Artificial Intelligence in China

Introduction

Artificial intelligence (AI) has become a central pillar of China’s national development strategy, permeating economic, social, and geopolitical spheres. Over the past decade, China has rapidly advanced from a technology follower to a global contender in AI, driven by a combination of state-led planning, massive investment, a robust talent pipeline, and a unique regulatory environment. As of early 2026, China’s AI ecosystem continues to be characterized by a dynamic interplay between government policy, industry innovation, academic research, and international engagement. This report provides a comprehensive analysis of the current state of AI in China, examining strategic goals, regulatory frameworks, leading companies and research institutions, technological breakthroughs, sectoral adoption, international collaborations, and the challenges and criticisms that shape China’s AI trajectory.

China’s National AI Strategy and Strategic Goals

Historical Context and Strategic Vision

China’s AI ambitions are rooted in a series of top-level policy documents and long-term plans. The 2017 State Council’s “New Generation Artificial Intelligence Development Plan” (AIDP) marked a watershed moment, setting the explicit goal for China to become the world’s primary AI innovation center by 2030¹². This plan outlined a three-phase roadmap:

• By 2020: Achieve parity with leading countries in AI technology and applications.
• By 2025: Attain global competitiveness in core AI technologies and establish a robust governance framework.
• By 2030: Become the world’s primary AI innovation center, with AI deeply integrated into all sectors of society and the economy.

These ambitions are reinforced by subsequent Five-Year Plans, including the 14th (2021-2025) and the forthcoming 15th (2026-2030), which emphasize technological independence, industrial modernization, and the integration of AI into manufacturing, public services, and national security³.

Strategic Objectives

China’s AI strategy is mission-driven and state-led, positioning AI as a transformative tool for economic modernization, social governance, and geopolitical influence. Key objectives include:

• Technological Self-Reliance: Reduce dependence on foreign technologies, particularly in semiconductors and foundational AI models.
• Industrial Upgrading: Embed AI across manufacturing, logistics, energy, and infrastructure to drive productivity and innovation.
• Social Governance: Utilize AI for public administration, urban management, healthcare, and education.
• Military-Civil Fusion: Integrate civilian and military AI R&D to enhance national security and defense capabilities⁴.
• Global Leadership: Shape international AI standards, governance frameworks, and capacity-building, especially in the Global South⁵.

Investment and Policy Instruments

China’s AI push is underpinned by massive public and private investment. The National AI Industry Investment Fund, launched in 2024 with an initial capital of ¥60 billion (US$8.2 billion), supports startups and foundational research². Local governments compete to attract AI firms through subsidies, tax incentives, and the establishment of AI innovation zones in cities like Beijing, Shanghai, Shenzhen, and Hangzhou⁶. The Bank of China’s AI Industry Development Action Plan commits over ¥1 trillion (US$137 billion) to AI infrastructure over five years.

Key Government Policies and Regulatory Framework

Regulatory Milestones

China has developed one of the world’s most comprehensive and proactive AI regulatory regimes. Key milestones include:

• Algorithm Recommendation Regulation (2022): Mandates registration, audits, and transparency for algorithmic personalization in internet services⁷.
• Deep Synthesis Regulation (2023): Requires labeling of AI-generated content (text, audio, images, video), risk assessments, and prohibits deepfakes without consent⁸.
• Interim Measures for the Management of Generative AI Services (2023): The first administrative regulation specifically targeting generative AI, effective August 15, 2023. It mandates licensing, lawful training data, content moderation, and alignment with “core socialist values”⁹¹⁰⁷.
• Labeling Rules (2025): Effective September 1, 2025, these rules require both explicit and implicit labeling of AI-generated content across all media types.

National Standards and Technical Guidelines

China’s technical standards for AI are rapidly evolving, with the Standardization Administration of China (SAC) and the National Information Security Standardization Technical Committee (TC260) playing leading roles. Recent standards include:

• Cybersecurity Technology-Generative Artificial Intelligence Data Annotation Security Specification: Sets requirements for data labeling in generative AI training¹¹.
• Security Specification for Generative Artificial Intelligence Pre-training and Fine-tuning Data: Ensures the security and integrity of datasets used in model development.
• Basic Security Requirements for Generative Artificial Intelligence Service: Covers user data security, model protection, and service reliability.
• Labelling Method for Content Generated by AI (2025): Details mandatory labeling protocols for AI-generated content.

