In its 15th Five-Year Plan (2026-2030), China has placed artificial intelligence at the centre of its industrial strategy, but with apparently clashing goals. China wants technological self-reliance and substantial safeguards around its domestic market, but also freely adaptable AI models, international diffusion of Chinese AI models and greater cooperation in global AI governance.
However, while these objectives appear contradictory, a closer look reveals that they’re different facets of a unified strategy. Chinese President Xi Jinping detailed the strategy at the 17-20 July 2026 World AI Conference in Shanghai. He presented AI as a “rare, historic opportunity” for open development, collaboration and sharing, and promoted the newly established Shanghai-based World Artificial Intelligence Cooperation Organization (WAICO) as a vehicle for a “just and equitable” system of global AI governance (Xi, 2026).
In this analysis, we examine how China seeks to assert AI leadership with itself at the centre of innovation, diffusion, standard-setting and adoption, while also insulating its domestic market. We also compare this strategy with those of the United States and the European Union.
Economic competitiveness
China wants AI to drive the productivity gains it needs to mitigate its structural deceleration: the exhaustion of its investment-driven growth model, especially in real estate, and the demographic challenges it faces (García-Herrero, 2025). China’s 15th Five-Year Plan lists AI as one of the “new quality productive forces” (新质生产力) in its technology-driven growth model. The plan also mentions artificial general intelligence as a national ambition, meaning its aspirations go beyond near-term industrial applications and extend to cutting-edge frontier model development.
Chinese policymakers expect AI to increase productivity beyond industrial automation into service sectors such as healthcare and education. For example, a Chinese ‘Action Plan for Artificial Intelligence + Education’, calls for use of AI to prepare lessons and to match students who have particular skillsets with universities and training programmes for emerging industries (MoE, 2026). Authorities are betting that AI can sustain growth even as the economy tilts towards services – a transition that slowed productivity in Europe and elsewhere (Duernecker et al, 2023).
The economic motive is of course not unique to China. The global dominance of US tech firms makes remaining at the AI frontier commercially imperative, while the EU has identified AI and digital infrastructure as essential to meet competitiveness challenges (Draghi, 2024).
Strategic autonomy through technological sovereignty
A second, and in many ways more consequential, reason for the pursuit of AI is the desire to reduce technological dependency. China sees this as a geopolitical necessity. Its dependence on foreign components for AI infrastructure was exposed by US export controls on advanced semiconductors in October 2022, further tightened in 2023, on the basis that allowing China access to advanced semiconductors represented a national security vulnerability.
In April 2025, Xi Jinping urged a nationwide mobilisation to achieve “self-reliance and self-strengthening” (自立自强) and build an “independent and controllable” (自主可控) AI ecosystem using domestic hardware and software (Chang et al, 2025). China is already embarking on this strategy with the development of a domestic AI stack – the technology required for successful development and dissemination of AI, from AI chips through machine learning to large language models. This domestic drive is paired with external engagement: encouraging adoption of Chinese AI platforms expands their ecosystem and positions Beijing rather than Washington at the centre of global AI development.
The US faces a different sovereignty vulnerability: the concentration of advanced semiconductor manufacturing in Taiwan. To remedy this, Washington has pressured and offered grants, loans tax credits and other incentives to encourage Taiwan’s leading chipmaker, TSMC, to produce cutting-edge semiconductors in Arizona.
Europe’s position is structurally the weakest. It remains dependent on US cloud providers and foundation models and on Taiwan for advanced chip fabrication. Its ‘digital sovereignty’ push (Di Marco et al, 2025) has yet to produce infrastructure capable of substantially reducing these dependencies, although locally hosted foreign models may mitigate some concerns over foreign control and leverage.
Military advancement
China’s third objective is AI for military capability, and is the least publicly acknowledged but arguably the most significant goal.
AI is already integrated into autonomous systems, intelligence analysis, logistics and battlefield decision support in both the US and Chinese armed forces (Fedasiuk et al, 2021). While the US maintains the world’s largest military-AI investment, China is integrating civilian technological and industrial capabilities with military research, procurement and production, and explicitly conceptualises an AI-enhanced military (军事智能化). The military aspect was, however, largely absent from Xi’s WAIC address.
Europe’s military-AI capabilities remain fragmented, although rearmament spurred by Russia’s invasion of Ukraine in 2022 has begun to stimulate defence-related AI investment (EDA, 2024).
China’s coordinated industrial policy across the full AI stack
China’s approach to AI is best understood as a state-coordinated innovation ecosystem, in which planning, industrial policy, infrastructure investment and selective regulation reinforce one another across every layer of the stack – echoing Xi’s call for a ‘secure and controllable’ ecosystem.
