China accelerates sovereign AI-computing and software buildout amid NVIDIA export controls

China reportedly targets 9,800 eflops of AI-computing capacity by 2030, approximately 3.8 trillion yuan in infrastructure investment, and clusters reaching 100,000 cards. Its software-modernization roadmap adds targets for 20,000 major software firms, 100 agent applications, and five open-source projects by 2028; domestic GPU substitution remains constrained by memory, manufacturing, and ecosystem gaps.