大涨！阿里大消息：联手英伟达！

Core Insights - Alibaba Cloud announced a partnership with NVIDIA in the field of Physical AI at the 2025 Hangzhou Yunqi Conference, integrating NVIDIA's Physical AI software stack into its AI platform PAI to enhance development cycles for applications like embodied intelligence and assisted driving [1][4] - Following the announcement, Alibaba's stock surged, with a nearly 10% increase in pre-market trading in the US and a 9% rise in Hong Kong [3] Group 1: Partnership and Technology Integration - The integration of NVIDIA's software stack includes tools for data preprocessing, simulation data generation, model training evaluation, robotic reinforcement learning, and simulation testing, creating a comprehensive service platform [1][4] - Physical AI aims to extend AI technology into the physical world, enabling interaction with real-world environments, with applications in humanoid robots and advanced autonomous driving systems [4] Group 2: Future Vision and Infrastructure - Alibaba Group's CEO emphasized that the development of Artificial General Intelligence (AGI) is a certainty, with the ultimate goal being the creation of Super Artificial Intelligence (ASI) [4] - Alibaba Cloud plans to significantly increase its global data center energy consumption by 10 times by 2032 compared to 2022, indicating a substantial investment in computational power [5] Group 3: AI Model Advancements - The newly launched Qwen3-Max model outperforms competitors like GPT-5 and Claude Opus 4, ranking among the top three globally, and has become the most widely adopted model among Chinese enterprises [7] - Alibaba Cloud has seen a fivefold increase in AI computing power and over fourfold growth in AI storage over the past year, reflecting its commitment to enhancing AI infrastructure [7][8] Group 4: Hardware Innovations - Alibaba Cloud introduced the new generation of Panjiu 128 ultra-node AI servers, designed for high density and performance, capable of supporting multiple AI chips and achieving significant improvements in inference performance [10]