This month the Chinese government rolled out its 15th Five-Year Plan, a central orienting document for Chinese policy through 2030. The plan, as well as the annual output from the National People’s Congress, signals both targets for technological advancement and some of the tools the government is likely to use to hit them. This is the first five-year plan since the popularization of the current paradigm of artificial intelligence and the first since the United States and its partners implemented export controls aimed at stemming China’s access to advanced semiconductors. At the same time, official Chinese ambitions to advance in high-tech sectors stretch back decades. We asked an array of specialists: In the broad area of technology, what stands out about the Chinese government’s approach in 2026?
This DigiChina Forum was edited by Johanna Costigan and Graham Webster.
Note: Additional responses have been added since this Forum went live. Check the bottom for the latest!
KENDRA SCHAEFER
Partner, Trivium China
The 15th Five-Year Plan indicates that U.S. export controls on advanced semiconductors are pushing China to craft policies aimed at compute utilization efficiency as the country attempts to squeeze more output from the hardware it already has. The plan pursues this goal on three fronts: national infrastructure planning, local government oversight, and enterprise cloud adoption.
At the national level, the FYP calls to continue building the National Unified Computing Power Network and related state-backed scheduling platforms. The purpose of this network is to pool compute resources nationally and route workloads to underutilized data centers, ensuring that existing hardware is fully utilized.
At the local level, the plan moves to prevent local officials from wasting capital on ill-considered data center projects. In recent years, local governments raced to build data centers to hit regional digitalization targets and national KPIs, often without regard for local grid capacity, renewable energy availability, AI readiness, or actual demand. The FYP's directive to "coordinate the layout and orderly construction of computing infrastructure" reasserts central planning authority over where capacity gets built.
At the enterprise level, the FYP calls to "actively develop public cloud services," signaling a deliberate effort to migrate organizations onto hyperscaler platforms such as Alibaba Cloud and Huawei Cloud, another efficiency measure. State-backed think tank CAICT data places China's cloud utilization rate at approximately 28%, against roughly 60% in the United States, a gap that reflects years of state-owned enterprises and government agencies defaulting to on-premises deployments. The result is scattered, fragmented data centers all competing for the same limited pool of AI-optimized semiconductors. These smaller installations cannot match the hyperscalers’ ability to efficiently allocate resources through greater expertise and advanced tooling.
Why the focus on wringing out waste? At this point, the strength of China's high-end semiconductor manufacturing capabilities is an engineering problem. There is little more the state can do to accelerate R&D timelines. Infrastructure optimization, by contrast, is precisely where direct state intervention can move the needle. And regardless of how the chip export controls ultimately play out, a leaner, better-coordinated compute infrastructure will serve China well over the long term.
JEREMY WALLACE
A. Doak Barnett Professor of China Studies, Johns Hopkins School of Advanced International Studies
From a climate-oriented perspective, the 15th Five Year Plan is a disappointing, conservative document. To be sure, Chinese plans on clean energy have tended to be “under promise, over deliver” affairs. And there is the new Ecological and Environmental Code, which some see as laying the groundwork for a future Climate Law. Still, this document refuses to plan for the transition. Instead, it extends the current situation—where clean electricity covers all new demand but little or nothing is done to displace coal—for another five years.
It’s unclear to me why this conservatism is present. Of course, electricity grids are famously focused on reliability. Blackouts are serious political problems to be avoided, and memories of inadequate generation leading to shortages in 2021 are only five years old. Yet China has added a once-unfathomable amount of solar generating capacity over those five years. An additional 55 GW came online in 2021, then a conspicuously lucky 88 GW in 2022, followed by 216 GW, 278 GW, and 315 GW in the past three years, respectively.
That’s 950 GW in just solar. The country’s numerous solar manufacturers can now pump out 1,000 GW of panels every year. China could race to become the first electrostate, but instead, if the Five-Year Plan is any indication, it remains frozen. Even absent democratic decision making, it’s politically hard to lay off miners and cut off the tax-paying companies of the coal sector. It’s even harder when the mood of the country, especially its urban middle and upper classes, is demoralized by the policy-induced bursting of the real estate bubble. That being said, solar has come so far and so fast in the past five years that I dare to dream about what could be, despite a disappointing five-year plan. Perhaps the 15th FYP will be eclipsed as dramatically as the 14th FYP was in this space. After all, the green tech revolution is a huge mess that just might save the world.
ANGELA HUYUE ZHANG
Professor of Law, Gould School of Law, University of Southern California
China’s new Five-Year Plan offers a revealing answer to a question that observers often get wrong: What role does law play in Chinese technological development?
