Core Insights - Huawei has planned three series of Ascend chips: 950, 960, and 970 series [1] - The Ascend 950 series includes two chips: 950PR, which will be launched in Q1 2026, and 950DT, expected in Q4 2026 [1] - The Ascend 960 chip is set to be released in Q4 2027, while the Ascend 970 chip is anticipated for Q4 2028 [1] - Since the release of the Ascend 310 chip in 2018 and the Ascend 910 chip in 2019, Huawei has continuously invested in AI computing power research and innovation [1] - The company aims to establish a solid foundation for AI computing power in China and globally through the evolution of Ascend chips [1]

Core Viewpoint - Huawei has launched new supernode products, significantly enhancing computing power and interconnect bandwidth, positioning itself as a leader in the AI chip industry [6][7][8]. Group 1: Huawei's New Products - The Atlas 950 supernode, based on the Ascend 950DT chip, supports 8192 Ascend 950DT chips, achieving a total computing power of 8E FLOPS for FP8 and 16E FLOPS for FP4, with an interconnect bandwidth of 16PB/s [7]. - The Atlas 960 supernode, based on the Ascend 960 chip, can support up to 15488 cards, with a total computing power of 30E FLOPS for FP8 and 60E FLOPS for FP4, and an interconnect bandwidth of 34PB/s [8]. - The Atlas 950 supernode is set to launch in Q4 2026, while the Atlas 960 is expected in Q4 2027, both significantly outperforming competitors like NVIDIA's upcoming products [7][8]. Group 2: Market Demand for Optical Modules - The demand for optical modules is projected to increase, with estimates for 2026 indicating a need for 3000-3200 million units, driven by major companies like Microsoft and NVIDIA [12]. - The 800G optical module market is expected to exceed expectations, particularly due to Microsoft's procurement strategies [12]. - The ratio of GPUs to optical modules varies by company, with NVIDIA at 1:3-1:4.5 and Google at approximately 1:14, indicating a growing need for optical modules in the industry [17]. Group 3: Key Suppliers and Market Dynamics - Major suppliers for optical modules include companies like 旭创 (Acacia), 菲尼萨 (Finisar), and 新易盛 (NewEase), with varying market shares across different clients [18]. - For 2026, the optimistic demand for 800G and 1.6T optical modules could reach nearly 50 million units, highlighting a potential supply gap [16]. - The competitive landscape shows that 旭创 is a dominant supplier for Google, while 新易盛 holds significant shares with AWS [18].

Core Insights - Huawei's Vice Chairman Xu Zhijun unveiled the company's AI computing power roadmap at the HC 2025 conference, detailing the evolution of the Ascend chip series over the next three years, with a focus on doubling computing power annually [3][4] Group 1: AI Computing Power Development - The Ascend 950 series will launch in Q1 2026, followed by the Ascend 960 in 2027 and the Ascend 970 in 2028, with significant advancements in low-precision data formats, vector computing power, interconnect bandwidth, and self-developed HBM [3][4] - The Atlas 950 and Atlas 960 supernode products were introduced, with the Atlas 950 featuring 8192 cards and an FP8 computing power of 8E FLOPS, set to launch in Q4 2026, significantly surpassing NVIDIA's concurrent offerings [3][4] Group 2: Expansion into General Computing - Huawei is extending its supernode concept to general computing with the TaiShan 950 supernode, which aims to outperform traditional databases and high-end servers by achieving a 2.9x performance boost without requiring database modifications [4] - The self-developed "UnifiedBus" interconnect protocol will be open-sourced from version 2.0, promoting a truly open supernode ecosystem [4] Group 3: Market Reaction and Collaborations - Following the announcements, Huawei's computing power concept stocks surged, with companies like Fenghuo Communication hitting the daily limit, and others like Xingtuxinke and Xinghuan Technology seeing significant gains [6][9] - Companies in the Huawei ecosystem, such as Fenghuo Communication and Xingtuxinke, are actively collaborating with Huawei on various AI solutions and have received certifications for their compatibility with Huawei's Ascend platform [9] Group 4: Industry Outlook - Analysts from Zhongyin Securities noted a rising industry sentiment, with domestic computing power expected to see significant growth, driven by high utilization rates in wafer manufacturing and semiconductor equipment orders [10]

