Workflow
对话徐直军:华为整个计算战略在五个字上
是说芯语·2025-09-20 07:13

Core Viewpoint - Huawei's advancements in AI and computing capabilities, particularly through its Ascend chip series and supernode technology, position it as a strong competitor in the global AI infrastructure landscape, despite facing challenges from U.S. sanctions and competition from companies like NVIDIA [2][3][4]. Group 1: AI and Computing Strategy - Huawei's core strategy revolves around "supernodes + clusters," which aims to provide continuous AI computing power despite limitations in chip manufacturing technology [4][8]. - The company has developed the CloudMatrix 384 supernode, which is the largest commercial supernode in China and boasts the highest effective computing power globally [8][9]. - Huawei plans to release several new AI chips and supernodes over the next few years, with a roadmap that includes the Ascend 950PR, 950DT, 960, and 970 series, aiming for a doubling of computing power approximately every year [12][13]. Group 2: Competitive Positioning - Huawei's Ascend 950 supernode is projected to maintain its status as the world's strongest computing supernode for several years, significantly outperforming NVIDIA's upcoming products in various metrics [13][15]. - The company emphasizes that its supernode architecture is designed to overcome traditional computing limitations, moving away from the von Neumann architecture to a self-defined "equal interconnection architecture" [9][16]. - Huawei's innovations are seen as a response to U.S. sanctions, pushing the company to explore alternative paths for technological advancement [7][8]. Group 3: Ecosystem and Market Dynamics - Huawei is focused on building a robust ecosystem around its hardware, with an emphasis on open-source software to attract developers and enhance hardware sales [17][18]. - The company does not rely on the CUDA ecosystem, instead developing its own MindSpore framework to ensure independence from Western supply chains [18]. - Huawei's long-term vision includes not just chip development but also providing comprehensive system integration capabilities to foster healthy collaborations within the AI industry [18][19].