These standards are complemented by sector-specific guidelines in finance, healthcare, and automotive industries, reflecting a risk-based, sectoral approach to AI governance¹².

Institutional Actors and Governance Architecture

China’s AI governance is characterized by a centralized yet multi-agency architecture:

• Cyberspace Administration of China (CAC): Lead regulator for AI, responsible for algorithm registration, content moderation, and cybersecurity.
• Ministry of Science and Technology (MOST): Oversees research policy, national labs, and ethical guidelines.
• Ministry of Industry and Information Technology (MIIT): Manages industrial implementation, AI standardization, and corporate policy.
• National Development and Reform Commission (NDRC): Coordinates strategic investment and industrial planning.
• Other Ministries: Including Education, Public Security, Health, and the National Radio and Television Administration, each with sectoral oversight⁷.

Local governments operate AI innovation pilot zones, serving as regulatory sandboxes for new technologies and business models¹.

AI Ethics, Safety Frameworks, and Standards

China’s approach to AI ethics is codified in several key documents:

• New Generation AI Governance Principles (2019): Emphasize harmony, fairness, inclusivity, privacy, safety, shared responsibility, openness, and agile governance.
• Ethical Norms for New Generation AI (2021): Require AI to enhance human well-being, promote fairness, protect privacy, ensure controllability, strengthen accountability, and improve ethical literacy.
• Trial Measures for Scientific and Technological Ethics Review (2023): Mandate ethics reviews for AI research involving sensitive areas, with committees established in universities, research institutions, and corporations¹³.
• AI Safety Governance Framework 2.0 (2025): Introduces a taxonomy of AI risks (including open-source model misuse, labor market impacts, and catastrophic risks) and calls for sectoral risk assessment and technical countermeasures.

A distinctive feature of China’s AI ethics regime is the explicit requirement for alignment with “core socialist values,” including prohibitions on content that threatens national security, social stability, or the Party’s image⁷.

Enforcement and Penalties

Regulators possess broad enforcement powers, including security assessments, supervisory inspections, fines (up to ¥50 million or 5% of annual revenue), business suspension, and criminal liability for severe violations¹⁰. Foreign companies providing AI services to Chinese users are subject to these regulations, with the CAC empowered to take technical measures against non-compliant overseas providers.

Major Chinese AI Companies and Their Focus Areas

The Chinese AI landscape is dominated by a mix of tech giants, specialized startups, and research-driven spin-offs. Use the explorer to filter companies by theme, then open the source table when needed.

Analysis

China’s AI giants have developed vertically integrated ecosystems, combining proprietary models, cloud infrastructure, and increasingly, domestic AI chips. Baidu, Alibaba, and Tencent (the “BAT” trio) have each released frontier LLMs (ERNIE, Qwen, Hunyuan) and are rapidly scaling enterprise and consumer deployments. Huawei’s Ascend chip ecosystem, while still trailing Nvidia in efficiency and software maturity, is positioned as the backbone of China’s push for semiconductor independence¹⁵¹⁴.

Startups like DeepSeek, MiniMax, Moonshot, and Zhipu AI have emerged as global contenders, leveraging open-source strategies and academic-industry linkages. DeepSeek’s R1 and V3 models, for example, have achieved near-parity with leading Western models on reasoning and language benchmarks, often at a fraction of the compute and cost¹⁶¹⁷.

Leading Chinese AI Startups and Research Labs

China’s AI startup ecosystem is vibrant and highly competitive, with several “AI Tigers” gaining international prominence:

• DeepSeek: Released DeepSeek-R1 in January 2025, an open-source reasoning LLM rivaling OpenAI’s o1, and continued scaling deployment into 2026. Known for cost-effective training and rapid iteration, DeepSeek’s models are widely adopted in both academic and commercial settings¹⁶¹⁷.
• MiniMax: Specializes in efficient LLM training and conversational AI; its M1 model is noted for high performance on limited GPU resources.
• Moonshot AI: Focuses on multimodal LLMs and innovative applications (e.g., Kimi).
• Zhipu AI: A university spin-off with strengths in medical AI and domestic deployment.
• Baichuan AI: Founded by Tsinghua affiliates, known for large-scale open-source models.
• Stepfun: Developed the first trillion-parameter Chinese AI model.
• 01.AI: Pivoted to enterprise AI apps, leveraging DeepSeek’s models.