The conventional view is that China advances innovation through industrial policy and state subsidies, while law is largely irrelevant. But the 15th Five-Year Plan makes clear that this view is misleading. In fact, regulation features prominently in Chapter 5, which addresses emerging and future industries, including electric vehicles, advanced manufacturing, biopharmaceuticals, embodied AI, quantum technology, and brain-computer interfaces.
More specifically, the Plan explicitly calls for regulatory approaches including “sandbox regulation” and “trigger-based regulation,” as well as for legislation in emerging sectors such as biopharmaceuticals and “intelligent driving.” This points to a distinctive Chinese model of legal governance, one I have explored in a recent paper: Law should not be seen merely as a brake on technological change, but rather as critical infrastructure for innovation.
Consider first “trigger-based regulation.” The idea is straightforward. Instead of attempting to design comprehensive rules for every technology in advance, regulators allow innovation to move forward and intervene quickly when concrete risks emerge. In effect, the state tolerates uncertainty at the front end, often delaying sweeping legislative or regulatory intervention until problems become highly visible, systematic, and politically salient. There are certainly downsides to such an approach, as sharp regulatory swings can also spook investors and entrepreneurs, as seen in the previous tech crackdown. Nonetheless, permissive and tolerant regulation was central to the early growth of China’s internet economy, and a similar stance has now been applied to generative AI.
Now consider the other concept detailed in the Plan: “sandbox regulation.” Here the state does the opposite. In some sectors, innovation cannot proceed unless government acts first. Autonomous driving is the clearest example. Existing legal frameworks were not designed for fully driverless vehicles on public roads. If the law is not updated quickly, the technology remains trapped in the lab. So regulators have created bounded zones for experimentation, authorized pilot programs, and gradually constructed rules for liability, insurance, and technical standards. In those settings, law is not merely reacting to innovation; it is clearing a path for it.
This helps explain why China has moved so quickly in autonomous driving. Robotaxis, for instance, have now expanded to more than 20 Chinese cities, while commercial deployment in the United States has so far been permitted in only six, even though companies such as Waymo began experimenting years earlier.
The point is not that China is simply “better” at regulating technology. It is that China’s hierarchical political system can align legislation, administrative approval, and local experimentation more quickly when it decides that a sector is strategic. Democratic systems, by design, are more decentralized and often move more slowly. That can be a virtue, offering stronger protections for individual liberty and public safety. But it can also impose real coordination costs in fast-moving sectors.
Law, in other words, is not peripheral to China’s innovation model. It is central to it. The 15th Five-Year Plan makes that plain.
PAUL TRIOLO
Honorary Senior Fellow, Center for China Analysis, Asia Society Policy Institute
China’s 15th Five-Year Plan (2026–2030) marks a sharper, more security-driven technology agenda than its predecessor. The 14th Five-Year-Plan was drafted before Washington’s sweeping 2022–2025 semiconductor-related export controls and investment restrictions—and before generative AI became the defining field of technological competition.
As a result, the 14th FYP treated semiconductors, AI, quantum, and digital infrastructure as major development priorities, but kept them within a familiar framework of modernization and industrial upgrading. The new Plan outline is different in tone and design: it is much more explicit about breaking bottlenecks, securing “strategic initiative,” and reorganizing state support around resilient control of critical technologies and rapid deployment at scale.
What stands out most is that AI is no longer just one priority sector among many; it is the organizing logic for a broad industrial transformation. The new outline calls for progress on the comprehensive “AI+” action plan—building on implementation guidance released last summer. AI+ links AI to science, industry, culture, public services, and governance, and pairs AI development with a three-part buildout of computing power, algorithms, and data. It calls for national hub computing clusters, explores building ultra-large intelligent computing clusters, supports government purchasing of computing services and compute leasing, promotes coordinated siting of green electricity and compute, and aims to lower compute costs for small and medium-sized firms. It also pushes “model-chip-cloud-application” (模芯云用) coordination—while promoting multimodal models, agents, embodied intelligence, and even “explore” the development of artificial general intelligence. Just as important, it calls for national AI data corpora, sector-specific high-quality datasets, and a framework for the reasonable use of AI training data.
On semiconductors, the new plan is less about a single headline performance or capacity target than situating the chip ecosystem into a broader sovereign digital stack. The outline explicitly calls for upgrading high-end processors, optoelectronic devices, basic software, and industrial software, while highlighting the need for “decisive breakthroughs in key core technologies,” signaling unprecedented emphasis on self-reliance. In other words, Beijing is no longer thinking only in terms of catching up in one manufacturing node; it is trying to thicken the entire ecosystem around compute infrastructure, software, data, and application deployment under external pressure. This approach was also reflected in an important paper released just before the NPC by semiconductor industry leaders.