For chip manufacturing, the state provides direct and sizable support. China’s National Integrated Circuit Industry Investment Fund deployed $20 billion in its first phase from 2014, $29.5 billion in its second phase from 2019 and $47.5 billion in its third phase from 2024. Chang et al (2025) estimated that Huawei anchors the semiconductor ‘national team’, with access to up to 70 percent of China’s advanced capacity from China’s leading semiconductor company (SMIC). An $8.2 billion National AI Fund was established in early 2025, targeting computing power, algorithms, data and applications (赋能应用) across the entire AI industry value chain.
For model and application development, the state enables rather than directs. Instead of funding large language model development outright, Beijing offers compute subsidies through local voucher programmes (up to 500 million renminbi annually in Shenzhen) and government guidance funds – a form of public-private partnership that incentivises where private capital should flow (Beraja et al, 2024). Domestic firms are further shielded by regulatory barriers that make it difficult for foreign providers to offer generative-AI services in China. Although the infant-industry rationale remains contested in the economic literature (Harrison and Rodríguez-Clare, 2009), in China, a protected home market has allowed domestic platforms to accumulate users and data and generate revenues to fund model development.
Since 2024, Chinese AI policy has turned more decisively towards economy-wide adoption. The 2024 ‘AI+’ initiative (expanded in 2025) aims to facilitate AI adoption across industrial and service sectors. It builds on earlier initiatives for open-innovation platforms and AI innovation application pilot zones.
China’s strategy to publish the weights behind its leading AI models to facilitate adoption, localise user data and crowd-source improvements, is its sharpest competitive tool. Beijing embraced open-weight development (see footnote 1) to lower entry barriers at home and to reduce reliance on sanction-vulnerable foreign intellectual property. Shared on Hugging Face and GitHub, Chinese models have become central to the global AI landscape (Meinhardt et al, 2025).
However, this openness is not a concession. It is both a response to export controls and a bid for global standard-setting – the goals laid out at WAICO. For European firms, the practical implication is uncomfortable: the most accessible frontier-adjacent models are increasingly Chinese, potentially creating a new dependency.
Thus, the Chinese model is a form of tailored state support, matching the type of intervention to the vulnerability of each layer – strong where import dependencies bite and light where closed domestic markets protect. The heaviest direct support is reserved for capital-intensive semiconductors, with less-intensive enabling measures higher up and a protected market in which domestic platforms have a built-in advantage.
The US: private-sector primacy and its risks
Differently to China, the US authorities enable rather than direct innovation via defence procurement, basic research, export-control policies and regulatory facilitation. Investment, meanwhile, is controlled by a handful of very large private firms. NVIDIA dominates hardware, while infrastructure investment is concentrated among a handful of US hyperscalers. Alphabet, Amazon, Meta, Microsoft and Oracle are expected to jointly account for around $750 billion in capital expenditure in 2026 alone (S&P Global Ratings, 2026), equivalent to roughly 0.6 percent of global GDP. That investment is funded by cloud revenues and advertising margins, which fund AI divisions that would be loss-making on a standalone basis.
Washington has grown more interventionist, mainly on governance and export controls. The Biden administration established the first federal framework for AI governance focused on safety and privacy rather than catalysing innovation, while the Commerce Department introduced export controls in 2022 and tightened them in 2023 to restrict China’s access to advanced AI chips and lithography machines. President Donald Trump has since pivoted to deregulation by revoking Biden’s governance framework through executive action, while making chip controls targeting China more transactional with negotiated export licenses that give Chinese purchasers access to chips while returning a portion of revenues to the US government (NVIDIA, 2025).
The broader strategy carries two risks. Private investors may prove less patient than a state as valuations demand returns, and the sheer speed of US compute build-out is outrunning public electricity grids. Whereas China’s binding constraint is access to chips, the emerging US constraint is energy and the patience of capital.
The European Union: regulatory leadership, policy fragmentation
Europe’s AI approach is distinctive: it has produced the world’s most comprehensive AI governance framework, which is designed to complement policies aimed to stimulate innovation. However, the EU still cannot finance a coherent industrial policy to close the gap with the US and China. The EU AI Act (Regulation (EU) 2024/1689), in force since August 2024, is a risk-based framework that bans the most harmful uses of AI, imposes strict conditions for high-risk systems and transparency obligations for the rest.
But regulatory leadership has not brought a surge in innovation or adoption. Compliance costs may be modest for large firms and costly for small and medium-sized firms (Mariniello, 2026) Binding restrictions and uncertainty over applicable standards make it difficult to build infrastructure in the EU, especially for startups and scale-ups. Recognition that the bloc was falling behind in AI infrastructure and adoption has led to a regulatory simplification push, with measures including delayed application of requirements on high-risk AI and clarification that legitimate interest can be a legal basis for processing personal data in AI development and training. Meanwhile, a range of EU initiatives, policies, programmes and financing vehicles that could focus on AI share the weakness that they are fragmented and hard to navigate.