These startups benefit from close ties to top universities (notably Tsinghua and Zhejiang), state-backed venture funding, and regulatory sandboxes that facilitate rapid productization².

Chinese Big Tech Firms’ AI Capabilities and Product Deployments

• Baidu: Baidu has transitioned from a search engine giant to a full-stack AI company. Its ERNIE series of LLMs, culminating in ERNIE 5.0 (2025), are natively omni-modal, supporting text, image, audio, and video understanding and generation¹⁸¹⁹. Baidu’s AI cloud business is valued at $34 billion, with its latest disclosed quarterly result (Q3 2025) showing 21% year-over-year growth heading into 2026. The company’s Apollo Go autonomous driving platform has completed over 17 million rides, making it the world’s largest robotaxi service. Baidu’s Kunlunxin chip unit is emerging as a major domestic AI chip supplier, with a five-year roadmap to fill the gap left by Nvidia’s restricted exports¹⁴.
• Alibaba: Alibaba’s Qwen3 LLM and open-source strategy have attracted over 90,000 enterprise users. The company’s Hanguang AI clusters and cloud platform support large-scale inference and deployment. Alibaba Health leverages AI for triage, patient management, and drug discovery.
• Tencent: Tencent’s Hunyuan 3.0 LLM, with 90 billion parameters, is optimized for e-commerce and recommendation systems. Tencent Miying, its AI medical imaging platform, serves over 100 hospitals with 90% accuracy in early cancer screening. WeChat’s AI-powered health mini-programs reach over 80 million users.
• Huawei: Huawei’s Ascend 910C/D chips and CloudMatrix 384 system form the backbone of China’s domestic AI compute infrastructure. The company’s MindSpore framework and CANN programming environment are maturing alternatives to Nvidia’s CUDA ecosystem, though still lagging in global adoption¹⁵.
• ByteDance: ByteDance’s Doubao LLMs power content generation and recommendation algorithms across TikTok/Douyin and Toutiao. The company is investing heavily in AI research and infrastructure, with a focus on global scalability.

Chinese AI Hardware and Semiconductor Ecosystem

Overview

China’s AI hardware ecosystem is at the center of global technological competition. The country has made significant strides in domestic chip design, fabrication, and system integration, though it remains constrained by U.S.-led export controls on advanced GPUs and semiconductor manufacturing equipment²⁰¹⁴.

Key Players and Technologies

• Huawei Ascend Series: The Ascend 910C/D chips, built on 7nm nodes by SMIC, deliver up to 780 TFLOPS (BF16) and are deployed in large-scale clusters (e.g., CloudMatrix 384). While trailing Nvidia’s H100/B200 in per-chip performance and efficiency, Huawei compensates through horizontal scaling and integration with domestic suppliers⁶.
• Baidu Kunlunxin: Baidu’s AI chip unit is rapidly gaining market share, with a roadmap for high-performance chips (M100, M300) targeting LLM training and inference¹⁴.
• SMIC: China’s leading foundry, producing 7nm and (in limited volumes) 5nm chips using DUV lithography. Yields remain below TSMC, but capacity is expanding.
• Cambricon, Biren, Moore Threads, Hygon: Domestic AI accelerator startups supplying chips for inference, video analytics, and edge AI.
• CXMT, YMTC/XMC: Domestic memory (HBM, DRAM, NAND) suppliers, critical for AI compute clusters.

Supply Chain and Export Controls

U.S. export controls, initiated in 2022 and tightened in 2023-2025, continue shaping China’s access to Nvidia’s A100, H100, and H20 chips, as well as advanced semiconductor manufacturing equipment in 2026²⁰¹⁵. Chinese firms have responded by stockpiling GPUs, sourcing via intermediaries, and accelerating domestic R&D. The government’s “Tech Self-Reliance” doctrine and the $48 billion semiconductor fund (2025) continue to subsidize 7nm and 5nm fabrication, with SMIC’s SN3 facility targeting 30,000 5nm wafers monthly by 2026⁶.

Despite progress, domestic chips lag in efficiency, software ecosystem maturity, and yield rates. Chinese AI developers still prefer Nvidia hardware for training frontier models, resorting to black markets and overseas cloud services to access restricted chips¹⁵.

