China’s Accelerated Rise in AI Power and Global Competition

China's pursuit of artificial intelligence leadership represents one of the most consequential technology campaigns of the twenty-first century. Anchored by the State Council's 2017 New Generation Artificial Intelligence Development Plan and continually reinforced through subsequent policy actions, the strategy has produced measurable results: China's core AI industry surpassed 1.2 trillion yuan in 2025, exceeding the plan's 2030 target five years ahead of schedule. Yet the campaign unfolds against a backdrop of intensifying US export controls, persistent hardware bottlenecks, and a rapidly evolving regulatory environment that seeks to balance innovation with political control. What follows is a comprehensive account of where China's AI strategy stands, how it arrived here, and what its trajectory means for the global order.

The foundational document driving China's AI ambitions is the New Generation Artificial Intelligence Development Plan, formally issued by the State Council on July 8, 2017, and publicly released on July 20 of that year. Drafted under the coordination of the Ministry of Science and Technology, the National Development and Reform Commission, and the Chinese Academy of Engineering, the plan laid out a three-step roadmap with escalating targets. By 2020, the plan called for China's AI technology and applications to reach globally advanced levels, with a core AI industry exceeding 150 billion yuan and related industries exceeding one trillion yuan. By 2025, it envisioned major breakthroughs in foundational AI theory, a core industry exceeding 400 billion yuan, and related industries surpassing five trillion yuan. The final milestone, set for 2030, targeted world-leading levels across AI theory, technology, and applications, with a core AI industry of one trillion yuan and related industries reaching ten trillion yuan. At 2017 exchange rates, the headline core industry figure of one trillion yuan translated to roughly 150 billion US dollars, a figure widely cited in Western analyses and broadly accurate as an approximation.

A common claim holds that the plan established 2049 as the target date for China to become the world's premier AI innovation center. This is imprecise. The plan's furthest explicit deadline is 2030, at which point China aims to be a "world primary AI innovation center." The 2049 reference derives from the plan's invocation of the Chinese Communist Party's broader "Two Centennial Goals" and the "great rejuvenation of the Chinese nation," which implicitly links to the centenary of the People's Republic. Some analysts and commentators have extrapolated this into an explicit 2049 AI target, but the original text does not contain such a milestone. The distinction matters because the plan's actual timelines have proved more aggressive than its drafters anticipated, with several targets already met or surpassed.

The plan's targeted sectors span a wide range: intelligent manufacturing, healthcare, smart cities, agriculture, national defense, transportation, finance, education, elderly care, environmental protection, and judicial services. The frequently cited shorthand of manufacturing, healthcare, finance, transportation, and national defense is accurate but incomplete. Notably, the original document was candid about China's weaknesses at the time of drafting, acknowledging "significant gaps in basic theory, core algorithms, key equipment, high-end chips, major products and systems, basic materials, components, and software." This frank self-assessment helps explain the intensity of subsequent government investment.

The policy landscape has evolved substantially since 2017. The Fourteenth Five-Year Plan, published in March 2021, elevated AI to the top of China's frontier technology priorities, ranking "new-generation AI" first among strategic technology fields alongside quantum information, integrated circuits, and brain science. It called for breakthroughs in foundational algorithms, dedicated AI chips, and deep learning frameworks while setting a seven percent annual growth target for overall research and development expenditure. The plan specified ten application domains spanning intelligent transportation, smart energy, intelligent manufacturing, smart agriculture, smart education, and smart healthcare, among others. In March 2024, Premier Li Qiang introduced the "AI Plus" initiative in the Government Work Report, modeling it on the earlier "Internet Plus" campaign and signaling a new phase of deep integration between artificial intelligence and the broader economy. By August 2025, the State Council had issued a comprehensive set of opinions on implementing the AI Plus initiative, establishing three new target phases: by 2027, smart terminal and AI agent adoption rates exceeding seventy percent; by 2030, adoption exceeding ninety percent with the intelligent economy becoming a major growth engine; and by 2035, China fully entering an intelligent economy and intelligent society phase. This August 2025 document, formally designated State Council Document 2025 Number 11, effectively superseded the 2017 plan's targets with updated and more ambitious goals. The Fifteenth Five-Year Plan, formally adopted at the National People's Congress in March 2026, gave AI unprecedented prominence, mentioning the term eight times in its planning suggestions compared to just twice in the Fourteenth Five-Year Plan. It called for fully implementing the AI Plus initiative, building a national integrated computing power network, and using AI to lead the transformation of scientific research paradigms.

