目前， scaleX万卡超集群已完成400多个主流大模型、世界模型的适配优化，依托国家超算 互联网，更可接入上千款应用，链接更多AI产业生态伙伴， 为 万亿参数模型训练、高通量推 理、 AI for Science等大规模AI计算场景 ，提供高效普惠的算力支持。 根据业内报道， scaleX万卡超集群在去年12月HAIC大会首次真机亮相，2个月后即以超3万 卡的AI算力规模落地。 有观点指出， 这标志着国产万卡集群已正式迈入规模化部署与实战应 用的新阶段 。随着 "中国算网"全面铺开，国产集群有望迎来更广阔的市场空间。 数据显示，截至 2025年底，国家超算互联网平台已服务100多万用户，应用商品超7200多 个，单日处理作业峰值103万个，迄今已累计支撑运行1.96亿次作业。据悉，这种"算力+应 用"一体化服务模式，大幅降低了算力成本和使用门槛，可以有效赋能AI创新提速。 其中，中科曙光的 scaleX万卡超集群系统为此次核心节点提供了关键算力底座。 据介绍， scaleX万卡超集群 基于 AI计算开放架构 设计 ，可全面兼容 CUDA等主流软件生 态，支持多品牌国产加速卡混合部署， 不仅 降低 了 开发者 ...

Group 1 - The China Securities Information Technology Application Innovation Industry Index decreased by 0.05% as of February 6, 2026, with mixed performance among constituent stocks [1] - Geer Software led the gains with an increase of 8.45%, while ZTE Information fell by 2.81%, indicating volatility in the sector [1] - The Xinchuang ETF (562570) experienced a decline of 0.27%, with the latest price at 1.47 yuan [1] Group 2 - The Xinchuang ETF saw a significant growth in scale, increasing by 24.46 million yuan over the past week, leading its peers [2] - The ETF has recorded continuous net inflows over the past four days, with a peak single-day net inflow of 30.01 million yuan, totaling 52.30 million yuan and an average daily net inflow of 13.08 million yuan [1] - Financial analysts believe that the rapid iteration of domestic models is capturing user AI interaction points, which is expected to accelerate the demand for reasoning-side computing power [2] Group 3 - 2026 is projected to be a pivotal year for the deployment of domestic super nodes, with numerous domestic manufacturers releasing next-generation super node solutions [2] - The Xinchuang ETF closely tracks the China Securities Xinchuang Index, emphasizing domestic full-stack replacement and integrating key components such as storage chips, CPUs, and AIPC [2] - The ETF is currently the largest in its category, reflecting strong policy certainty and a comprehensive approach from hardware to industry intelligence [2]

Group 1 - The core viewpoint of the news highlights the mixed performance of the AI sector, with specific stocks like Aobo Zhongguang and Xingchen Technology showing gains, while Kunlun Wanwei led the declines [1] - The AI Artificial Intelligence ETF (512930) is currently priced at 2.2 yuan, with a turnover rate of 3.3% and a transaction volume of 1.09 billion yuan during the trading session [1] - Over the past month, the average daily transaction volume of the AI Artificial Intelligence ETF reached 2.52 billion yuan, and its latest scale is 33.63 billion yuan [1] Group 2 - According to Zhongyin Securities, the uncertainty in AI application business models and hardware demand is a necessary phase for industry deepening rather than a trend ending, indicating that short-term corrections do not alter the long-term upward logic of the industry [2] - The report suggests that vertical application vendors are likely to build new barriers in the AI era by leveraging their industry experience and data advantages, with hardware demand expected to grow alongside software maturity [2] - The AI Artificial Intelligence ETF closely tracks the Zhongzheng Artificial Intelligence Theme Index, which selects 50 listed companies involved in providing basic resources, technology, and application support for AI [2] Group 3 - As of January 30, 2026, the top ten weighted stocks in the Zhongzheng Artificial Intelligence Theme Index accounted for 57.27% of the index, including companies like Zhongji Xuchuang and Hikvision [3] - The AI Artificial Intelligence ETF has several off-market connections, including Ping An Zhongzheng Artificial Intelligence Theme ETF Initiated Link A, C, and E [3]