Recent Technological Breakthroughs and Benchmark Achievements

Open-Source LLMs and Model Performance

• DeepSeek R1/V3: Achieved near-parity with OpenAI’s o1 on reasoning, math, and code benchmarks, often with significantly lower compute and cost. DeepSeek’s open-source approach has catalyzed global adoption and downstream innovation¹⁶¹⁷.
• Alibaba Qwen3: Ranks near the top of global leaderboards in reasoning and language tasks, with strong multilingual capabilities.
• MiniMax M1: Demonstrated competitive performance on just 512 Nvidia H800 GPUs, highlighting efficiency innovations.
• Baidu ERNIE 5.0: Natively omni-modal, supporting joint modeling of text, images, audio, and video. Powers over 400,000 enterprise apps and the world’s largest robotaxi fleet¹⁹.

Hardware and System-Level Innovations

• Huawei CloudMatrix 384: An AI server rack integrating 384 Ascend 910C processors, forming a high-bandwidth, all-optical network. While more power-intensive than Nvidia’s GB200, it demonstrates China’s ability to scale domestic hardware for large model training⁶.
• SMIC 7nm/5nm Production: Despite lacking EUV lithography, SMIC has achieved 7nm production using DUV multi-patterning, though yields and costs remain challenges.

Academic and Research Achievements

• Tsinghua University: Leads the world in AI citations and patents, with over 900 AI-related patent filings in 2024 alone. Tsinghua-affiliated startups (e.g., DeepSeek, Baichuan, MiniMax, Moonshot, Zhipu) are at the forefront of LLM development and commercialization²¹²².
• China’s AI Research Output: In 2024, China produced more AI research publications than the US, UK, and EU combined, with over 30,000 active AI researchers and a rapidly growing talent pipeline²¹.

International Collaborations and Multilateral Initiatives

Global AI Governance

China has positioned itself as a champion of multilateral AI governance, advocating for the United Nations as the primary venue for international rulemaking⁵. Key initiatives include:

• Global AI Governance Action Plan (2025): A 13-point roadmap emphasizing infrastructure development, sectoral application, data quality, open-source collaboration, sustainability, and capacity-building for developing countries⁵.
• Bletchley Declaration (2023): China signed the UK-led declaration on AI safety.
• UN Resolutions (2024): Co-sponsored resolutions on AI safety and closing the AI access gap.
• Paris AI Declaration (2025): Joined 60 nations in committing to shared standards and international coordination.

China actively participates in international standards bodies (ISO, IEC, ITU) and hosts AI standardization forums. The Digital Silk Road exports AI infrastructure and expertise to Southeast Asia, Africa, and Latin America.

US-China and Western Relations, Export Controls, and Their Effects

The US has implemented a series of export controls since 2022, targeting advanced AI chips (Nvidia A100, H100, H20), semiconductor manufacturing equipment, and memory (HBM, DRAM)²⁰¹⁵. These measures aim to slow China’s AI progress, particularly in military and dual-use applications. The controls have:

• Constrained Compute: Chinese AI firms face a tenfold gap in compute capacity compared to US counterparts, with limited access to cutting-edge GPUs.
• Accelerated Domestic R&D: Spurred investment in domestic chip design, fabrication, and alternative supply chains.
• Prompted Stockpiling and Smuggling: Chinese companies have stockpiled Nvidia chips and resorted to black markets and overseas cloud services.
• Shifted Market Dynamics: Domestic chipmakers (Huawei, Baidu Kunlunxin, Cambricon) are gaining market share, though software ecosystem and performance gaps remain.

In November 2024, the two countries agreed to avoid giving AI control over nuclear weapons systems. However, fundamental differences persist on values⁵.

Academic and Research Institutions Driving AI in China

• Tsinghua University: The epicenter of China’s AI ambitions, leading in citations, patents, and talent production. Tsinghua-affiliated startups are at the forefront of LLM development and commercialization²³²².
• Peking University, Shanghai Jiaotong, Zhejiang University: Major contributors to AI research, talent training, and industry partnerships.
• Chinese Academy of Sciences (CAS): Drives fundamental research and innovation.
• Shanghai AI Laboratory (SHLAB): State-affiliated, leading open-source platforms (OpenMMLab) and AI safety research.