Estimating Chinese government spending on AI remains notoriously difficult due to the opacity of budgeting, the blurred boundary between state and enterprise expenditure, and the role of local government guidance funds. Georgetown University's Center for Security and Emerging Technology assessed with moderate confidence in 2020 that China's public AI research and development investment was on the order of a few billion dollars in 2018, far below widely circulated but poorly sourced claims of tens of billions. By 2025, the picture had changed dramatically. Bank of America projected total Chinese AI capital expenditure of 600 to 700 billion yuan in 2025, with government investment accounting for up to 400 billion yuan and internet companies contributing roughly 172 billion yuan. The Federal Reserve Board's October 2025 analysis found that over the preceding decade, China had invested 912 billion dollars through local government venture capital funds in key sectors, with approximately 184 billion dollars flowing to nearly ten thousand AI-related firms. In March 2025, China announced an additional 138 billion dollars over twenty years in venture capital guidance funds targeting AI and quantum technology, and a separate National AI Industry Investment Fund of 600 billion yuan was launched. China's 2025 central government science and technology budget reached 398 billion yuan, roughly 55 billion dollars, representing a ten percent increase from the prior year, with AI among the priority allocations. For context, total unclassified US federal AI research and development spending in fiscal year 2025 stood at roughly 3.3 billion dollars, though US private AI investment of 109.1 billion dollars in 2024 far exceeded China's 9.3 billion dollars in private funding, illustrating the fundamentally different structures of the two countries' AI investment ecosystems.

The claim that AI could add up to seven trillion dollars to China's gross domestic product by 2030 originates from PwC's "Sizing the Prize" report, released in June 2017 at the World Economic Forum's Annual Meeting of the New Champions in Dalian. PwC projected that AI could contribute 15.7 trillion dollars to global GDP by 2030, with China capturing a 26.1 percent boost equivalent to approximately seven trillion dollars, and North America receiving a 14.5 percent boost worth 3.7 trillion dollars. A separate analysis by Accenture and Frontier Economics, also from 2017, estimated that AI could increase China's annual growth rate by 1.6 percentage points to 7.9 percent by 2035, adding more than seven trillion dollars in gross value added. McKinsey's 2018 analysis was more conservative, projecting thirteen trillion dollars in global AI output by 2030 without specifying a China figure of that magnitude. These estimates were generated before the pandemic, before US export controls, and before the generative AI explosion, and no major updated comprehensive estimate specifically revising the seven trillion dollar figure for China has been published as of early 2026. The figures remain directionally plausible but should be understood as maximum-potential projections made under assumptions that have since been partially invalidated.

China's data environment is frequently cited as a strategic advantage rooted in the country's 1.4 billion population, and the claim contains a kernel of truth but requires significant qualification. China has over 1.1 billion mobile application users, massive e-commerce platforms, and deeply integrated super-applications like WeChat and Alipay that generate enormous volumes of behavioral data. The relatively permissive domestic data collection environment, compared to many Western democracies, further facilitates large-scale data aggregation. However, the relationship between data volume and AI capability is more nuanced than raw population figures suggest. China's data ecosystem suffers from fragmentation across proprietary platforms, limited public-sector data accessibility, and cross-border data flow restrictions that reduce the utility of domestically collected data for global applications. Healthcare data, often cited as a domain of particular promise, illustrates the challenge: only 16.6 percent of cohort studies achieve ten or more years of follow-up, and just 10.6 percent use standardized data-sharing platforms. The real advantage lies less in population size per se than in the scale of digital platform deployment, state capacity for resource mobilization, and a culture of rapid adoption. A 2024 survey by SAS and Coleman Parkes found that 83 percent of Chinese professionals reported using generative AI, compared to 65 percent in the United States and 54 percent globally.

The execution of China's AI strategy depends heavily on a cohort of technology giants that have become de facto instruments of national policy. Baidu has positioned itself as a comprehensive AI platform company, with its Ernie Bot amassing over 200 million users by early 2025 and its Apollo autonomous driving program logging more than fifty million kilometers of public road testing across thirty cities. Alibaba has emerged as perhaps the most consequential player in the global open-source AI ecosystem: its Qwen model family became the most popular open-weight model series globally in 2025, with over one hundred models released, more than forty million downloads, and the Qwen3 family trained on thirty-six trillion tokens across 119 languages. Alibaba committed 380 billion yuan to cloud computing and AI infrastructure over three years and led private sector research and development spending at 67 billion yuan in 2025. ByteDance's Doubao chatbot became China's most popular AI assistant, exceeding 100 million downloads by September 2024, while the company triggered an industry-wide price war by offering its model at 0.8 yuan per million input tokens, described as 99.3 percent below prevailing industry prices. Tencent integrated its Hunyuan model across over two hundred services including WeChat and QQ, while investing approximately fifteen billion dollars in AI development between 2023 and 2026. Huawei, though not formally designated as part of the "national AI team," plays a pivotal infrastructure role through its Ascend chip series and MindSpore framework, and announced plans to fully open-source its AI software stack by the end of 2025.