The plan is also more ambitious on robotics and future industries. It names robots, particularly humanoid versions, as among critical strategic industries—and elevates embodied intelligence to the same tier as quantum technology, brain-computer interfaces, 6G, and nuclear fusion. It also introduces structured mechanisms for the state to support future industries: large-scale application-demonstration programs, national AI application pilot and testing bases, concept-verification centers, and explicit risk-sharing mechanisms for investment.
Finally, the financing modes around government support for advanced technology is evolving. Major Chinese banks last week began ramping up lending to technology firms, with some targeting around 30% growth in loans to innovation-oriented companies; loans to small and medium tech firms had already risen to 3.63 trillion yuan by end-2025. That matters because the 15th Plan is not just prioritizing AI, semiconductors, and robotics rhetorically. It is building the financial plumbing—through banks, state planning, compute procurement, and industrial pilots—to fund them at scale in a far more hostile external technology environment.
Expect to see a lot more Chinese AI and advanced semiconductor and robotics firms tap capital markets in 2026, following on the heels of major IPOs in Shanghai and Hong Kong at the end of 2025 and early this year, including GPU players Biren, Moore Threads, Illuvatar, and Enflame, and AI model developers Zhipu, Minimax, and, soon, Moonshot. China’s five year plans have long since given up their planning aura in favor of high-level guidance, and the 15th iteration reflects both Beijing’s industrial priorities and its flexibility on how precisely they are achieved.
MARTIN CHORZEMPA
Dennis Weatherstone Senior Fellow, Peterson Institute for International Economics
China’s latest five-year plan reveals how central AI has become for China’s leadership. The plan mentions AI 52 times and includes a standalone AI+ action plan. By contrast, the last five-year plan from 2021 mentioned the term six times. AI has gone from a sideshow listed among other critical technologies to a key part of government ambitions for its economy and society.
China’s AI strategy can seem like the opposite of America’s: U.S. labs are framed as racing to artificial general intelligence (AGI) with closed source models, while Chinese are fast followers focused on diffusion of open models. The U.S. aims for limited regulation, while China organizes a massive industrial policy effort to develop indigenous capacity and reduce dependence on the United States for key semiconductors and equipment.
Yet, China’s approach has striking similarities to the U.S. AI Action Plan. Both frame AI as crucial to strategic competition. Both aim to accelerate AI development and diffusion, and both are positive on open-source AI despite safety risks.
The U.S. aims for “unquestioned and unchallenged global technological dominance.” China’s plan deploys military metaphors like “storm heavily fortified positions,” alongside a narrative of “win[ing] the strategic initiative” in “fierce international competition.”
Point after point in China’s plan calls to “accelerate” initiatives to develop and diffuse AI, reminiscent of the second Trump Administration’s support for AI “accelerationists” over more cautious voices accused of being “doomers.”
China’s government work report said that “China-made large AI models spearheaded the development of the global open-source AI ecosystem” and vowed to “support the development of open-source AI communities.” The U.S. plan more explicitly notes the “geostrategic value” of ensuring the United States has competitive open models to rival China’s. Notably, China’s plan does not preclude labs, such as Alibaba, going in a more closed direction.
Where the two countries differ the most is the role of the state. The U.S. plan is more focused on eliminating perceived barriers to AI development and adoption through deregulation. China’s more interventionist plan calls for the state to take on a more active role to foster “self-reliant, controllable” inputs for AI that do not depend on the United States. The plan thus includes efforts to make more advanced chips for training and inference, forming nationwide networks of computing power, and a national market for data—a longstanding ambition that is now more important as it could serve as a training data source for technical AI applications.
BEN MURPHY
Translation Manager, Center for Security and Emerging Technology (CSET), Georgetown University
One of the major problems the 15th Five-Year Plan is attempting to address is the innovation-application mismatch in China. Compared with competitors such as the United States, a much greater proportion of scientific research breakthroughs in China occur in state-run universities and research institutes rather than in corporate R&D labs. The disconnect between the academically oriented, state-funded researchers who make discoveries and the Chinese corporations and startups with the market knowledge and profit incentive to exploit innovations is a perpetual concern for policymakers.
Hence the 15th Five-Year Plan’s emphasis on the “mainstay status” (主体地位) of companies in technological innovation. The 14th Five-Year Plan (2020-2025) featured a lone call for “strengthening the status of enterprises as the principal entities of innovation,” and the fact that the 15th Five-Year Plan repeats similar language and elevates the phrase in importance (it is the subject of Chapter 10 of the Plan) is a tacit acknowledgement by China’s leaders that this problem has not been resolved and is in fact becoming worse.
