AGI Has Quietly Become Central to Beijing’s AI Strategy

Executive Summary:

• Pursuit of artificial general intelligence (AGI) is a top-priority project within the Party’s increasingly centralized technology planning apparatus. Its success would both close the gap with U.S. firms and bind AGI models to Party-state governance, shaping how intelligent systems are aligned, deployed, and exported.
• Xi Jinping’s 2018 Politburo session operationalized the New Generation AI Development Plan, defining frontier breakthroughs as a lever of national power. Starting in 2020, Beijing and other provinces had institutionalized AGI in local initiatives, and by August 2025 the State Council’s AI+ Action plan codified AGI-linked targets into national modernization benchmarks.
• The Party-state’s approach rests on two inseparable pillars: frontier breakthroughs to secure sovereign control of general intelligence, and diffusion across the real economy to sustain political legitimacy and commercial value. The two are treated as mutually reinforcing, not competing.
• AGI is now written into China’s operating system for modernization. The MIIT meeting in June and the AI+ Action plan in August 2025 tied frontier models to industrial upgrading, governance standards, and long-term milestones to 2027, 2030, and 2035.
• Under the “new national system,” state institutions, elite labs, and firms are mobilized in concert, while outward-facing efforts such as Alibaba’s global intelligent network strategy show that Beijing views AGI not only as a domestic modernization tool but also as a lever of international power.

The State Council’s August release of the Opinions on Deeply Implementing the ‘AI+’ Action reignited debate over whether Beijing is serious about artificial general intelligence (AGI) or focused only on embedding applied AI across the economy (State Council, August 26). Commentary has often leaned toward the latter, but Party sources and policy documents show a clear throughline: since the 2017 New Generation AI Development Plan first defined “generalizable” intelligence as a research horizon, AGI has steadily moved from implicit objective to explicit policy goal. By 2025, that goal was tied to modernization benchmarks and reinforced through Party-state planning at every level.

Beijing has built its AI strategy on two inseparable pillars: frontier breakthroughs in general models and broad deployment across the real economy. Central directives since 2018 have emphasized sovereignty over core technologies and original innovation, while local governments piloted AGI (通用人工智能) initiatives beginning in 2020. By April 2025, Shanghai had classified distinct AGI subfields, and General Secretary Xi Jinping urged the Politburo to “seize the decisive opportunity and win the advantage” (占领先机、赢得优势) in AI. Under this logic, diffusion and frontier research are mutually reinforcing mandates: deployment underpins political legitimacy and commercial value, while frontier breakthroughs secure “self-reliant” (自力更生) capabilities and open the path to AGI as a sovereign capability

This dual-track approach undercuts the notion that the People’s Republic of China’s (PRC) AI policy is “pragmatic” or application-first. To assume the Party is indifferent to AGI is to miss the architecture of the “new national system” (新型举国体制). Within this system, the state builds compute centers, funds competitions and research programs, and sets standards for model evaluation, safety, and ethics, while firms like Alibaba and ByteDance are expected to operationalize breakthroughs at scale (China Brief, February 28, June 30). The launch of DeepSeek R1 in early 2025, and pronouncements such as Alibaba CEO Wu Yongming’s (吴泳铭) claim that “AGI is a certainty … the ultimate goal is ASI [artificial superintelligence]” (AGI已是确定性事件 … 终极目标是 … ASI), highlight that the Party-state now views general intelligence as both attainable and necessary (Equal Ocean, September 24). Tracking policy documents alongside industry behavior shows that Beijing’s strategy is not about choosing between diffusion and AGI but about fusing them into a comprehensive technological capacity. For competitors, this makes the AGI race a contest between political systems—one in which the PRC has tied frontier innovation and applied deployment into a single, state-directed project of geoeconomic power.

AGI Defines the Party’s Comprehensive AI Strategy

From the outset, Beijing’s leadership has treated AI as both an infrastructure for industrial upgrading and a frontier national project, setting the stage for the PRC’s evolving pursuit of AGI as a critical objective of state technology planning.

