数字经济周报（2026年第5期）：智能体获关注，地方政府拟推“龙虾”支持措施-20260309

Group 1: Policy Focus - The Shenzhen Longgang District proposed the "Lobster Ten Measures" to support the development of OpenClaw, aiming to attract global intelligent agent developers and One Person Companies (OPC) with a "zero-cost startup" approach [5][6] - The measures include free deployment and development support, enhanced data and computing resources, innovation rewards, and talent attraction initiatives [6][7] Group 2: Market Overview - From March 2 to March 6, 2026, the A-share and Hong Kong technology sectors experienced a downward trend due to geopolitical factors, while the U.S. tech sector showed slight fluctuations [8][13] - The Sci-Tech Innovation Board and Hang Seng Technology Index rebounded on March 6 following the release of the government work report emphasizing technology-led productivity [8] Group 3: AI Industry Dynamics - The global AI industry is characterized by a "dual-core drive" from the U.S. and China, with significant growth in application, ecosystem, and capital dimensions [2][21] - The weekly usage of global AI models reached 1.38 billion times, with China's MiniMax M2.5 leading at 17.3 trillion calls, showcasing the competitive landscape [21][22] Group 4: Chinese Developments - The 2026 "Two Sessions" highlighted the focus on artificial intelligence and digital economy, introducing the concept of "building a new form of intelligent economy" [32][34] - The Ministry of Industry and Information Technology emphasized the integration of AI with manufacturing and the iterative update of new intelligent terminals [33][34] Group 5: International Developments - The European Union proposed the "Industrial Accelerator Act" to enhance "EU manufacturing" requirements, aiming to increase the manufacturing sector's GDP share to 20% by 2035 [36] - The U.S. government reached an agreement with tech giants on electricity costs for AI-driven data centers, addressing public concerns about energy demand [37] Group 6: Technological Frontiers - The competition among large AI models has intensified, with OpenAI's GPT-5.4 surpassing human performance in desktop manipulation tasks, marking a significant advancement in AI capabilities [40][41] - Physical AI has emerged as a new hotspot, with companies like Jizhi Vision and Wuwen Zhike securing funding to develop world models and data infrastructure for AI applications [43]