How China Navigates the Global AI Chip Landscape
For years, global observers have watched China's rapidly expanding technological ambitions, often characterized by a blend of state-led initiatives and private sector ingenuity. Today, as Artificial Intelligence reshapes industries worldwide, China has intensified its pursuit of self-reliance, charting a course that prioritizes domestic innovation across its entire AI technology stack. This strategic pivot, driven by national security concerns and geopolitical dynamics, reshapes its technological landscape and challenges established global leaders.
Source: Own illustration
Quick Summary
- Strategic Imperative: China aims for AI leadership and self-reliance, driven by national and economic security concerns.
- US Export Controls: Restrictions on advanced AI chips have accelerated China's domestic development efforts.
- AI Stack Development: China is building its own solutions across chips, machine learning frameworks, and large language models (LLMs).
- Hardware Challenges: Chinese-made AI chips still lag Nvidia’s performance, but domestic market share is growing.
- Software & Open Source: China actively contributes to global open-source ML frameworks and develops its own alternatives.
- LLM Advancements: Significant progress in LLMs like DeepSeek, fueled by private funding and open-source models.
- Talent & Funding: China is rapidly attracting and retaining top AI talent, with increasing domestic investment.
- Future Outlook: Despite challenges, China’s strategic focus aims for an increasingly independent AI ecosystem.
The Strategic Imperative for AI Self-Reliance
China aims for a leading role in Artificial Intelligence (AI), quantum technology, and other cutting-edge fields, as outlined in its
15th Five-Year Plan. This plan mandates increased original scientific research to foster the nation's self-sufficiency. China views AI as critical for both national and economic security, prompting a comprehensive drive for self-reliance across the entire technology stack.The United States implemented export controls on advanced AI chips, chip design software, and manufacturing equipment in October 2022, intensifying these restrictions in 2023 and 2024. These measures directly targeted China's AI development efforts. In response, Beijing declared a "self-sufficient and controllable" AI ecosystem a primary objective. Xi Jinping further underscored this commitment in April 2025, calling for a nationwide mobilization to achieve "self-reliance and self-strengthening" in technology.
❝ self-reliance and self-strengthening ❞
President of China
Source: en.wikipedia.org
Xi Jinping has called for nationwide mobilization to achieve "self-reliance and self-strengthening" in technology.
China's AI Stack: Building from the Ground Up
A simplified AI stack consists of three layers: chips for computation, machine learning frameworks for model creation, and applications like Large Language Models (LLMs). China has made substantial progress in developing its own solutions across all these layers, from LLMs like DeepSeek to the foundational semiconductors driving AI workloads.
AI Chips: The Foundational Layer
At the base layer, chips remain a significant challenge. While China boasts several AI chip design companies such as Cambricon, Moore Threads, and Kunlun, these firms still depend on ARM microarchitecture and Western electronic design automation (EDA) software. Huawei, however, leads efforts to indigenize EDA software and serves as a primary actor coordinating chip production, working closely with Semiconductor Manufacturing International Corporation (SMIC). SMIC stands as the only Chinese company capable of producing advanced 7-nanometer chips today.
The “Big Fund,” established in 2014, funnels significant capital into domestic chip development and manufacturing. Its third phase, announced in 2024, allocated 340 billion CNY for this purpose. Huawei’s investment arm, Hubble, strategically invests across the supply chain, often co-investing with state-backed funds like the
Shenzhen Major Industry Investment Group.Despite these efforts, the performance of Chinese-made AI chips has not yet matched that of market leader Nvidia. In 2024, Nvidia sold over 1 million H20 chips in China, while Huawei sold only 200,000 AI chips during the same period, despite offering lower prices. DeepSeek, for instance, found Huawei's Ascend 910C insufficient for large-scale LLM training. Nevertheless, China’s market share for domestic AI accelerator servers reached nearly
41 percent by 2025, indicating a significant structural shift in the market. Huawei aims to secure 50% of the Chinese AI chip market by 2026.
Source: webpronews.com
Huawei’s Ascend 910C chip has been found insufficient for large-scale LLM training, showcasing the ongoing challenge in matching Nvidia’s performance.
Here's a comparison of AI chip market share in China (2025 estimates):
| Vendor | Estimated Shipments (Units) | Market Share (%) |
|---|---|---|
| Nvidia | 2,200,000 | 55% |
| Huawei | 812,000 | ~20% |
| Alibaba T-Head | 265,000 | ~7% |
| Baidu Kunlunxin | 116,000 | ~3% |
| Cambricon | 116,000 | ~3% |
| Other Chinese Vendors | ~457,000 | ~12% |
| AMD | 160,000 | ~4% |
| Total Estimated Market | ~4,126,000 | 100% |
Machine Learning Frameworks and Open Source Contributions
Machine learning frameworks, such as PyTorch and TensorFlow, both open-source platforms originally developed by US companies, form the middle layer of the AI stack. Chinese developers contribute to these frameworks, integrating support for domestic GPUs, and also develop their own alternatives. Baidu's PaddlePaddle, launched in 2016, enjoys widespread industrial adoption within China. Huawei's MindSpore, open-sourced in 2020, functions as a full-scenario AI computing framework. Recognizing the importance of global collaboration, Huawei joined the PyTorch Foundation as a Premier member in 2023.
