更多资料加入知识星球:水木调研纪要 关注公众号:水木纪要 2026 年电子行业十大预测 20260104 Q&A 2026 年电子行业的十大预测中,云端算力板块有哪些重要趋势和变化? 2026年,云端算力板块尤其是国产算力将迎来显著发展。首先,从产能角度来 看,预计到 2026年底,包括国内头部晶圆厂在内的 N2 制程将释放充沛的产能, 为国内 GPU 厂商提供足够的供给。这意味着产能瓶颈不再是国产算力发展的核心 问题。 从需求端来看,无论是字节跳动等龙头公司的投放消耗,还是国产算力 原厂商的客单价情况,都显示出强劲增长。例如,在2025年 5月,字节跳动单 日数据消耗约为 15-16 万亿,到 9 月底已达到 30 万亿,并在 12 月初超过 50 万 k 调研纪号 争 狗 - 国产 GPU 产能释放,不再是算力瓶颈,字节跳动等龙头企业数据消耗量激 . 增,预示国产算力需求端强劲增长,2026年中或突破百万亿级别,为国 产算力公司业绩兑现奠定基础。 寒武纪和海光等龙头 GPT 公司在产能供给和地方政府招标中占据优势,同 ● 时,推理芯片需求增加,ASIC 芯片有望与 GPU 并驾齐驱,新元股份和翱 捷等公司 ...

Investment Rating - The report maintains an "Overweight" rating for the electronic industry [1] Core Insights - The report highlights the competitive landscape of Switch chip manufacturers, with NVIDIA dominating the market through proprietary protocols, while Broadcom is gaining traction with its open-source SUE architecture [6][13] - It emphasizes the potential for domestic manufacturers to replace imported Switch chips, with companies like Shengke Communication leading the way in Ethernet chip production [35] - The report suggests investment in key players such as Haiguang Information and Shengke Communication, while also recommending attention to companies like ZTE and Lanke Technology [6][35] Summary by Sections 1. Switch Chip Manufacturers - NVIDIA's NVSwitch is noted as the highest bandwidth and most mature private solution in the Scale-Up segment [11] - Broadcom holds a 90% market share in cloud data center switches and has introduced the SUE architecture for open Ethernet interconnects [13] - Astera Labs is recognized for its complete product chain, being the only company that integrates switch chips, extension lines, and software management platforms [20] 2. Achieving Domestic Replacement of Switch Chips - Shengke Communication is identified as a leading domestic Ethernet switch chip manufacturer, with 12.8T and 25.6T chips entering customer promotion stages [35] - Other domestic players like Shudao Technology and Lanke Technology are making strides in the PCIe segment, with Shudao expected to achieve breakthroughs in domestic replacement by the end of 2025 [41] - Major manufacturers such as Haiguang, Huawei, and ZTE are also developing self-researched chips to support domestic replacement efforts [43][45][50] 3. Investment Recommendations - The report recommends focusing on Haiguang Information and Shengke Communication for investment opportunities, while suggesting to keep an eye on ZTE, Wantong Development, and Lanke Technology [6][35]

