffuse AI efficiently enough to achieve the necessary results?
• And, in terms of economic competition, can it diffuse AI more rapidly than competitors like the US?

Jeffrey Ding, assistant professor of Political Science at George Washington University and author of Technology and the Rise of Great Powers, argues that:

• “China faces a diffusion deficit because its capacity to innovate and pioneer new technologies far outstrips its capacity to diffuse those advances throughout the entire economy”

According to Ding, that’s because:

• “China’s top-down, centrally planned approach [is not as effective as the US] for diffusing general purpose technologies [like AI, because they] need to be adopted by small and medium-sized businesses in countless economic sectors, and that adoption must be driven by market forces to be efficient and fast”
• While China can train its top AI talent in top-tier universities, it lacks strong institutions for training ordinary engineers, creating a barrier to adoption outside of certain high-tech companies and institutions.

We think there’s some truth to this – and we’d add this to his argument: The current level of digital transformation in large industrial firms in China is relatively low, especially among state-owned enterprises (SOEs).

• In 2022, only 4.5% of SOE execs surveyed said that internal digital transformation efforts were “meeting expectations.”
• 60% said their enterprises were in the early stages of transformation, with digitization impacting only an isolated business unit or process.

Yet it is precisely these SOEs – the giants of heavy industry – that the state wants to lead the adoption of AI tools.

• Regardless of what the state wants, however, many of these companies lack a sufficiently high level of digitization to support the efficient and effective adoption of AI.

At the same time, this is also true: China’s robust ICT infrastructure and widespread adoption of digital platforms and payments lay the foundation for the rapid diffusion of new digital technologies among consumers, even if those technologies take longer to penetrate industrial and business processes.

We’ll leave you with one final thought: Even though economic efficiency is the end goal, no Chinese policy we are aware of, including this one, lays out clear targets or metrics for measuring AI’s contribution to economic growth.

• How widely must AI be diffused before efficiency benefits are gained?
• Will AI efficiency gains be outweighed by energy demands?
• Are there specific industries in which AI uptake would generate more benefits?
• How much efficiency does AI have to contribute before it offsets population loss or other headwinds?

China doesn’t know, and neither does anyone else, because there is no globally accepted standard for accurately calculating exactly how much of an efficiency boost AI gives to an economy.

• Chinese and US economists are still trying to quantify the scope of the digital economy 1.0 – AI will add a whole new layer of complexity.
• Until they sort that out, benchmarking the efficacy of this policy and subsequent AI policies will be exceedingly difficult, with results unlikely to be clear for many years to come.