National Orchestration and Provincial Competition: China’s Industrial Policy for AI Dominance

America has no peer competitor in artificial intelligence outside of China. Several years ago, when the trajectory of AI model development was uncertain enough to allow middle powers like the United Kingdom and France to nurse hopes of remaining competitive, this may have sounded like an overconfident pronouncement. Today, it is well-earned conventional wisdom. Competing at the frontier of AI requires coordinating talent, data, energy, and compute infrastructure at unprecedented scales—scales that only the United States and China are realistically capable of delivering on, although each in their own distinctive ways. Understanding China’s approach to developing and diffusing AI is thus of existential importance to understanding America’s relative position in the world to come.

America has been surprised by China’s AI prowess before. In January 2025, the release of DeepSeek was widely described as a Sputnik moment by policymakers and business commentators. While DeepSeek’s technical advances were overstated, the media’s reaction revealed the extent to which many in the United States had become complacent about China’s lag in AI capabilities.¹ Against available evidence, too many American observers believed that China was incapable of discovering AI breakthroughs on its own, whether because of constraints on their access to advanced semiconductors, or the persistent myth that Chinese companies can only copy but not innovate. Even now, many still seem to believe that Chinese AI models will remain behind American models in perpetuity, offering lesser capabilities but at a fraction of the price. Yet offering a good enough product at ultra-low prices and thereby cornering the market on less exquisite technologies and manufacturing inputs is exactly how China became a peer competitor to the United States in the first place. In AI, we are thus primed to be surprised once again.

The pervasive indifference that characterizes America’s overconfident view of its place in the AI race stems from grading the Sino-American AI race against our own preferred rubric: frontier model benchmarks, the scale of the data center buildout, and timelines to artificial general intelligence (AGI). Rarely do we measure American performance against the categories that the Chinese themselves choose to emphasize. The party-state and various Chinese companies are clearly trying to unleash AI capabilities, and Beijing’s desire for international AI leadership is beyond dispute. But their methods and benchmarks of success are different from ours, evincing a fundamentally distinct understanding of the nature of the competition.

American readers inclined to dismiss China’s focus on open source AI diffusion and applications as a case of settling for less than the frontier should consider an alternative interpretation: that the Chinese state has made a sincere and potentially well-founded judgment about where the benefits from AI development will accrue in the medium- to long-term horizon, and its leaders are organizing the many arms of the state to support their industry ecosystem accordingly. The primary questions explored by this essay are (1) how that AI-focused industrial policy is orchestrated, and (2) what the subnational dynamics between China’s provinces, municipalities, and central government reveal about their model of AI development, for which America has no equivalent.

China’s AI Division of Labor

Before describing Chinese industrial policy for AI, it is necessary to explain two paradigmatic differences in the ways that the Chinese and American governments perceive AI development and diffusion, as well as how those different perspectives influence tangible policy outcomes.

The first is the difference between how the two countries approach hardware versus software. The United States has myriad regulatory barriers to physical infrastructure buildouts that coexist with an engrained hesitation to regulate algorithms and models. China is almost the inverse. The PRC actively regulates the algorithmic layer of AI, requiring registries of proprietary data and, in some cases, imposing “ethical committees” to oversee algorithmic usage.² At the same time, the Chinese state aggressively organizes and subsidizes physical infrastructure and deployment: data centers, compute vouchers, industry funds, procurement mandates, start-up incubators, and more.

In contrast, the U.S. federal government maintains a relatively light-touch approach to software but is mired by legal and regulatory constraints on physical infrastructure, including power generation, transmission lines, data centers, heavy industries, and until recently, chip fabrication. It is too simple to frame this as American lawyers litigating the physical buildout of AI while skilled Chinese engineers speed ahead at constructing power generation, transmission, and roboticized factories. Nevertheless, this asymmetry is a good starting place for understanding how the respective AI strategies of the United States and China diverge.

Second, high-level discourse in China concerning AI’s technological potential is conceptualized quite differently relative to the English-language AI community. In the United States, the AI race is widely seen as a sprint to AGI, an autonomous system capable of outperforming expert humans in virtually every domain. Developing AGI in a way that benefits humanity is the explicit mission of OpenAI, for instance, reflecting the influence of early AI safety thinkers from the Effective Altruist (EA) and rationalist communities, in particular. While AI development is proceeding along a continuous spectrum, AGI is considered a particularly momentous threshold, beyond which progress rapidly accelerates toward superintelligent systems with the potential to transform every aspect of our economy and society, while giving the first company or country to achieve AGI decisive economic and military advantages.