Group 1 - The core point of the article is the launch of the scaleX 30,000-card supercluster system by Zhongke Shuguang, marking the largest domestic AI computing power pool in China, capable of supporting trillion-parameter model training and high-throughput inference [1][3] - The scaleX supercluster represents a significant advancement in the deployment and practical application of domestic AI computing infrastructure, having been showcased at the HAIC conference and operational within two months [3] - Zhongke Shuguang has overcome key technical challenges such as high-speed interconnection, tight coupling of storage and computing, and efficient resource scheduling to enable the scaleX supercluster [3] Group 2 - The AI industry is experiencing explosive growth in global intelligent computing demand, necessitating infrastructure that is ecologically compatible and flexible to support diverse applications [3] - The scaleX supercluster is designed to be compatible with mainstream software ecosystems like CUDA, allowing for mixed deployment of various domestic acceleration cards, thus lowering the adaptation barrier for developers [3] - The supercluster has completed optimization for over 400 mainstream large models and can connect to thousands of applications, fostering a collaborative innovation ecosystem in the AI industry [4] Group 3 - The scaleX supercluster provides deep empowerment across multiple scenarios, including support for trillion-parameter model training and fault recovery, and has enhanced inference efficiency for major internet users [5] - In the AI for Science domain, the supercluster has supported a domestic materials research model that achieved top rankings internationally, significantly improving protein research efficiency by 3 to 6 orders of magnitude [5] - Zhongke Shuguang aims to deepen technological research and application practices, promoting domestic intelligent computing to serve various sectors of economic and social development more efficiently and inclusively [5]

Core Viewpoint - The AI sector is experiencing a short-term pullback, but the long-term growth trajectory remains intact, with opportunities for investment in AI applications, cloud services, and storage as the industry matures [2] Group 1: Market Performance - As of February 6, 2026, the CSI Artificial Intelligence Theme Index (930713) shows mixed performance among its constituent stocks, with Zhongke Shuguang leading at a 0.41% increase, while Xinyi Sheng fell by 4.95% [1] - The AI Artificial Intelligence ETF (512930) is currently priced at 2.17 yuan, with a total scale of 3.363 billion yuan [1] - Over the past five trading days, the AI ETF has seen net inflows on three occasions, totaling 158 million yuan, averaging daily net inflows of 31.606 million yuan [1] Group 2: Industry Insights - According to Zhongyin Securities, the uncertainty surrounding AI application business models and hardware demand is a necessary phase in the deep integration of AI technology into industries, rather than a fundamental threat to the industry's prospects [2] - Vertical application vendors are expected to leverage their industry experience and data advantages to build new barriers in the AI era, with hardware demand anticipated to grow alongside software maturity [2] - The CSI Artificial Intelligence Theme Index includes 50 listed companies that provide essential resources, technologies, and application support for AI, reflecting the overall performance of AI-related securities [2] Group 3: Key Holdings - As of January 30, 2026, the top ten weighted stocks in the CSI Artificial Intelligence Theme Index are: Zhongji Xuchuang, Xinyi Sheng, Hanwha Technology, Lanke Technology, Keda Xunfei, Zhongke Shuguang, Hikvision, Haowei Group, Chipone, and Kingsoft, collectively accounting for 57.27% of the index [2]

Group 1 - The Digital Economy ETF (工银561220) opened at 1.480 yuan, experiencing a decline of 0.47% [1] - Major holdings in the ETF include companies like SMIC, which fell by 1.48%, and Haiguang Information, which dropped by 1.00% [1] - The ETF's performance benchmark is the China Securities Index for State-Owned Enterprises in the Digital Economy, managed by ICBC Credit Suisse Asset Management [1] Group 2 - Since its inception on May 21, 2025, the ETF has achieved a return of 48.72%, with a monthly return of 0.79% [1]

日前，国家超算互联网核心节点在郑州正式上线试运行，这一全国一体化算力网络的重大基础设施，凭 借3套中科曙光scaleX万卡超集群系统的同步部署，一举成为全国首个实现3万卡规模部署、且可实际投 入运营的最大国产AI算力池。 作为国家部委、河南省、郑州市多方协同发力、重点部署的战略工程，该核心节点的落地承载着打破算 力壁垒、优化资源配置的重要使命，其战略价值不言而喻。此次开创性完成国产万卡超集群的规模化部 署，不仅创下多项业界纪录，更打造了一个可复制、可验证、可普及的国家级算力建设样板间，为全国 一体化算力网络发展提供了宝贵的实践参考。 中科曙光scaleX万卡超集群的三大核心优势，构筑起这一国产算力底座的硬核实力，全方位领跑业界发 展。在集成度上，该集群采用全球首创的高密度单机柜设计，融合超高密度刀片、浸没相变液冷等前沿 技术，将单机柜算力密度提升20倍之多，同时将电能利用效率（PUE）控制在1.04的超低水平，实现了 算力与节能的双向突破，呼应了"双碳"目标下算力基础设施绿色发展的要求。 众所周知，算力作为数字经济时代的核心生产力，堪比农业时代的水利、工业时代的电力，而我国算力 网络建设的核心目标，正是打破算 ...