Conclusions
China’s approach to AI is resourceful and highly strategic, focusing on state backing for development of champions, directing compute to them, protecting their home market and promoting their open-weight models abroad within a coherent framework. The US approach is less coordinated but more market-driven, relying on titanic private investment along with state intervention through research funding, export controls and incentives for semiconductor investments. For Europe, the critical question is whether its institutions can translate its ambition to lead in AI into comparably coherent action at the speed the technology necessitates.
References
Chang, W., R. Arcesati and A. Hmaidi (2025) China’s drive toward self-reliance in artificial intelligence: from chips to large language models, MERICS Report, Mercator Institute for China Studies, available at https://merics.org/en/report/chinas-drive-toward-self-reliance-artificial-intelligence-chips-large-language-models
Beraja, M., W. Peng, D.Y. Yang and N. Yuchtman (2024) ‘Government as Venture Capitalists in AI’, NBER Working Paper 32701, National Bureau of Economic Research, available at https://www.nber.org/papers/w32701
Di Marco, D., S. Thabit Gonzalez, A. Kotsev, A. Friis-Christensen, M. Minghini and M. King (2025) ‘Open but Not Powerless: Towards a Common Understanding of EU Digital Sovereignty’, Science for Policy Brief, European Commission Joint Research Centre, available at https://publications.jrc.ec.europa.eu/repository/handle/JRC144908
Draghi, M. (2024) The Future of European Competitiveness, European Commission, available at https://commission.europa.eu/topics/competitiveness/draghi-report_en
Duernecker, G., B. Herrendorf and Á. Valentinyi (2023) ‘Structural change and productivity growth in Europe — Past, present and future’, European Economic Review 151: 104329, available at https://www.sciencedirect.com/science/article/abs/pii/S0014292122002094?via%3Dihub
EDA (2024) 2024 Coordinated Annual Review on Defence (CARD) Report, European Defence Agency, available at https://eda.europa.eu/docs/default-source/documents/card-report-2024.pdf
EUCLERA (2026) Outline of the 15th Five-Year Plan (2026-2030) for National Economic and Social Development of the People’s Republic of China, available at https://www.wko.at/ktn/aussenwirtschaft/euclera-translation-15th-five-year-plan-2026-2030-.pdf
European Commission (2025) ‘AI Continent Action Plan’, COM(2025) 165 final, available at https://eur-lex.europa.eu/legal-content/EN/TXT/?uri=COM:2025:165:FIN
Fedasiuk, R., J. Melot and B. Murphy (2021) Harnessed Lightning: How the Chinese Military is Adopting Artificial Intelligence, Center for Security and Emerging Technology, Georgetown University, available at https://cset.georgetown.edu/publication/harnessed-lightning/
García-Herrero, A. (2025) ‘Ten challenges facing China’s economy’, Analysis 25/2025, Bruegel, available at https://www.bruegel.org/analysis/ten-challenges-facing-chinas-economy
Harrison, A. and A. Rodríguez-Clare (2009) ‘Trade, Foreign Investment, and Industrial Policy for Developing Countries’, NBER Working Paper 15261, National Bureau of Economic Research, available at https://www.nber.org/papers/w15261
Mariniello, M. (2026) ‘The right balance: how to fix European Union artificial intelligence regulation’, Policy Brief 12/2026, Bruegel, available at https://www.bruegel.org/policy-brief/right-balance-how-fix-european-union-artificial-intelligence-regulation
Martens, B. and T. Schenk (2026) ‘How can Europe address its pressing AI compute infrastructure shortfall?’ Policy Brief 18/2026, Bruegel, available at [ADD DOI]
Meinhardt, C., S. Nong, G. Webster, T. Hashimoto and C.D. Manning (2025) ‘Beyond DeepSeek: China’s Diverse Open-Weight AI Ecosystem and Its Policy Implications, Stanford HAI/DigiChina’, Issue Brief, Stanford HAI/DigiChina, available at https://hai.stanford.edu/policy/beyond-deepseek-chinas-diverse-open-weight-ai-ecosystem-and-its-policy-implications
MoE (2026) 教育部等五部门关于印发《“人工智能+教育”行动计划》的通知, Ministry of Education of the People’s Republic of China, available at http://www.moe.gov.cn/srcsite/A16/s3342/202604/t20260410_1433240.html?utm_sou
S&P Global Ratings (2026) ‘Will Rising Capex Test Hyperscalers’ Credit Strength?’ CreditWeek, 14 May, available at https://www.spglobal.com/ratings/en/regulatory/article/creditweek-will-rising-capex-test-hyperscalers-credit-strength-s101685676
Xi, J. (2026) ‘Joining Hands to Build a Just and Equitable System for Global AI Governance’, keynote address at the opening ceremony of the 2026 World AI Conference and High-Level Meeting on Global AI Governance, Shanghai, 17 July, available at https://www.fmprc.gov.cn/eng/xw/zyjh/202607/t20260717_11984766.html