Among smaller but strategically significant companies, SenseTime achieved record revenue exceeding five billion yuan in 2025 with thirty-three percent year-over-year growth, while its SenseNova model series claimed benchmark-leading performance with a fraction of the training compute used by Western competitors. iFlytek deepened its partnership with Huawei, co-developing its Spark reasoning model trained entirely on domestic Ascend computing power. Zhipu AI, a Tsinghua University spinoff, released its GLM-5 model in February 2026 trained entirely on Huawei Ascend chips and completed an initial public offering on the Hong Kong Stock Exchange. The sheer number of active players is striking: by 2025, China counted over six thousand AI enterprises, and over 302 generative AI services had completed filing with regulators.

No development better illustrates the shifting dynamics of China's AI sector than the emergence of DeepSeek. Founded in July 2023 as a research spinoff of the quantitative hedge fund High-Flyer, DeepSeek is led by Liang Wenfeng, a Zhejiang University alumnus who funded the venture entirely from High-Flyer's resources without external venture capital. The company attracted global attention with the release of DeepSeek-R1 on January 20, 2025, a reasoning model with 671 billion parameters that matched or exceeded OpenAI's o1 model on key benchmarks, scoring 91.6 percent on MATH-500 compared to o1's 85.5 percent, while reportedly costing only 5.6 million dollars to train in its final run. The predecessor model, DeepSeek-V3, was trained for approximately six million dollars on 2,048 Nvidia H800 GPUs, roughly one-eighteenth the training cost and one-tenth the team size of comparable Western models. The market reaction was seismic: on January 27, 2025, Nvidia lost 589 billion dollars in market capitalization in a single trading session, the largest single-day loss for any company in US stock market history, as investors questioned whether massive compute expenditures were truly necessary for frontier AI performance. DeepSeek's daily active users surpassed thirty million within weeks, and it briefly became the top-rated application on the US iOS App Store. Chinese academic analyses characterized the episode as evidence that US chip restrictions had "accidentally triggered reverse innovation," while Western commentators invoked the phrase "AI Sputnik moment." DeepSeek continued iterating rapidly, releasing V3.1 in August 2025 with 840 billion parameters and V3.2 in November 2025 with sparse attention mechanisms for improved efficiency. By early 2026, nine of the top ten open-weight AI models globally originated from Chinese developers.

The hardware dimension of China's AI ambitions is simultaneously the area of greatest vulnerability and most determined investment. Huawei's Ascend 910C, the most advanced domestically produced AI chip, employs a dual-die chiplet design on SMIC's second-generation seven-nanometer process and delivers approximately 640 to 800 teraflops in half-precision floating point, placing it at roughly eighty percent of Nvidia's H100 performance. The chip is priced at approximately 20,000 yuan per unit, about one-fifth the market price of competing Nvidia products. Huawei's CloudMatrix 384 system, clustering 384 Ascend 910C chips, demonstrated inference efficiency exceeding Nvidia's H100 and H800 on DeepSeek-R1 workloads, though it uses 2.67 times more chips and consumes more power. Huawei's executive chairman Xu Zhijun publicly acknowledged that "constrained by inaccessible advanced chip manufacturing processes, Huawei's single-chip computing power still lags Nvidia," but emphasized compensation through cluster architecture innovation. The company announced an ambitious roadmap at its 2025 Connect conference: the Ascend 950 with Huawei's self-developed high-bandwidth memory in the first quarter of 2026, followed by annual generational upgrades through the Ascend 970 in 2028, each targeting a doubling of compute power. Cambricon, another key domestic chip maker, experienced a financial breakout in 2025, with first-half revenue of 2.88 billion yuan representing a forty-four-fold year-over-year increase and its first-ever half-year profit. Its Siyuan 590 chip, fabricated on SMIC's seven-nanometer process, delivers roughly eighty percent of Nvidia A100 performance. Other contenders include Enflame, backed by Tencent and notably still with access to TSMC fabrication, and Moore Threads, which filed for a Shanghai IPO in mid-2025.