While the discourse in America regarding the correct path for AI development is uniform—AGI is brought into existence as an emergent property of scaling LLMs and related infrastructure—the same cannot be said for China. To be sure, some Chinese researchers and thinkers do share this perspective.³ Especially at model-developing start-ups, such as Moonshot and Zhipu, where each company’s CEO has explicitly stated that his mission is to achieve AGI through LLM scaling, there is a symmetry between the American and Chinese perspectives. Another perspective is the idea of Embodied AI (EAI), a viewpoint which has been mentioned in recent high-level national documents, such as the Fifteenth Five-Year Plan. Chinese proponents of the EAI perspective tend to believe that scaling LLMs is not the path to AGI.⁴ Instead, EAI advocates see integrating model development with physical applications, such as robotics and self-driving cars that have self-improving capacities from interactions with the tangible world, as the best path toward AGI.⁵

But in practice, as well as in policy, perhaps the most influential Chinese perspective is the one that treats AI as a general purpose technology, much like electricity. This view is best articulated in Taiwanese scientist and entrepreneur Kai-Fu Lee’s popular 2018 book AI Superpowers, where he describes AI in these terms: “Often, once a fundamental breakthrough has been achieved, the center of gravity quickly shifts from a handful of elite researchers to an army of tinkerers—engineers with just enough expertise to apply the technology to different problems.”⁶ In this case, the power of AI will accrue not to the most sophisticated model developers but instead to those who find novel applications for the technology. This is a view that explicitly rejects imminent AGI. Lee himself describes it as decades or centuries away, if it is feasible at all.⁷ It should not be assumed that the Chinese government agrees with this assessment, and there are new reasons to believe that party officials are taking the prospect of superintelligence seriously.⁸

Rather, the central government should be seen as not having yet decided on any single approach, indulging a range of development paths through national-level and subnational policies. The manifold arms of the state have been orchestrated to support AI development and diffusion, but we have not yet seen what Chinese industrial policy for a single AI development strategy would look like.

When it comes to China, America’s worst habit is assuming that they think and behave just like us. What was the misguided belief that China would liberalize politically as it developed economically if not a projection, or mirroring, of our own attitudes onto a country and civilization with its own distinct internal logics? If the United States is going to truly compete with China in the race to achieve AI leadership, as we believe it should, and gain the technological edge that seems to be contested only between our two countries, then we need to understand exactly what we are competing against.

How the Central Government Sets Direction

China’s ambition to be a leader in AI development and diffusion has been clear for more than a decade. Over that same period, the PRC’s AI industrial strategy has moved from bureaucratic jargon to concrete policy implementation. National-level state interest in AI dates to the 2015 “Internet+” Action Plan, which first mentioned artificial intelligence in a national policy document and particularly emphasized physical integration in automobiles, factories, and homes.⁹ The Made in China 2025 Plan, released in 2015, set specific goals and deadlines for Chinese industrial upgrades across a range of technologies, many fundamentally connected to the development of a globally competitive yet indigenous AI stack.¹⁰

Two years later, the 2017 New Generation AI Development Plan set the ambitious, if not entirely specific, goal to “build China’s first-mover advantage in the development of AI” and to be a “global power in science and technology.”¹¹ The 2021 Fourteenth Five-Year Plan included AI among its list of frontier industries, a general term indicating that the technology, among others, is of importance to the party.¹² Since the last five-year plan, however, the state’s attitude has clearly shifted from indefinite support toward more targeted goal setting.

Following the release of the summer 2025 White House AI Action Plan, the Chinese State Council released its own AI Plus plan in August.¹³ AI Plus doesn’t mention competition at all, nor does it even mention America. Its focus is on diffusing and implementing AI in science, manufacturing industry, and administrative duties, including government.¹⁴ The plan sets specific metrics and timelines for this diffusion: 70 percent “deep integration of artificial intelligence (AI) with six key areas” by 2027 and application integration of 90 percent in these areas by 2030.¹⁵ A readthrough of the AI Plus plan also hints at the different policy levers used by the central government to orchestrate interprovincial cooperation and competition, such as the “Eastern Data, Western Computing” initiative.