Core Viewpoint - The article discusses advancements in lithography technology, particularly the development of "Beyond-EUV" (B-EUV) lithography, which utilizes a wavelength of 6.5nm to 6.7nm, potentially allowing for resolutions below 5nm, surpassing current EUV technology [2][3][18]. Summary by Sections Lithography Technology Evolution - The evolution of lithography has progressed from contact to projection methods, with current mainstream technology using extreme ultraviolet (EUV) light at a wavelength of 13.5nm [3][5]. - The need for higher resolution in lithography can be achieved by either increasing numerical aperture (NA) or shortening the wavelength [3]. Challenges of EUV and B-EUV - EUV technology faces challenges due to its high absorption rates by materials, necessitating advanced mirrors for effective light reflection [5][6]. - B-EUV technology is still in its infancy, with no industry-standard methods for generating the required 6.7nm wavelength radiation [6][19]. Innovations in Light Sources - Various companies are exploring new light sources for lithography, including the development of high-efficiency laser systems and compact, high-power sources [10][11][13]. - Inversion is working on Laser Wakefield Acceleration (LWFA) to create high-energy light sources, while xLight is developing free electron lasers (FEL) to enhance EUV power output significantly [13][16]. Breakthroughs in Photoresist Materials - Johns Hopkins University has made significant progress in photoresist materials, discovering that metals like zinc can effectively absorb B-EUV light and trigger chemical reactions for fine pattern etching [18][19]. - The development of a new technique called Chemical Liquid Deposition (CLD) allows for the creation of thin films that can be used in semiconductor manufacturing, highlighting the importance of matching materials with the appropriate wavelengths [19][20]. Future Prospects - The advancements in B-EUV technology and associated materials could lead to significant improvements in semiconductor manufacturing efficiency and cost reduction, although challenges remain before widespread adoption [20].

Core Viewpoint - Computing power is crucial for artificial intelligence and will continue to be a key factor for China's AI development, as stated by Huawei's rotating chairman Xu Zhijun [3]. Group 1: Huawei's Chip Development - Huawei plans to launch the Ascend 950PR chip in Q1 2026, the Ascend 950DT in Q4 2026, the Ascend 960 chip in Q4 2027, and the Ascend 970 chip in Q4 2028 [3]. - The company has introduced the Atlas 900 A3 SuperPoD, with over 300 units deployed, serving more than 20 customers, and will release the Atlas 950 SuperPoD in Q4 2026, supporting 8192 Ascend cards with a computing power of 8 EFLOPS FP8 / 16 EFLOPS FP4 [5]. Group 2: Supernode Infrastructure - Huawei is developing the Atlas 960 SuperPoD, expected to support 15488 Ascend cards with a computing power of 30 EFLOPS FP8 / 60 EFLOPS FP4, planned for Q4 2027 [7]. - The company announced the Atlas 950 SuperCluster and Atlas 960 SuperCluster, with computing power exceeding 500,000 cards and reaching one million cards, respectively [9]. Group 3: General Computing and Open Technology - Huawei revealed the roadmap for the Kunpeng chip, with the Kunpeng 950 chip set to launch in Q4 2026 and the Kunpeng 960 chip in Q1 2028 [9]. - The company introduced the TaiShan950 SuperPoD, the world's first general computing supernode, expected to launch in Q1 2026, which aims to replace various large and small machines as well as Exadata database integrated machines [9]. - Xu Zhijun announced that the supernode technology will be fully open, including the Lingqu 2.0 technology specifications, which encompass foundational, firmware, and operating system reference design specifications [9].

Core Viewpoint - The company has successfully developed 10 microwave MMIC products, including power amplifiers, low-noise amplifiers, and digitally controlled phase shifters, which are applicable in communication and radar scenarios [1] Group 1: Product Development - The company has completed performance verification for its microwave MMIC products [1] - Microwave components such as S-band and Ku-band TR components are currently in the research and development phase [1] - The company is actively developing new products in clock management circuits and buffers, demonstrating relevant technical capabilities [1] Group 2: Commercialization and Production - The company will strictly follow R&D progress, market demand, and industry certification requirements for large-scale production and commercialization [1]

格隆汇9月18日丨鼎信通讯(603421.SH)在投资者互动平台表示，公司芯片开发采用平头哥的CK802 32位 内核架构，平头哥与公司签订有全面技术授权协议。 ...