Despite this progress, the hardware gap remains substantial. The RAND Corporation estimated in August 2025 that the United States holds a tenfold advantage over China in total AI compute capacity. US technology giants collectively controlled approximately three million H100-equivalent GPUs in 2024, projected to exceed 12.4 million by 2025, compared to China's roughly 360,000 H100-equivalents. China controls an estimated fifteen percent of global AI compute versus seventy-five percent for the United States. SMIC, China's most advanced foundry, can produce seven-nanometer chips but cannot scale production due to the absence of extreme ultraviolet lithography equipment, with yields estimated at roughly twenty percent compared to an industry standard of sixty percent. Nearly all domestic AI chips rely on foreign high-bandwidth memory from Samsung and SK Hynix, a dependency that became more acute after the United States imposed first-time controls on high-bandwidth memory exports in December 2024. DeepSeek itself has repeatedly stated that its single biggest constraint is access to AI compute, and reports indicate the company was forced to restrict API access after the R1 launch due to inference compute shortages. The software ecosystem presents another challenge: Nvidia's CUDA platform dominates AI computing, and Chinese alternatives like Huawei's CANN framework have been described by independent analysts as bug-prone, poorly documented, and unstable compared to the mature CUDA ecosystem.

The US semiconductor export control regime has been the single most significant external constraint on China's AI ambitions. The controls began with the October 7, 2022 rules that banned exports of advanced AI chips including Nvidia's A100 and H100 to China, restricted semiconductor manufacturing equipment, and prohibited US persons from supporting Chinese chip development. Nvidia quickly developed the A800 and H800 chips to circumvent the rules by reducing interconnect bandwidth below regulatory thresholds while maintaining near-equivalent training performance. The October 2023 update closed this loophole by introducing total processing performance metrics, capturing the workaround chips, and expanding the geographic scope to restrict third-country routes. Further tightening in December 2024 added 140 Chinese entities to the Entity List, imposed first-time controls on high-bandwidth memory, and expanded foreign direct product rules. The Biden administration's final major action, the January 2025 AI Diffusion Rule, established a three-tier global licensing framework that amounted to a near-total ban on advanced AI chip imports by China. The Trump administration initially continued the restrictive trajectory, blacklisting additional entities and imposing a sweeping ban on Huawei Ascend chips globally in May 2025, but subsequently modulated policy in response to industry lobbying, granting Nvidia licenses to resume H20 sales in July 2025 and permitting H200 shipments to approved Chinese customers under controlled conditions in December 2025.

The effectiveness of these controls is a matter of genuine debate. They have severely constrained China's domestic chip production capacity and forced reliance on architecturally inferior alternatives. But they have not significantly limited China's ability to train cutting-edge models, as demonstrated by DeepSeek's algorithmic efficiency gains. Smuggling remains a persistent problem, with an estimated 40,000 to 50,000 restricted chips believed to have reached China through informal channels. Chinese firms stockpiled high-bandwidth memory supplies after reports of impending restrictions leaked in mid-2024. China has responded with retaliatory export bans on germanium, gallium, and rare earth materials, an antitrust probe against Nvidia, bans on foreign AI chips in state-funded data centers, and massive state investment in domestic semiconductor capabilities. Peking University announced a prototype extreme ultraviolet lithography light source in March 2025, though it has not yet produced working chips. The overall dynamic is one in which controls have imposed real costs and delays on China's AI hardware development but have simultaneously incentivized innovation in algorithmic efficiency, domestic chip design, and alternative computing architectures.

The Stanford Human-AI Institute's 2025 AI Index Report provides the most rigorous benchmarking of the competitive landscape. In 2023, the gap between top US and Chinese models on key benchmarks stood at 17.5 percent on MMLU, 13.5 percent on MMMU, 24.3 percent on MATH, and 31.6 percent on HumanEval. By the end of 2024, these gaps had shrunk to 0.3 percent, 8.1 percent, 1.6 percent, and 3.7 percent respectively. On the LMSYS chatbot arena, the gap between top US and Chinese models narrowed to within thirty Elo points. However, this convergence at the frontier level masks persistent asymmetries: US private AI investment remains roughly twelve times China's, the US produces more of the most-cited research papers, and American companies continue to lead in frontier model development and commercial AI monetization. China leads in AI patent filings, holding approximately seventy percent of global AI patents, and in total publication volume, accounting for 23.2 percent of global AI research output. Multiple experts caution against a simplistic "arms race" framing, noting that the two countries are in many respects "running in very different lanes," with the United States focused on frontier model development and service applications and China prioritizing industrial integration and mass deployment. A significant talent dynamic is also in play: at least eighty-five prominent Chinese scientists left US institutions for China since early 2025, with the reverse brain drain accelerating under the pressures of visa scrutiny, research budget cuts, and immigration enforcement in the United States.

