Workflow
超节点产品
icon
Search documents
新华三图灵小镇跑出“贵安速度”:打造智算枢纽的西部新范式
Huan Qiu Wang Zi Xun· 2025-09-04 05:43
Core Viewpoint - The article highlights the strategic upgrade of Unisplendour Corporation's subsidiary, H3C Group, in response to the urgent demand for computing power infrastructure in the AI era, emphasizing the establishment of the Turing Town model as a replicable and promotable industrial model [1][11]. Group 1: AI Infrastructure Development - H3C Group emphasizes the importance of building a robust computing power infrastructure as a foundation for AI applications, likening it to "repairing the road" for a long-distance run [3]. - The company proposes a "Computing Power × Connectivity" concept to address challenges such as low resource utilization and network congestion in traditional computing clusters [3][4]. - H3C has developed leading super-node products that enhance training and inference efficiency by 25% and 62.5% respectively, supporting large model training [4]. Group 2: Turing Town Model - The Turing Town model aims to create a complete commercial closed loop, addressing the traditional challenges of high investment, operational difficulty, and low profitability in computing centers [5][6]. - H3C collaborates with local government platform companies to ensure efficient utilization of computing resources through joint operations [5]. - The model has established four capability centers, forming a comprehensive AI industrial ecosystem that covers hardware adaptation, model optimization, scenario incubation, and industry empowerment [6]. Group 3: Market Demand and Differentiation - H3C's approach is demand-driven, identifying key areas such as internet, research, and model training to ensure high utilization of computing resources [6]. - The company differentiates itself in a competitive AI server market by offering integrated solutions rather than just hardware, transforming from a product supplier to a solution participant [6][7]. Group 4: Data Value Activation - H3C introduces the "Intelligent Trusted Data Space" solution to facilitate data circulation while ensuring data security, addressing the issue of data ownership transfer [8]. - The "Data Bridge" tool allows for data analysis without transferring ownership, maintaining the data sovereignty of the provider [8]. Group 5: Future Goals and Expansion - H3C aims to expand its computing power infrastructure in Guizhou and the western region, focusing on the Turing Town as a regional intelligent computing hub [11]. - The company plans to attract AI model and algorithm application enterprises to build a full-chain industrial ecosystem and promote large-scale AI applications in various sectors [11].
国泰海通:scale up带动交换芯片新需求 国内厂商市场份额有望逐步提升
智通财经网· 2025-08-24 23:35
Group 1 - The core viewpoint is that domestic manufacturers are expected to gradually increase their market share in high-end switching chips due to continuous breakthroughs and increased overall AI spending, with projected market sizes for 2025, 2026, and 2027 being 257 billion, 356 billion, and 475 billion yuan respectively, representing year-on-year growth rates of 61%, 39%, and 33% [1] - The current overall domestic substitution rate of switching chips is low, especially in the high-end chip market, where companies like Broadcom, Marvell, and NVIDIA dominate, indicating significant room for domestic chip replacement [1] Group 2 - The evolution of large models and the expansion of Scale up clusters are identified as important trends, with large language model parameters evolving from hundreds of billions to trillions and beyond, employing various strategies to address the limitations of model size [2] - The communication requirements for tensor and expert parallelism are stringent, making high-bandwidth, low-latency Scale up networks the mainstream technical solution in the industry [2] Group 3 - The ongoing upgrade of overseas AI chips to Scale up sizes is driving new demand for switching chips, with current GPU Scale up interconnects reaching dozens of cards and evolving towards hundreds, while AI custom chip interconnects are expanding from dozens to thousands [3] - Domestic AI companies are launching their own supernode products equipped with Scale up switching nodes, with Huawei's Ascend supporting interconnects of 384 chips and Baidu's Kunlun supporting 32/64 card interconnects [3] - Various domestic manufacturers, including ZTE and H3C, are providing foundational engineering capabilities for domestic chips to transition to supernodes, with ZTE's supernode server achieving GPU communication bandwidths of 400GB/s to 1.6T/s [3] - In the Scale up switching domain, Ethernet, PCIe, and private protocols (such as NVLink and UB) are expected to coexist, while Ethernet is anticipated to dominate the Scale out domain due to its open ecosystem and cost advantages [3]