Core Insights - The evolution of AI large models from single-modal to multi-modal is driving an exponential increase in computing power demand, with supernodes becoming a key foundation to overcome computing bottlenecks [1] - Huawei aims to provide scalable and effective computing power for the global AI industry through innovations in "clusters + supernodes," marking supernode deployment as a significant focus for its AI infrastructure strategy by 2026 [1] - Huawei's self-developed AI computing cluster, Atlas900 supernode, has been deployed over 300 sets since its launch, serving more than 20 clients across various industries, indicating a new phase of large-scale application [1] - The demand for computing power and communication efficiency is growing exponentially as model parameters evolve from hundreds of billions to trillions, making supernodes essential for supporting complex computing needs [1] Industry Developments - Supernodes are seen as the ecological foundation for transitioning computing power from "usable" to "usable effectively," pushing the industry from "single-point performance competition" to "system-level efficiency competition" [2] - Major Chinese tech giants are rapidly occupying the supernode landscape, with companies like Alibaba Cloud and Tencent Cloud releasing their own supernode solutions, while infrastructure firms focus on hardware breakthroughs [3] - Huawei's supernode technology has achieved breakthroughs in three key areas: ultra-large bandwidth, ultra-low latency, and unified memory addressing, enabling "computation without waiting for communication" and eliminating traditional cluster data exchange issues [3] - The supernode architecture is becoming the mainstream form of future intelligent computing centers, with leading domestic companies establishing a clear first-mover advantage in this field [3] Collaborative Efforts - Wuhan Yangtze Computing Technology Co., a subsidiary of FiberHome Technologies, has partnered with Ascend to release the Ascend Intelligent Manufacturing series solutions, while StarRing Information Technology has announced collaboration with Huawei in intelligent computing products [4] - The development of supernode technology is recognized as a crucial direction for AI advancement, with two main paths emerging: single vendor solutions for extreme computing efficiency and collaborative ecosystems to leverage multiple capabilities [4] - Domestic leading enterprises' focus on supernode technology is gradually altering the international computing power competition landscape, emphasizing the importance of ecosystem collaboration for optimizing computing efficiency and fostering sustainable innovation [4]

2月5日，国家超算互联网核心节点在河南郑州上线试运行。该节点是国家超算互联网平台上线以来接入 的最大单体国产AI算力资源池，可提供超过3万卡的国产AI算力，满足大规模AI计算需求。 据介绍，该节点的算力资源由中科曙光scaleX万卡超集群系统支撑，基于AI计算开放架构，兼容主流软 件生态，支持多品牌国产加速卡混合部署，并已完成400多个大模型的适配优化。在国家超算互联网平 台一体化算力调度下，可接入上千款应用。 截至2025年底，国家超算互联网平台已服务100多万用户，上架超7300个应用商城商品。 来源：新华社 作者: 温竞华 杨琳 ...

麦格米特近日发布的公告显示，截至1月26日，葛卫东关联人葛贵莲从该公司前十大股东名单中消失，结合过往公告看，葛贵莲在今年1月至少减持76万 股。 从麦格米特股价表现看，2024年初股价（前复权价）最低不到18元，近期股价则一度超过140元，区间最高涨幅超六倍。从该公司定期报告看，葛卫东关 联人李萍和葛贵莲分别于2025年二季度和三季度在该公司前十大股东名单中现身，浮盈丰厚。 "葛卫东"减持 2月5日，麦格米特发布了向特定对象发行股票之上市公告书，公司最新前十大股东情况浮出水面。截至1月26日，葛卫东关联人王萍持有麦格米特1148.92 万股，和去年底相比持股数量不变。但是葛卫东另一位关联人葛贵莲则退出前十大股东之列。 记者梳理发现，截至2025年底，葛贵莲持有麦格米特606.76万股。截至2026年1月26日，麦格米特第十大股东南方中证500ETF持股数量为529.98万股，这 意味着，开年以来，葛贵莲至少减持麦格米特76万股。 | 序号 | 股东名称或姓名 | 股份数量(股) | 持股比例(%) | 持有有限售条件的股份数量(股) | | --- | --- | --- | --- | --- | | 1 ...