Military applications of AI are deeply embedded in China's strategic vision. The 2017 plan explicitly called for strengthening military-civil fusion in AI and supporting AI technology for "command decision-making, military simulation, and national defense equipment." The People's Liberation Army's modernization framework proceeds through three overlapping phases: mechanization, informatization, and intelligentization, with the third phase actively underway. In October 2022, Xi Jinping urged the PLA to "speed up the development of unmanned, intelligent combat capabilities," and in April 2024, China established the Information Support Force as a strategic-level branch dedicated to network information systems. PLA theorists have articulated a conceptual framework for intelligentized warfare organized around four transformations: from systems confrontation to algorithm competition, from information-led to machine-dominant warfare, from human decision-making to intelligent decision-making, and from disrupting networks to extreme warfare. Procurement documents analyzed by Georgetown's CSET reveal extensive investment in autonomous unmanned systems across air, ground, sea, and underwater domains, as well as AI-enabled intelligence analysis, cognitive domain operations, and war-gaming simulation. The China Electronics Technology Corporation demonstrated a swarm of 119 fixed-wing drones as early as 2017. Chinese analysts estimate the military edge-AI market will reach 377.5 billion yuan by 2030. Nevertheless, the PLA lacks real-world combat experience with these AI capabilities, open-source information on fielded systems remains limited, and chip bottlenecks constrain military AI compute just as they do civilian applications.

China's AI surveillance apparatus raises distinct data privacy and security concerns that exist in tension with the country's formal regulatory framework. China has constructed a sophisticated data governance regime anchored by three major laws: the Cybersecurity Law effective June 2017, the Data Security Law effective September 2021, and the Personal Information Protection Law effective November 2021. The Personal Information Protection Law, comparable to Europe's General Data Protection Regulation in many respects, requires lawful bases for processing personal information, mandates separate consent for sensitive biometric data including facial recognition, requires personal information protection impact assessments, and grants individuals the right to demand explanations of automated decisions with significant impact. Maximum fines reach fifty million yuan or five percent of prior-year revenue. Facial recognition regulations adopted in September 2024 prohibit compulsory face scanning for routine services and require facial data to be stored locally by default. Yet meaningful constraints on government use of AI surveillance technologies remain conspicuously absent. The Personal Information Protection Law does not regulate the social credit system, courts have prioritized facial recognition's role in "maintaining social stability and promoting economic development" over personal information protection, and the facial recognition regulations contain a notable loophole exempting research and development and algorithm training purposes. The result is a dual-track system in which private companies face increasingly stringent data protection requirements while state surveillance capabilities operate under far fewer restrictions.

China's AI regulatory framework has emerged as one of the most active and distinctive in the world. Rather than pursuing a single comprehensive AI law, China has adopted what might be described as a modular approach, issuing targeted regulations addressing specific AI applications and risks. The Algorithmic Recommendation Regulations, effective March 2022, require transparency, user opt-out rights from personalized recommendations, and mandatory algorithm filing with the Cyberspace Administration of China. The Deep Synthesis Provisions, effective January 2023, mandate real-name user registration, explicit and implicit labeling of AI-generated content, and filing to the algorithm registry. The Interim Measures for Generative AI Services, effective August 2023, establish the principle of "balancing development and security" with "inclusive and prudent" supervision, require content to uphold "core socialist values," and mandate security assessments and algorithm filing for services with public opinion influence, while notably exempting research and development activities not providing services to the domestic public. In March 2025, the Cyberspace Administration issued what has been described as the world's most comprehensive AI content labeling framework, requiring both explicit visible labels and implicit metadata watermarks for all AI-generated content, supported by a mandatory national standard. By April 2025, the CAC's algorithm registry contained 3,739 generative algorithmic tools from approximately 2,353 unique companies, with roughly 250 to 300 new registrations approved monthly, constituting the world's only comprehensive publicly accessible registry of generative AI services.