Finally, by March 2026, the National People’s Congress approved the Fifteenth Five-Year Plan, which mentions artificial intelligence more than fifty times, nearly five times the frequency of the 2021 plan.¹⁶ By now, China’s AI ambitions have become legible and concrete. Beyond stating that the state will “fully implement the AI+ initiative,” the plan also sets out ambitions to “seize the commanding heights of AI industrial applications,” and “Accelerate breakthroughs in fundamental theories and core technologies of artificial intelligence.”¹⁷ The emphasis on artificial intelligence and the rhetoric around self-reliance in technology, especially for chips and fabrication technology, is well understood in English-speaking media. There’s no reason to take these documents at face value, but we should question why the Chinese narrative seems to ignore the race-to-­dominance framework. Additionally, the exact mechanisms China uses to support innovation across the national and subnational levels, as well as the relationship between the many layers of government and actors within it, remain underdiscussed. At the central government level, three mechanisms deserve special attention: state funding structures, orchestrated infrastructure buildouts, and algorithmic standards-setting.

State funding from China’s central government flows through a layered system designed to align provincial incentives with central priorities. The National AI Industry Investment Fund, seeded at 60 billion RMB ($8.2 billion) and overseen by the Ministry of Finance and the Ministry of Industry and Information Technology (MIIT), operates with provincial matching requirements.¹⁸ Additionally, the Bank of China announced plans in January of last year to launch 1 trillion RMB in funding to support the full AI stack.¹⁹ While requirements and specifics vary, provinces that want access to national capital usually need to commit their own funds. Additionally, provincial money, once committed, creates an incentive for local officials to demonstrate results. The competition for AI capital is thus not only between firms competing for investment but between provinces competing to be the most attractive local jurisdiction for that investment.

Beyond channeling funding to provinces, China’s central government also orchestrates major infrastructure buildouts. The best example in the AI context is the aforementioned program of “Eastern Data, Western Computing,” formally the National Integrated Computing Network.²⁰

Launched by China’s National Development and Reform Commission (NDRC) in 2022, it routes compute demand from China’s coastal economic centers to energy-rich western provinces through eight planned national computing hubs and ten data center clusters.²¹ This policy, perhaps more than any other, cuts against the central government’s tendency to encourage competition among provinces and municipalities. In this case, the application of central planning forces infrastructure specialization for western provinces, including Guizhou and Inner Mongolia, while providing more reliable infrastructure for the heavily populated coastal provinces and cities. In practice, eastern coastal cities generate the demand for AI training and inference while western provinces supply cheap hydroelectric power, low temperatures, and abundant land. In essence, the central state has orchestrated the interprovincial infrastructure needed for a start-up in Shanghai to connect with and train models on data centers located in Inner Mongolia.

No equivalent national compute-routing system exists in the United States. On the contrary, data center siting in the United States is primarily driven by private capital and local permitting rather than coordinated allocation, while cross-country broadband infrastructure investment is both overwhelmingly private and largely coupled to the rights-of-way provided by the Interstate Highway System. Similarly, transmission infrastructure is managed by a convoluted assortment of Regional Transmission Organizations and public utility commissions overseen by the Federal Energy Regulatory Commission. Coordinated investments that combine all of these domains—compute, connectivity, and energy transmission—is thus severely inhibited by inconsistent and overlapping authorities and jurisdictions. Absent these constraints, the United States would be well positioned to mirror China’s compute-routing strategy by locating multi-gigawatt data center projects near interregional grid and fiber hook-ups on federal lands in western states.²²

The third way that the national government in China shapes AI development is by setting standards, from algorithms to embodied applications. In March 2026, the MIIT established a dedicated Humanoid Robot and Embodied Intelligence Standardization Technical Committee and released the first national standard system covering the humanoid robot industry’s full stack.²³ Similarly, the MIIT’s AI+Manufacturing Initiative, jointly issued by eight other agencies in January 2026, provides the sector-specific implementation guidance. The document targets three to five general purpose industrial models deployed in manufacturing by 2027, along with one thousand industrial AI agents and five hundred demonstration scenarios across steel, petrochemicals, automotive, aerospace, and pharmaceuticals.²⁴ While the document sets out specific numbers, they are best thought of as directional indicators of the central government’s priorities. Though every province must pursue AI+Manufacturing and AI development more broadly, details are left up to local cadres. Setting these standards may be burdensome, causing smaller developers to deal with months of paperwork and delays. But in other cases, especially in hardware, regulations may push manufacturers toward more widely usable and interoperable designs.