In the Party’s own lexicon, “frontier” has always encompassed AGI models: the State Council’s 2017 New Generation AI Development Plan defines new-generation AI to include “data-driven general artificial intelligence mathematical models and theory” (数据驱动的通用人工智能数学模型与理论) and advances in “artificial intelligence with high explainability and strong generalization capabilities” (实现具备高可解释性、强泛化能力的人工智能), establishing generalizable intelligence as a core research target and policy horizon (State Council, July 20, 2017). In October 2018, Xi Jinping convened the Politburo for a collective study session on AI. This was one of the first times the top leadership explicitly framed AI as a strategic lever for national power. He urged strengthening basic theory and called on scientists to “bravely venture into the uncharted zones of AI’s frontiers” (勇闯人工智能科技前沿的“无人区”), aiming for transformative, disruptive breakthroughs in theory, methods, tools, and systems (在 … 理论、方法、工具、系统 … 取得变革性、颠覆性突破) (Xinhua, October 31, 2018).

Xi’s 2018 directives operationalized the New Generation AI Development Plan’s AGI objectives while linking them to economic development and technology self-reliance, describing AI as “a key strategic lever for us to seize the initiative in global scientific and technological competition and a vital strategic resource for promoting leapfrog development” (是我们赢得全球科技竞争主动权的重要战略抓手，是推动我国科技跨越发展、产业优化升级、生产力整体跃升的重要战略资源). His speech framed AI as a strategic frontier with open-ended, system-shaping implications: Xi emphasized multidisciplinary integration, theoretical leadership, and sovereignty over key technologies—a blueprint that presupposed AGI ambitions even if the term itself was not used.

Xi’s next major intervention came nearly seven years later, in April 2025, just months after the release of DeepSeek R1, the first Chinese model widely perceived as a credible frontier breakthrough (China Brief, February 11). Xi told the Politburo that AI was already “deeply changing human production and lifestyles” (深刻改变人类生产生活方式) and stressed the need to “seize the decisive opportunity and win the advantage” (占领先机、赢得优势) in this domain (Xinhua, April 26). In keeping with the Party’s lexicon of new generation AI (新一代人工智能) – first codified in 2017 to include generalizable intelligence – Xi placed equal weight on frontier breakthroughs and broad deployment. [1] He called for advances in “basic theory, methods, and tools” (基础理论、方法、工具) and the construction of a “autonomous and controllable, collaboratively operating AI basic software and hardware systems” (自主可控、协同运行的人工智能基础软硬件系统).

For the PRC’s pursuit of AGI, this balance is telling: frontier research and diffusion are not competing priorities but mutually reinforcing pillars of strategy. Diffusion secures immediate returns in legitimacy and growth, while frontier breakthroughs open the possibility of achieving AGI as a sovereign capability. Xi made clear that the strategic stakes remain control across the entire technology spectrum: theory, models, chips, software, data, infrastructure, and governance. Rather than narrowing to applied AI, the Party’s goal is comprehensive, “gaining the upper hand and securing an advantage” (占领先机、赢得优势) across all domains. In the wake of DeepSeek R1, this amounted to a tacit endorsement of AGI as an attainable horizon. The April 2025 Politburo study session thus marked the culmination of a seven-year evolution: from the New Generation AI Development Plan’s broad objective-setting to Xi’s 2018 call for frontier breakthroughs, to a post-DeepSeek posture in which AGI is treated as a realistic extension of the Party’s drive to prevail in global technological competition.

AGI Moves From Horizon to Policy Goal

Beginning in 2017–2018, PRC AI policy emphasized national sovereignty over core technologies, frontier research, and integration with the economy, but it stopped short of invoking “通用人工智能” (AGI) by name. Soon, however, references to AGI began to appear in formal policy and local initiatives, signaling a shift from implicit horizon to explicit agenda. With central benchmarks in place, local governments served as testbeds, using competitions, subsidies, and research platforms to translate the Party’s frontier objectives into outcomes. This interplay between central direction and local experimentation laid the groundwork for AGI’s explicit elevation to national strategy.

