China's AI development

Notes

• China is promoting an open-source/open-weight AI strategy, exemplified by models like Moonshot AI's Kimi K3.
• Kimi K3 is a 2.8 trillion parameter model, demonstrating that China is overcoming GPU export restrictions through abundant electricity and the use of larger quantities of less advanced chips.
• Chinese firms like Alibaba, Tencent, and Baidu are encouraged by the state to develop AI capabilities, leading to widespread adoption of their models on platforms like HuggingFace.
• The Chinese government is actively protecting its AI firms from foreign acquisition and restricting key personnel travel to maintain domestic control over the AI ecosystem.
• Open-source models offer cost predictability and transparency, potentially challenging the dominance of Western 'captive' models (e.g., OpenAI, Anthropic) that restrict source code access.
• There is a global trust deficit regarding Chinese AI models, particularly in India, due to geopolitical concerns and potential data security risks.
• India is focusing on developing domestic AI champions (e.g., Sarvam AI) while maintaining a cautious approach toward Chinese capital in the technology sector.
• China's future AI strategy aims to transition from digital applications to physical world integration.

Part of a longer story

This is day 2 of 6 in China's AI development and governance, which has been running since 18 July 2026. Reading it whole is usually worth more than reading today alone — the exam asks how something developed.

• China's AI governance initiative Before
• China's AI development and global strategy After
• The whole story 6 entries