Subnational Competition and Dynamics

That gap between central government direction and the competition and specialization between subnational units of government, such as provinces and municipalities, is where many of the most interesting dynamics in Chinese AI industrial policy emerge. The Chinese system relies on subnational competition to translate central priorities into local action, and this competition provides a rubric for the success of local cadres. Provincial and municipal officials advance by delivering visible achievements in strategically designated sectors within their tenures. This dynamic promotes a sort of convergence around successful policies that might be initially tested in one province before being copied by other competing party officials.

For example, computing power vouchers, first encouraged nationally by the NDRC in December 2023, are now issued by Beijing, Shanghai, Shenzhen, Hangzhou, Chengdu, Shandong, Henan, and others.²⁵ The mechanism is straightforward: local governments subsidize the cost of renting compute time at data centers for AI model training, typically covering 30 to 60 percent of costs. But the competitive dynamic around vouchers is not straightforward at all, as provinces can bid against each other on generosity. Standard vouchers tend to run from $140,000 to $200,000, but the city of Hangzhou has offered up to $1.1 million for individual vouchers. Each locality tries to make its compute environment more attractive than the next, pulling AI firms toward its data centers and innovation zones. Beyond vouchers, jurisdictions compete through municipal AI industry funds, Ministry of Science and Technology-designated innovation zones, specialized action plans, and start-up incubator programs that offer free office space and even loan guarantees to individual AI founders.

This dynamic has no American equivalent. The United States has no national compute voucher program for AI start-ups. It has no cross-sector deployment mandate comparable to AI+. It has no common institutional mechanism through which federal priorities automatically trigger competitive responses from state governments. While the United States does have block grant programs designed to nudge states toward federal priorities, the incentives they provide are often weak and poorly calibrated. The chips and Science Act’s Regional Innovation Engines program, which used a competition for federal grants to entice localities into investing in science and technology sectors, is perhaps the closest analogue. But as a creature of Congress, the program’s dollars were spread thin to appease multiple constituencies; it was overly focused on job creation per se and has had limited follow-through.

At the same time, the direction of China’s central government allows provinces to specialize their roles in contributing to China’s AI stack, rather than all competing in an involuted race to all offer the same range of incentives. Some provinces, particularly inland provinces, have self-selected into distinct roles as providers of land, power, and data center infrastructure within what amounts to a national architecture for AI development. The result is convergence in the basic policy instruments employed at a local level combined with salutary differentiation in each locality’s economic function as part of a coherent whole.

Regional Specialization: Innovation Density in the East

Beijing could best be characterized as the center of China’s governance and technical talent. For technical talent, the concentration is structural and longstanding: Tsinghua and Peking Universities, the Chinese Academy of Sciences, and several leading model developers, such as Baidu, Moonshot, and Zhipu AI, are located there. Additionally, Beijing operates a voucher system that offers 40 percent reimbursement for using PRC-made compute, compared to a 30 percent reimbursement for foreign compute.²⁶

Hangzhou, the capital of Zhejiang province, is dense with AI companies and adjacent firms. It is home to DeepSeek, Alibaba, and Unitree Robotics. Like other cities, Hangzhou uses a computing power voucher program that amounts to 10 billion RMB over four years, with subsidies for up to 60 percent of compute costs, among the most generous in China.²⁷ Like several other provinces and municipalities, Hangzhou is developing a $13 billion state fund for supporting the full AI stack.²⁸

Additionally, Hangzhou offers specific rewards to AI companies depending on the specific application of their product as well as the specific characteristics of the model. For instance, the city government rewards the decision to make a model open source with 1 million RMB, while products with applications in manufacturing and heavy industry receive five times as much.²⁹ Hangzhou’s technical talent for model development largely derives from Zhejiang University and institutional bridges to commercialization like Zhejiang Lab, a joint venture established in 2017 between the provincial government, Zhejiang University, and Alibaba. The three-way structure between the state, academia, and technology firms is a replicable form articulated in China’s New R&D Institute (NRDI) model.³⁰ These labs vary in size and capability, but they all retain some level of government involvement and have played a key role in AI development at the local level, with other NRDI labs in Shanghai and other coastal cities.³¹