Talent, Publications, and Patents

China leads the world in AI patent filings, with over 378,000 effective AI invention patents by the end of 2023—a 40% year-on-year growth rate²². The country accounts for over 70% of global generative AI patents (2014-2023) and produces more AI research publications than the US, UK, and EU combined²². Early AI education (beginning at age six), massive STEM graduation rates, and government-backed talent programs (e.g., Thousand Talents Plan) ensure a robust pipeline of AI professionals.

Sectoral AI Adoption: Healthcare, Finance, Transportation, Manufacturing

• Healthcare: AI is transforming China’s healthcare sector, with applications in diagnostics, medical imaging, drug discovery, and hospital management. The National Medical Products Administration (NMPA) classifies all AI medical software as Class III devices, subject to stringent review. AI-powered platforms serve hundreds of millions of users, and China leads globally in healthcare AI patents¹².
• Finance: AI is widely used for credit scoring, fraud detection, automated trading, and risk management.
• Transportation: China is a global leader in autonomous driving, with Baidu’s Apollo Go completing over 17 million rides.
• Manufacturing: AI-driven automation, robotics, and predictive maintenance are central to the “Made in China 2025” initiative.
• Other Sectors: AI is also transforming agriculture, education, legal services, and public administration.

Ethical, Human Rights, and Censorship Concerns and Criticisms

• Content Moderation and Censorship: AI regulations require strict alignment with “core socialist values,” prohibiting content that threatens national security or the Party’s image. AI models are trained on censored datasets, leading to biases and limitations in expression¹⁷.
• Privacy, Data Protection, and Bias: The Personal Information Protection Law (PIPL) and Data Security Law establish comprehensive data privacy frameworks. However, challenges remain in ensuring data quality, preventing algorithmic bias, and balancing transparency with confidentiality¹².
• Human Rights and Surveillance: AI is extensively used for surveillance (Skynet, Sharp Eyes) and social governance, raising concerns about privacy and civil liberties, including the targeting of ethnic minorities such as the Uyghur population.

Challenges, Constraints, and Future Outlook

Challenges and Constraints

• Compute and Hardware Bottlenecks: China’s AI development is constrained by limited access to cutting-edge GPUs and semiconductor manufacturing equipment. Domestic chips lag in efficiency, yield, and software ecosystem maturity¹⁵.
• Talent Drain and Research Quality: Concerns persist about the quality and global impact of research, with many top Chinese AI scientists working for US universities and tech firms²³.
• Regulatory Complexity and Innovation Drag: Pioneering AI regulations may slow innovation for public-facing products, and fragmentation between agencies creates uncertainty².

China’s Global Positioning and Comparative Strengths vs. US/EU

China’s strengths include the largest AI talent pool, user base, and data resources; rapid diffusion and commercialization; state-led coordination; a strong open-source strategy; and patent leadership. The US retains advantages in foundational model innovation and advanced semiconductor design.

Future Outlook and Likely Policy Trajectories

China’s AI trajectory will be shaped by:

• Technological Independence: Continued investment in domestic chips and foundational models.
• Agile Regulation: Ongoing refinement of AI laws, standards, and risk assessment frameworks.
• Global Governance Leadership: Active participation in international standard-setting, especially through the UN.
• Sectoral Deepening: Expansion of AI adoption in healthcare, manufacturing, transportation, and public services.
• Military-Civil Fusion: Further integration of civilian and military AI R&D.

China’s 15th Five-Year Plan (2026-2030) will prioritize technological independence, industrial modernization, and high-quality growth, positioning AI as a central driver of economic and social transformation³.

Conclusion

China’s AI ecosystem in 2026 is marked by remarkable progress, strategic ambition, and complex challenges. The country has established itself as a global leader in AI research, patents, and sectoral adoption, with a unique blend of state-led planning, industry innovation, and academic excellence. Regulatory sophistication, open-source strategies, and international engagement have enabled China to close the gap with the US and shape the global AI landscape. However, persistent constraints in hardware, talent retention, and international trust, as well as ethical and human rights concerns, present ongoing hurdles. The coming years will determine whether China can sustain its momentum, achieve technological self-reliance, and realize its vision of becoming the world’s primary AI innovation center by 2030.