Beijing was the clear frontrunner in moving AGI from horizon to policy goal. In 2020, under the guidance and support of the Beijing Municipal Party Committee and government, the Beijing Institute for General Artificial Intelligence (BIGAI; 北京通用人工智能研究院) was established, jointly backed by the Ministry of Science and Technology and the Ministry of Education and partnered with Peking and Tsinghua Universities. Founded and directed by UCLA-returnee and leading computer vision expert Song-Chun Zhu (朱松纯), BIGAI was tasked with advancing strategic, forward-looking, and foundational research on building “general intelligent agents” with human core cognitive abilities (Peking University Institute for Artificial Intelligence, March 15, 2023). This institutional foundation was followed in 2023 by the release of municipal measures tying AGI to Beijing’s 2023–2025 plan for building a globally influential AI hub. These measures called for pooling and managing compute across multiple clouds, creating open pre-training datasets, funding large-model algorithm research, and “exploring new paths toward AGI” (探索通用人工智能新路径) through embodied and brain-inspired systems. They also emphasized building AGI evaluation platforms, developing operating systems and compilers for large models, and applying general intelligence to domains from governance to medicine, finance, and autonomous driving (Beijing Government, May 23, 2023). [2] This represented the first systematic subnational policy document to treat AGI as an achievable policy objective.

Other provinces followed suit. In December 2023, Inner Mongolia hosted the first “Tongzhi Cup” General Artificial Intelligence Innovation Application Competition in Ulanqab, which drew over 120 enterprises and was paired with the launch of a General AI–Industry Integration Innovation Center (Xinhua, December 23, 2023). Inner Mongolia is one of eight national “hub nodes” that are home to data center clusters, and Ulanqab’s Grassland Big Data Valley has attracted investment from Apple and Huawei, among others (Substack/Sinocities, November 20, 2024; China Brief, September 25). Just days later, the Chinese Computer Federation and Anhui government co-hosted the 2023 National General Artificial Intelligence Innovation Application Competition in Wuhu. This event was tied to Anhui’s “Three-Year Action Plan for General AI Innovation and Development” (2023–2025) and involved more than 300 project teams nationwide. Winning projects were eligible for financing, promotional support, and, in some cases, multimillion dollar provincial subsidies for projects that landed locally (Anhui Science and Technology Department, January 16, 2024).

By 2024–2025, momentum around AGI had moved beyond local pilots to higher-level institutions and national directives. The Chinese Academy of Sciences’ Institute of Software established an AGI High-Performance Software Innovation Center in 2024 to study large-model mechanisms, develop new paradigms for parallel software, and explore applications in scientific computing and related industries (ScienceHR Net, June 21, 2024). In April 2025, Shanghai followed suit with its Notice on New Generation AGI Innovation ‘Challenge and Leadership’ Work, which singled out embodied, scientific, spatial, group, and brain-inspired intelligence as subfields of a comprehensive AGI agenda (Shanghai Economy and Informatization Committee, April 21). The timing overlapped with the Politburo collective study session on AI, where Xi urged his comrades to “seize the decisive opportunity and win the advantage” (Xinhua, April 26). Together, CAS’s new AGI unit, Shanghai’s explicit program, and Xi’s renewed national guidance signaled that the Party now regarded breakthroughs in AGI as attainable.

AGI Anchors the Party’s Strategy for New-Type Industrialization

By mid-2025, AGI had crossed a threshold: what had begun as a series of local pilots was now embedded at the core of national industrial strategy. In June, the Ministry of Industry and Information Technology (MIIT) convened a meeting to study Xi’s latest instructions and chart a course for embedding general models into the modernization of manufacturing. The readout directed units in the industrial system to “coordinate the layout of general large models and industry-specific large models” (统筹布局通用大模型和行业专用大模型), backed by expanded computing infrastructure, software–hardware integration, and industry-specific datasets. Notably, the meeting extended the Party’s frontier-plus-application logic into industrial modernization: breakthroughs and deployment were to advance together, creating “powerful dynamism” (强大动能) for “new-type industrialization” (新型工业化) (MIIT, June 4). Officials ordered large models deployed in key industries for “full-process intelligent upgrading” (制造业全流程智能化升级), cultivation of “advantageous AI enterprises” (人工智能优势企业), and development of “high-level AI open-source communities” (高水平人工智能开源社区), while also mandating risk controls such as “deep synthesis detection technologies” (深度合成检测技术) and new ethical rules. Taken together, the meeting signaled that AGI had been written into the operating system of national industrial strategy, binding frontier models directly to the Party’s agenda of comprehensive economic modernization.