Shanghai occupies a somewhat different position. It was the first locality in China to pass an AI-specific law, the Shanghai AI Development Regulations of 2022, and has used that first-mover position to set the regulatory templates that other cities adopt.³² In an effort to support existing firms and attract talent, the municipality offers vouchers for training and inference costs, including 80 percent of training fees.³³ The city also hosts the National Local Joint Humanoid Robot Innovation Center, benefitting from its concentration of manufacturing sectors relevant to AI integration, from automotives and integrated circuits to biopharmaceuticals—sectors where Shanghai already has a prominent role.³⁴ Additionally, the city contains a national-level AI research lab.³⁵ The regulatory inversion between software and hardware relative to the United States is visible here: Shanghai can build local implementation frameworks for AI deployment precisely because national-level questions about algorithmic governance are already settled.

Next, Guangdong province is home to Shenzhen, a city that has become synonymous with efficiency and economies of scale in areas like manufacturing and microelectronics. Shenzhen has over fifty-seven thousand robotics-related enterprises, thirty-four publicly listed robotics companies, and nine unicorns in the sector. Shenzhen is thus notable for its concentration of firms contributing to “embodied AI,” such as BYD for electric vehicles and DJI for drones.

The city has even promoted an embodied AI action plan.³⁶ Additionally, it hosts Peng Cheng Laboratories, a Ministry of Science and Technology-designated national lab and research institute for AI and other digital technologies supported by provincial and municipal funding.³⁷ Shenzhen also offers a compute voucher subsidy of 50 percent for established firms and up to 60 percent for start-ups,³⁸ and it recently launched an initiative to support start-ups developing agentic AI applications in particular.³⁹ Shenzhen, like many other cities in China, supports a local AI fund.⁴⁰ But numbers alone do not capture what makes Guangdong’s position distinctive. Its physical proximity to Hong Kong and preexisting industrial bases—dense manufacturing clusters, deep supplier networks, and spillover effects from the electric vehicle supply chain—give Guangdong unparalleled qualitative advantages in physical AI deployment.

The Guizhou Model: Data Centers as Development Strategy

Then there is Guizhou province, an outlier compared to model developer-focused cities and in some ways the most instructive case of specialization within China’s AI industrial policy. As a mountainous, historically poor province in China’s southwest, Guizhou has no meaningful AI research base, no frontier model companies, and no high-profile robotics development. What it does have, however, are ideal conditions for data center infrastructure, including cheap hydroelectric power, a temperate climate, stable geology, and clean air. Known as China’s “digital valley,” Guizhou’s total computing power is projected to reach 190 exa-FLOPS this year across more than fifty data centers, with participation from hyperscalers such as Baidu, Huawei, Tencent, and NetEase.⁴¹ While substantial in the Chinese context, this puts Guizhou’s compute hub in a middle position relative to U.S. hyperscalers.

For perspective, Guizhou’s AI compute capacity, aggregated across multiple sites, is roughly comparable to forty-seven thousand H100 GPUs for low-precision workloads, which is equivalent to about a quarter of the compute xAI controls at its one Colossus data center in Memphis, Tennessee. China’s relative dearth in AI compute reflects both the enduring effects of U.S. export controls on advanced AI chips and American frontier AI companies’ deep conviction about compute scaling. Chinese hyperscalers thus offset their constrained access to domestic compute through data center localization projects in neighboring parts of Southeast Asia, particularly Malaysia, Thailand, and Indonesia.⁴²

Nevertheless, just as Shenzhen and Hangzhou benefit from the economies of scale that come from concentrating technical talent and manufacturing in one locale, Guizhou benefits from economies of scale in data center construction and operations.⁴³ Gui’an New Area, a development zone in Guizhou, alone contains more than twenty-six major data centers, while demand for Gui’an’s data centers rose by over 517 percent year-on-year in the first quarter of 2025. To match this demand, the Gui’an Power Supply Bureau constructed a 220-kilovolt substation, four 110-kilovolt substations, and 130 dedicated transmission lines to serve the data center cluster, with a five-hundred-kilovolt substation under construction that will add four gigawatts.⁴⁴ Guizhou is thus an example of the success of specialization, in contrast to competing provinces and municipalities. The speed at which this infrastructure was built also illustrates the regulatory inversion between China and the United States at the physical level. Equivalent data center buildouts in the United States face permitting timelines, environmental review, and grid interconnection queues that can stretch to a decade, forcing the adoption of off-grid power generation.⁴⁵