China has long championed the open-source movement to mitigate reliance on foreign technology companies. The Open Atom Foundation, established in 2020, promotes open-source development in advanced technology projects. Chinese developers constituted about 9 percent of all developers on GitHub in 2024. The Ministry of Industry and Information Technology (MIIT) also supported Gitee, a local hosting platform, as an alternative to GitHub in 2020. Alibaba Cloud (Aliyun) leads Chinese AI open-source endeavors; its Qwen series consistently tops open-source leaderboards, and it operates ModelScope, an open-source AI model platform similar to Hugging Face.
Large Language Models and Applications
China’s LLM landscape thrives with numerous academic and private ventures developing cutting-edge models. Ample private funding and access to global open-source models have fueled rapid progress. DeepSeek models, released in late 2024, positioned China at the forefront of global LLM development. DeepSeek’s R1 model, released in January of this year, demonstrated remarkable efficiency in computational resource utilization by combining existing technical solutions to drastically reduce computing demands. Many leading Chinese LLMs leverage the underlying architecture of Meta's LLaMA series, including Baichuan's Baichuan series and 01.AI's Yi series.
While the government provides less direct funding for LLMs and applications, it cultivates a supportive ecosystem. Provincial governments, such as Shenzhen’s, offer vouchers of up to 500 million CNY annually to startups for computing power. Hangzhou, an emerging AI innovation hub, hosts six AI startups, including DeepSeek, that drive this new wave of innovation.
Hardware limitations, specifically the scarcity of advanced AI chips resulting from US export controls, compel China's AI industry to increasingly focus on specialized applications. Qihoo 360 CEO Zhou Hongyi suggests a vertical approach, endorsing smaller models and proprietary data for resource-constrained Chinese companies. The government's "AI+ Initiative," prominently featured in the "Two Sessions" of 2024 and 2025, champions the integration of AI across manufacturing, electric vehicles, robotics, education, and medicine. This initiative also promotes smart data centers, often with implicit requirements for purchasing Chinese technology.
Talent, Funding, and Global Collaboration
China has rapidly gained ground in attracting and retaining elite AI talent. In 2022, a quarter of the world's top AI scientists earned their bachelor's degrees at Chinese universities. Twenty-eight percent of these top talents live and work in China, a substantial increase from 11 percent three years prior. China also accounted for approximately 40 percent of highly cited AI papers in 2021, surpassing the US share as early as 2016. However, a significant portion of this research emerged from collaborations with US colleagues, and AI research cooperation between China and the US has largely remained stable despite geopolitical tensions.
US investment in China’s AI ecosystem significantly declined, reaching a ten-year low of 1.3 billion USD in 2022. Chinese AI firms now seek alternative capital sources, including CNY-denominated funds and investments from Gulf states. In 2024, China recorded 715 AI sector deals totaling 7.3 billion USD, surpassing any other Asian country. In January, China launched a new AI investment fund worth 60 billion CNY (8.2 billion USD).
Challenges and the Path Ahead
Despite rapid advancements, challenges persist. Huawei chips, for instance, still lag Nvidia’s in FP8 support and manufacturing scale. Furthermore, SMIC’s lower yield rates in manufacturing pose a risk. Analysts at UBS predict that China’s overall chip self-sufficiency, encompassing DRAM, analog, and equipment, will rise to 27% by the end of 2025, up from mid-teens during the pandemic. Morgan Stanley foresees China’s AI chip self-sufficiency soaring to 82% by 2027, a significant leap from last year's 34%.
Limited access to advanced chips and the future of China's participation in the global open-source community remain key vulnerabilities. The Politburo Study Session on AI in March 2026 underscored self-reliance, scaling back rhetoric regarding openness and global integration. Data quality and availability for model training also present ongoing issues, prompting the National Data Administration to launch an initiative for better integrating China’s data market.
Source: dreamstime.com
Data quality and availability are ongoing issues, prompting the National Data Administration to integrate China’s data market.
What is China's primary goal regarding AI?
China aims to achieve a leading role in AI and full technological self-reliance, driven by national and economic security concerns. This includes developing domestic solutions across the entire AI technology stack.
How have US export controls impacted China's AI development?
US export controls on advanced AI chips and manufacturing equipment have accelerated China's efforts to develop its own domestic alternatives and foster a "self-sufficient and controllable" AI ecosystem.
What progress has China made in AI chips?
While Chinese AI chips, like Huawei's Ascend series, are improving and gaining market share, they still lag behind Nvidia’s top-tier performance, particularly for large-scale LLM training. However, domestic market share is growing significantly.
Is China involved in the global open-source AI community?
Yes, China actively contributes to global open-source machine learning frameworks like PyTorch and TensorFlow. It also develops its own open-source alternatives and platforms, like Baidu's PaddlePaddle and Alibaba's ModelScope.
What is the "AI+ Initiative"?
The "AI+ Initiative" is a government strategy to integrate AI across various sectors, including manufacturing, electric vehicles, robotics, education, and medicine, promoting the widespread application of AI models and smart devices.
Conclusion
China’s pursuit of AI self-reliance is a complex, multifaceted endeavor driven by both economic ambition and geopolitical necessity. Bolstered by substantial state investment, a thriving open-source community, and a growing pool of AI talent, China has demonstrated remarkable progress across the AI stack, from foundational chips to sophisticated large language models. While hardware limitations and geopolitical tensions present notable hurdles, particularly regarding advanced chip manufacturing, China's strategic focus on domestic innovation and application-oriented development underscores its determination to emerge as a global leader in artificial intelligence. The trajectory suggests an increasingly independent AI ecosystem within China, potentially reshaping global technological power dynamics for decades to come.