And so—despite the irony that five-year plans are one of the most obvious holdovers from China’s Stalinist planned economy era—the 15th Five-Year Plan advocates putting more power and resources into the hands of Chinese companies. It calls for “increasing enterprises’ level of participation in national major technological innovation decision-making.” It encourages major corporate technology projects to be included, in an unspecified way, in state technology plans. It grants “greater autonomy” to “leading enterprises” in choosing which technical routes to pursue, which entities they collaborate with, and how they allocate expenditures when working on state-funded technology projects. A notable example of this policy in the 15th Five-Year Plan is the emphasis given to developing China’s commercial space sector, subject of a recent national plan of its own. Commercial spaceflight was mentioned only once in passing in the 14th Five-Year Plan, but in the 15th, it is one of 10 “new industries” to be developed, along with perennial priorities such as integrated circuits and large aircraft.
Of course, like many policies before it, the 15th Five-Year Plan is long on aspirations and short on specifics. Making more scientific data from state-funded projects available to companies—another proposal in the Plan—sounds smart, but who decides which companies get the data, and how much? If state-funded universities and labs are to permit “micro-, small-, and medium-sized enterprises” to use their innovations per the Plan, how will issues like licensing fees and IP ownership be handled? It remains to be seen whether and how policymakers will translate aspiration to reality.
MEI DANOWSKI
Co-founder and Principal, Natto Thoughts
China’s 15th Five-Year Plan (15th FYP) outlines a strategy based on integrating cybersecurity into every layer of digital development—from core infrastructure to international governance. This is not surprising, given the rapid development of cybersecurity in China.
Accelerating the construction of a “cyber superpower” (网络强国) is one of the five superpower-building areas highlighted in Part II of the 15th FYP. (The other four areas mentioned are: manufacturing superpower, quality superpower, aerospace superpower, and transportation superpower.) Furthermore, Part XIV discusses “Safeguarding National Economic Security.” The section titled “Enhancing Cybersecurity Capabilities” provides the main details of China’s strategy for cybersecurity and building a cyber superpower.
Particular emphasis is placed on protecting critical information infrastructure through cybersecurity reviews and cloud service assessments, as well as improving technical resilience by building disaster recovery and backup systems. There are also safeguards for industrial control systems and emerging technologies and applications. This is the first time that technical resilience—especially building disaster recovery and backup systems—has been included in the FYP. This likely reflects China’s recognition of the urgency amid growing technological competition and geopolitical tensions with Western nations.
At the same time, the 15th FYP promotes the growth of the cybersecurity industry as a pillar of economic and technological development. It calls for increased innovation, the production and adoption of secure and reliable information products and services, and the integration of cybersecurity into the modernization of manufacturing through intelligent transformation, digital transition, and network integration (智改数转网联). “Secure and reliable” products and services are aimed at reducing dependence on foreign products and services.
Beyond technical and industrial priorities, the 15th FYP underscores the importance of shaping a “healthy” cyberspace. Initiatives such as the “Qinglang” (Clean and Bright, 晴朗) campaigns, which first began in 2016, continue to curb online misinformation, cyber violence, and other content regulators disapprove of. Crackdowns on cybercrime, enhanced personal information protection, and tighter oversight of platform algorithms and data practices reinforce state control over the digital environment.
Overall, the cybersecurity strategy in the 15th FYP, compared to the 14th FYP, introduces more granular requirements, such as product standards for the development and use of “safe and reliable” information products and services, as well as technical resilience through disaster recovery and backup systems. There are also new security assessments specifically for cloud computing services. How these requirements are further specified and implemented may be worth continued attention over the next five years.
JOHANNA COSTIGAN
Independent Writer and Editor
Sticking to a timeline is especially difficult when you stand to gain from both meeting and missing your target. That’s the predicament the leaders of major AI labs in the U.S. have put themselves in when it comes to “achieving” artificial general intelligence (AGI). In February, OpenAI CEO Sam Altman described AGI as “a near-term thing.” In January, Anthropic’s Dario Amodei predicted an AGI equivalent (he dislikes the popular term) could come in the next 12 to 24 months.
In response to remarks like these, markets tilt and Washington stays gridlocked. Plenty of people disagree with such predictions, either rejecting the premise that AGI will ever appear, or debating when it will. Still, in America, the AGI conversation takes place largely in reaction to statements from the labs. In China, government intervention in AI discourse has resulted in more emphasis on the AIs of now relative to possible AGIs of the future. Officials have given companies the cues for what to focus on, packaged in policies and slogans such as the “AI+ Initiative” and the long-pursued “technological self-reliance.” The details of what to do with AI follow: upgrade existing industries, expand enterprise applications, innovate toward the efficient use of China’s limited compute, and adhere to national AI regulations.