Aside from the AI infrastructure buildout specialization in western provinces, the overarching tendency in China is for local officials to crowd around similar policies and priorities, leading to duplication and redundancy. In July 2025, Xi Jinping thus criticized local officials for a lack of differentiation and careless adoption of faddish priorities: “When it comes to projects, there are a few things—artificial intelligence, computing power and new energy vehicles. Do all provinces in the country have to develop industries in these directions?”⁴⁶ Coverage from the People’s Daily similarly described it as an exhortation to cadres to modernize and industrialize prudently, rather than incurring debt and boosting short-term policies.⁴⁷

There seems to be a growing recognition that AI policy convergence, left unchecked, reproduces the kind of duplication that characterized earlier rounds of Chinese industrial policy in sectors from solar panels to electric vehicles. Interestingly, however, the provinces that have found valuable specialized roles in China’s AI stack did not receive instructions from Beijing specifying their function. Rather, they arrived at differentiation because their existing natural resources, characteristics, and industrial compositions made certain specializations natural and others impossible.

Resetting the Terms of the AI Race

The AI competition between the United States and China is often described as a race, and in many ways it is. But as we’ve seen, China’s understanding of the finish line differs markedly from the prevailing conception in Silicon Valley. While America’s top AI companies continually one-up each other’s benchmarks on the path to AGI or superintelligence, Chinese companies are shaped by their constrained access to compute, the orchestration of the central government in Beijing, and the competitive dynamics of local governments to focus on AI’s productive diffusion.

What we’ve described is a narrow slice of a larger industrial and policy environment that has facilitated China’s successes in AI development and diffusion to date. Not one policy exists without connection to others. From improving education and selection of promising talent to enabling and accelerating power generation, the roots of competitiveness in AI stretch into the entirety of Chinese society, and there are hundreds of pages more to be written about the various policies in those other areas. But one conclusion is clear: it is not enough to support model developers or keep the state out of the way in software. Dysfunction in training technical talent or permitting necessary infrastructure is detrimental to a country’s overall competitiveness in AI. In this way, competitive AI development appears to be possible only for a state with a certain level of harmony and alignment.

The Chinese state’s support for AI adoption and diffusion has no parallel in the West. From the central coordination of backbone infrastructure to the bidding war between provincial and local governments for talent and investment, China’s AI industrial strategy borrows from the same playbook it has used to develop technological prowess in myriad other sectors and industries.

Imagine the same level of orchestration in America: coastal cities such as New York, Miami, San Francisco, and Boston offer vouchers for compute, while each city operates a local fund with billions of dollars to invest in promising firms. At the same time, the federal government incentivizes clustering data centers and power generation in land-abundant western states such as Montana and Wyoming. And above all this, Washington sets various targets for AI integration into private firms by specific deadlines. In the context of American history and political dynamics, such an arrangement may appear incoherent or tyrannical in its overreach, or at the very least, as counterproductive to our traditional understandings of American innovation. Yet it is this model that China is currently betting will allow them to surpass the United States in AI.

In the Western imagination, AGI represents a true threshold technology, if not humanity’s final invention. From our perspective, this is not so much wrong as incomplete. The advent of AGI will represent a true turning point in human civilization, but it should not be counted on as a deus ex machina. A fuller conception of the AI race would instead synthesize China’s practical obsession with diffusion, integration, and embodiment with our valid focus on raw capabilities. Just as a tree that falls in the woods with no one to hear it doesn’t make a sound, a superintelligence that sits idle in a data center does little to increase prosperity. And while we may soon have genius-level AI coders and scientists at our fingertips, bridging the accelerating progress in the virtual world to the world of tangible, physical production remains the hardest step of all. Here, China retains an advantage, one that could very well offset America’s leadership at the technological frontier in the fullness of time.

This article originally appeared in American Affairs Volume X, Number 2 (Summer 2026): 103–15.

---

Notes