Despite this regulatory proliferation, China has not yet enacted a comprehensive national AI law. The State Council's legislative work plans for 2023 and 2024 mentioned preparations to submit a draft, but the item was removed from the 2025 legislative schedule, surprising observers. The government instead prioritized pilot programs, standards development, and targeted measures. The AI Safety Governance Framework, first issued by China's Technical Committee 260 in September 2024 and upgraded to version 2.0 in September 2025, evolved from a principled declaration to an operational manual with granular guidance on risk categories and governance mechanisms. Three national standards for generative AI security were released in April 2025 with an effective date of November 2025. The Fifteenth Five-Year Plan calls for strengthening AI governance and improving related laws, policies, application standards, and ethical guidelines, but the timeline for a comprehensive AI law remains unclear. Chinese scholars at the Chinese Academy of Social Sciences have characterized the approach as "actively exploring, legislatively cautious," noting that among major powers, comprehensive AI legislation has become an exercise in "cognitive domain shaping," with some countries advocating regulation for competitors while resisting it for themselves.

China's ethical framework for AI has developed in parallel with its regulatory apparatus but bears distinctive characteristics. The Beijing AI Principles, released in May 2019 by the Beijing Academy of Artificial Intelligence with endorsement from leading universities and the Chinese Academy of Sciences, introduced the concept of "optimizing symbiosis" between humans and AI, drawing from Chinese philosophical traditions of harmony. The National New Generation AI Governance Expert Committee, chaired by Tsinghua University's Xue Lan, published governance principles in June 2019 organized around eight values including harmony and friendliness, fairness and justice, and agile governance. A more detailed ethical specification followed in September 2021, and in March 2022, the Central Committee and State Council jointly established "ethics first" requirements designating AI as one of three key areas for ethics-focused regulation. China's approach differs from Western frameworks in its integration of state goals and "core socialist values" into ethical requirements, its preference for binding laws over voluntary principles, and its emphasis on harmony between humans and AI rather than purely rights-based framing. The gap between principle articulation and implementation enforcement, however, remains pronounced.

On the international stage, China has pursued a deliberate strategy to shape global AI norms and standards. President Xi Jinping launched the Global AI Governance Initiative at the October 2023 Belt and Road Forum, articulating principles of human-centered AI development, equal national development rights, opposition to ideological divisions and exclusionary blocs, and support for establishing an international AI governance body under the United Nations framework. China was among the twenty-eight signatories to the Bletchley Declaration on AI safety in November 2023, though it did not sign the Seoul Ministerial Statement at the May 2024 AI summit and declined to establish an AI safety institute. In July 2024, the United Nations General Assembly unanimously passed a China-sponsored resolution on strengthening international cooperation in AI capacity building, co-sponsored by over 140 countries. China actively participates in ISO and IEC standards development through its national mirror committees, and Chinese national standards increasingly reference international frameworks. Through the Belt and Road Initiative and Digital Silk Road, China exports AI-powered surveillance systems, smart city platforms, and data infrastructure to developing countries, often bundled with infrastructure financing. Huawei's Safe City surveillance platforms have been deployed in multiple partner nations. In 2024, China co-launched Africa's first Center of Excellence in AI and Digital Economy with the United Nations. The strategy amounts to building a parallel ecosystem of AI governance norms anchored in sovereignty, development rights, and opposition to unilateral technology controls, positioning China as the champion of Global South interests in AI governance while advancing its commercial and strategic objectives.

The trajectory of China's AI strategy reveals a nation that has moved from aspirational planning to substantive execution faster than most observers anticipated. The 2017 plan's core industry target of one trillion yuan by 2030 was exceeded five years early. DeepSeek demonstrated that algorithmic innovation can partially compensate for hardware constraints, though the compute gap remains a binding limitation on China's ability to deploy AI at scale. The regulatory framework, while more developed than any other country's targeted AI governance regime, has not yet coalesced into a comprehensive law, and the tension between promoting innovation and maintaining political control continues to shape policy choices. US export controls have imposed genuine costs but have also catalyzed domestic innovation and deepened China's resolve to achieve self-sufficiency in critical technologies. The competitive landscape is neither a simple race with a clear leader nor a symmetrical contest: the United States retains decisive advantages in compute infrastructure, private investment, and frontier model development, while China increasingly leads in industrial AI deployment, open-source model proliferation, patent filings, and state-directed resource mobilization. What is clear is that the era in which AI leadership could be assessed through a single metric or attributed to a single country is over, replaced by a more complex and consequential competition across multiple dimensions of technology, governance, economics, and geopolitical influence.