The State Council’s Opinions on Deeply Implementing the ‘AI+’ Action, issued in August 2025, again elevated AGI to the level of long-term national planning. The document framed AI as a force to “reshape human production and life patterns, drive a revolutionary leap in productive forces, and bring deep changes to relations of production” (重塑人类生产生活范式、促进生产力革命性跃迁和生产关系深层次变革) (State Council, August 21). Its emphasis on new-generation intelligent terminals and intelligent agents (新一代智能终端、智能体) signaled that Beijing viewed frontier model capabilities as central to economic and social transformation. Directing use of agents and foundation models as the bedrock of an “intelligent economy” (智能经济) and “intelligent society” (智能社会), the State Council effectively designated applied AGI as a national modernization benchmark.

The Opinions spelled out how AGI-adjacent capabilities were to be institutionalized across science, industry, consumption, governance, and global cooperation (China Brief, September 21). On the research front, it called for “accelerating the construction and application of scientific foundation models” (加速科学大模型建设应用) and for exploring “AI-driven new paradigms of scientific research” (人工智能驱动的新型科研范式). In industry, it urged firms to build “AI-native enterprises” (智能原生企业) with operating logics rooted in general intelligence, while promoting “model-as-a-service, agent-as-a-service” (模型即服务、智能体即服务) as the backbone of future applications. And in governance, it envisioned “a new picture of human–machine symbiosis in social governance” (社会治理人机共生新图景), extending intelligent agents into public administration and security. By linking frontier model development to every sector of the economy and tying it to 2027, 2030, and 2035 milestones, the Opinions confirmed that the Party leadership now treat AGI as a goal whose realization is integral to both state capacity and the Party’s claim on future global competitiveness.

That agenda is now being deepened through state–industry partnerships, as firms align their strategies with national directives and investors chase objectives defined by the state. Alibaba plans to make its Tongyi Qianwen model an open “Android for the AI era,” dovetailing with Beijing’s push for national data processing buildout and model services (Equal Ocean, September 24). Similarly, investment firm Shanda’s (盛大网络) recruitment of Tsinghua associate professor Dai Jifeng (代季峰) to build an AGI venture aiming to compete with DeepSeek shows how private capital is already moving to match Party priorities (Leiphone, August 4).

Conclusion

Beijing’s AGI strategy is aimed at building a complete ecosystem in which frontier research and applied AI advance together, supported by state-backed infrastructure, standards, and capital. Its pursuit of AGI is therefore not a quixotic moonshot, but part of an increasingly centralized political economy: breakthroughs in frontier areas are inseparable from state planning, infrastructure build-out, and policy direction. Under this “new national system,” policy is already embedding general models into industrial planning (China Brief, June 16, September 26).

In the short run, this will mean more prescriptive targets on compute, model capability, and agent deployment, paired with governance that keeps frontier work “safe, reliable, and controllable”; in other words, shaped by Party priorities for internal security and political-military integration. In the medium-term, Beijing is likely to promote new national programs that treat large models as cyberneticist operating systems for the real economy and public services. [3] If successful, the PRC will not only narrow the gap with U.S. firms but also bind frontier AI development to Party-state governance, shaping how intelligent systems are aligned, deployed, and exported.

Notes

[1] The document calls for policymakers to “promote the deep integration of AI technological innovation with industrial innovation, establish a collaborative innovation system led by enterprises that brings together industry, academia, research, and application, support the transformation and upgrading of traditional industries, and open new avenues for strategic emerging industries and future industries” (要推动人工智能科技创新与产业创新深度融合，构建企业主导的产学研用协同创新体系，助力传统产业改造升级，开辟战略性新兴产业和未来产业发展新赛道).

[2] These could be akin to Chile’s “Project Cybersyn” (see, for instance, Medina, Eden. Cybernetic Revolutionaries: Technology and Politics in Allende’s Chile. MIT Press, 2011).

[3] The full title of the measures is “Several Measures for Promoting the Innovation and Development of General Artificial Intelligence in Beijing” (北京市促进通用人工智能创新发展的若干措施).