Core Insights - Oracle projected its cloud infrastructure revenue to reach $144 billion over the next five years, supported by non-cancellable contracts [1] - Despite Oracle's optimistic outlook, it faces challenges such as a heavy debt load and significant capital expenditures required to realize this revenue [2] - Major cloud providers like Amazon, Microsoft, and Alphabet have opted not to pursue this opportunity, allowing Oracle to build data centers for them instead [3] Company Analysis Oracle - Oracle's stock surged following the revenue projection, but the company is not expected to be the primary beneficiary of its own cloud computing push due to financial constraints [2] - The uncertainty surrounding the economics of the projected revenue raises questions about Oracle's ability to capitalize on this growth [3] Nvidia - Nvidia holds a dominant position in AI infrastructure, with a 94% market share in the GPU market and a 56% increase in data center revenue to $41.1 billion [6] - The company has established a competitive advantage through its CUDA software platform and NVLink interconnect system, which enhances the performance of its GPUs [7][8] - Nvidia is expected to benefit significantly from Oracle's data center projects due to its close relationship with the company [9] Broadcom - Broadcom is positioned to benefit from Oracle's spending as customers seek to diversify their AI processing capabilities beyond Nvidia [10] - The company has been instrumental in developing custom chips for AI workloads, with a serviceable market opportunity estimated between $60 billion to $90 billion by fiscal 2027 [12] - OpenAI, a key customer for Oracle, is anticipated to utilize Broadcom's custom AI chips alongside Nvidia's GPUs, presenting a substantial growth opportunity for Broadcom [13]

Core Insights - The semiconductor sector is experiencing significant momentum, with the Shanghai Composite Index targeting 3900 points and the ChiNext Index rebounding strongly, driven by key players like Cambrian, North Huachuang, SMIC, and Haiguang Information reaching new highs [1] Group 1: Market Performance - The semiconductor ETF surged by 5.9%, while the largest chip ETF increased by 4.48%, reflecting strong investor interest in the sector [1] - The market saw a net inflow of 136 million in the Sci-Tech Semiconductor ETF, which rose by 5.9% after a previous increase of 3.6% [2] Group 2: Industry Catalysts - Recent developments include Alibaba and Baidu utilizing self-developed chips for AI model training, which has heightened market attention on the chip sector [1] - Alibaba's self-developed AI chip was featured on CCTV's "News Broadcast," reportedly matching key parameters of NVIDIA's H20 chip and surpassing the A800 chip [1] - Tencent announced the full opening of its AI capabilities, adapting to mainstream domestic chips [1] - Huawei's rotating chairman emphasized the importance of computing power for AI and revealed future plans for its Ascend chips [1] - The Federal Reserve's recent 25 basis point rate cut and China's strong performance in AI technology and biomedicine have attracted renewed interest from overseas investors [1] Group 3: Notable Products - The largest chip industry ETF, covering the entire semiconductor value chain, includes leading companies such as SMIC, Cambrian, Haiguang Information, North Huachuang, and Shengbang [1] - Key components of the Sci-Tech Semiconductor ETF include companies specializing in etching equipment, thin film deposition equipment, CMP equipment, 300mm silicon wafers, and silicon carbide substrates [2]

【环球网科技报道 记者 张阳】9月18日，华为在上海世博展览馆召开全联接大会（HC）。华为轮值董事长徐直军发表了题为《以开创的超节点互联技术引 领AI基础设施新方式》的主题演讲。在演讲中，徐直军首次公开了华为昇腾芯片演进路线图，以及面向超节点的互联协议灵衢架构，并将灵衢2.0技术规范 对外开放。 徐直军表示，自2019年发布昇腾910芯片，到2025年，基于昇腾910C芯片打造的Atlas900超节点的规模部署，业界对昇腾芯片就一直有很多诉求和期待。此 次，就正式公布昇腾芯片演讲路线图。 大规模超节点，把计算和通用计算的能力推向了新的高度，但是，大规模超节点带来的技术挑战也显而易见：如何满足长距离且高可靠的连接，并且还要能 实现大带宽且低时延。 徐直军表示，为了解决长距离铁高可靠性的问题，我们在互联协议的物理层、数据链路层、网络层、传输层等每一层都引入了高可靠机制。同时，我们在光 路上引入了纳秒级的故障检测和保护功能，当出现光模块闪断和故障时，让应用感知，并重新定义和设计了光器件、光模块以及互联芯片，这些创新设计让 光互联的可靠性提升了100倍，满足了高可靠全光互联、高带宽、低时延的要求，让大规模超节点成为可 ...