The signaling on AGI has been tellingly sparse—yet, arguably equally tellingly, existent. Rather than completely ignore the term, Beijing has acknowledged it as a worthwhile pursuit for researchers without staking the future on AGI’s safe, functional, and assured emergence. This restrained approach was on display in both the 15th Five-Year Plan (FYP) and the 2026 Government Work Report.
The work report didn’t mention AGI. It focused on implementing China’s AI+ strategy and ensuring Chinese models lead the global open-source ecosystem, among other objectives. The FYP, an articulation of China’s big-picture policy strategy from now until 2030, featured AGI modestly, encouraging the “exploration of development pathways toward general-purpose AI.” That’s as close to a mention of AGI as the FYP gets. The term used, 通用人工智能, translates to “general-purpose” AI, applying the generality to the use case of the AI tool. That’s in contrast to the AGI definitions used in the U.S., which typically emphasize the human-like intelligence of the AI system. Including AGI in the FYP without pinning all hopes to it is prudent: rejecting the possibility of AGI is almost as extreme as insisting we’re on the brink of it.
Timelines are key to managing the AGI narrative tightrope, rendering Beijing’s time-locked policies useful. One thing is clear: Developing AI systems that are “general purpose” is of interest to policymakers. But in invoking “pathways,” they suggest this is a longer-term goal—one that will take a few years, during which the economic utility of other forms of AI should be thoroughly explored. It makes it to the five-year plan, but isn’t pressing enough to be included in the one-year plan.
The result is that China’s AI stakeholders, including the major model developers, investment community, and media ecosystem, know the boundaries. AGI can be discussed (DeepSeek CEO Liang Wenfeng, for example, has publicly stated his aspiration to create it). But government input clarifies that chasing AGI should not become China’s singular approach to AI development. That stands in contrast to the Silicon Valley-driven discourse, which frames the market and geopolitical competitions as fundamentally about AGI. China’s technocratic state is made of officials who are equal parts practical and radical: they forgo Silicon Valley utopia-speak, yet inject vast resources to science and technology “bets” that make every other government look cautious and passive. State planners are willing to note the potential of AGI without losing sight of the AI implementation and oversight that has to happen in the meantime.
CHARLES SUN
Yale Sinovation Fellow, Yale School of Management
Signals only matter if someone acts on them. What stands out about China’s approach to AI in 2026 is not the ambition of its agenda, but how quickly private firms have been moving to capture the economic advantages that agenda creates.
The plan calls for “government procurement of compute services” (政府购买算力服务), “cultivating a self-reliant and controllable software and hardware ecosystem” (自主可控的软硬件生态), and “greater government procurement of indigenously innovated products” (加大政府采购自主创新产品力度). Read together, these aspirations authorize a procurement and subsidy architecture that makes alignment with state objectives the cheapest market option. The Ministry of Finance has proposed a 20% discount for domestic products in government procurement evaluations. Energy subsidies have reportedly reduced power costs by up to half for some data centers deploying Chinese-made AI processors. Provincial and municipal governments have layered on additional incentives that consistently favor domestic hardware.
Firms are responding accordingly. Alibaba Cloud, a central player in Beijing’s push to migrate workloads onto domestic hyperscalers, has shipped hundreds of thousands of its in-house AI chips and built domestic-silicon clusters serving over 400 clients—because the subsidy structure makes its vertically integrated stack more cost-competitive. Zhipu AI optimized its GLM models for Huawei Ascend and Cambricon chips before the plan was even finalized, because compatibility with domestic hardware has become a de facto requirement for accessing hefty public-sector contracts. DeepSeek's open-weight releases, most notably R1 under an MIT license, reflect a commercial bet that aligns with the plan’s support for open-source ecosystems: when compute is scarce, giving away the model layer and competing on efficiency is a rational strategy that happens to serve state goals.
The plan’s principle of “combining an effective market with a capable government” (有效市场和有为政府相结合) has appeared in five-year plans for years. In AI, it is becoming empirically visible in new ways. The state is not directing firms to build specific models or adopt particular chips. It is shaping the conditions under which commercial actors align with national priorities before they are formally required to do so. The incentives are local and tangible, and the cost of misreading direction is high. The more interesting question is not whether firms will align, but whether the AI bets embedded in China’s subsidy structure are technically sound, and what happens when they are not.