Summary of Conference Call Records Industry Overview - The domestic computing power sector is benefiting from policy implementation and overseas market trends, with an overall positive outlook for the second half of the year. There is potential for significant price increases from late August to the end of the year as major companies provide KPI guidance and mid-year financial reports [1][2][19]. Key Companies and Recommendations - **Cambricon (寒武纪)**: Recommended as a key player in the domestic GPU and ASIC segments. If the n+2 expansion proceeds smoothly by the second half of 2025, the expected capacity could reach over 600 billion, supporting a substantial market valuation [1][5]. - **New Yuan Co. (新元股份)**: Still presents investment opportunities in the second half of the year, particularly due to its collaboration with Alibaba and potential overperformance in film progress, which could lead to a strong stock price reaction [1][8]. - **Switch Chip Sector**: Expected to grow significantly, with domestic companies like Shenghe Communication (盛和通信) likely to see major catalysts. The market for non-NVIDIA computing chips is projected to reach approximately $3.8 billion by 2026 [1][10][11]. - **ZTE and Hua Hong Semiconductor**: Both companies have positive expectations for Q3 and Q4, with ZTE launching new technologies and products, indicating strategic importance [1][14]. Market Dynamics - The domestic computing power sector has shown significant market performance since June, driven by supply and demand changes and chip testing progress. The overall trend remains positive despite some fluctuations [2][19]. - The relationship between GPU and ASIC segments is characterized by shared growth opportunities, with both experiencing high demand and tight supply [6]. Financial Projections and Market Size - The market for PCIe switch chips is projected to grow significantly, with estimates suggesting a market size of approximately $3.8 billion by 2026, potentially doubling by 2027 as technology upgrades occur [11]. - **Yongxin Electronics (永新电子)**: Expected to benefit from advanced manufacturing and domestic GPU client integration, with projections indicating a potential market valuation of up to 400 billion if certain sales targets are met [18][21]. Investment Opportunities - The recent adjustment in the Hong Kong stock market provides a good opportunity for investors to reposition, particularly in companies like Hua Hong Semiconductor, which is expected to have significant future developments [15][16]. - The domestic advanced packaging sector, including companies like Yuxi Electronics (有锡电子) and Yongxin Electronics, is highlighted as having substantial growth potential due to their roles as suppliers to major GPU clients [20][21]. Conclusion - The domestic computing power sector is poised for growth, with several key players and segments showing promise. Investors are encouraged to focus on companies with strong strategic positions and potential catalysts for growth in the coming months [19].

Core Viewpoint - The global AI server market is expected to grow at a compound annual growth rate (CAGR) of approximately 27.2% from 2022 to 2029, with significant market size projections for 2025 and 2027 [1][4]. Market Dynamics - The shift in performance bottlenecks in the inference era is moving from "computing power" to "transport capacity," necessitating a transition from Scale-Out to Scale-Up network architectures to reduce communication latency and enhance throughput [1][2]. - The internal switch chips in AI servers can be categorized into CPU-GPU connections and GPU-GPU interconnections, which are crucial for high bandwidth and low latency communication within GPU clusters [2]. Competitive Landscape - The AI inference market is diversifying with various acceleration chips, moving away from NVIDIA's dominance, as open standards like PCIe6/7 and UALink create opportunities for third-party interconnect vendors [3]. - Astera Lab is emerging as a key player in the non-NVIDIA ecosystem, actively participating in AMD's UALink and AWS's Trainium NeuronLink, positioning itself for growth alongside AI infrastructure expansion [3]. Market Size Projections - The AI server market is projected to reach approximately 49.139 billion yuan by 2025 and expand to about 102.223 billion yuan by 2027, driven by increasing demand for GPU interconnect switches and CPU-GPU interconnect switches [4]. Investment Recommendations - Companies such as 盛科通信-U (688702.SH) and 海光信息 (688041.SH) are highlighted for their advancements in high-spec Ethernet switch chips and Chiplet interconnect technologies, respectively, with 澜起科技 (688008.SH) also recommended for its PCIe and CXL products [5].

Core Insights - The report from GF Securities highlights that the demand for single-node computing density, bandwidth, and low-latency interconnects is rising due to AI training and inference, making the enhancement of single-machine performance a priority for data center investments [1] - The rapid development of Scale up networks is directly driving the demand for Scale up switches, switch chips, interconnect chips, PCBs, and ODMs [1] - The transition from "AI's iPhone moment" to "AI's fission moment" and then to "AI's offensive moment" suggests a focus on core beneficiaries within the industry chain [1] Industry Trends - Scale up has become the core trend for computing power expansion in the AI era, with three main interconnects surrounding GPUs: business network interconnect, Scale out network interconnect, and Scale up network interconnect [1] - Scale up enhances computing density and data access efficiency, with GPU interconnects first focusing on Scale up before Scale out, indicating its importance for future AI scalability [1] Protocol Development - The Scale up protocol is diversifying, with various manufacturers launching self-developed or deeply customized switch chips and interconnect protocols, including PCIe, NVLINK, UALINK, and SUE [2] - There are two main paths for Scale up: high-density single-cabinet GPU solutions and multi-cabinet optical interconnects, with NVIDIA and Google being key representatives in these developments [2] - As GPU interconnects extend beyond single cabinets, optical interconnects are becoming essential for achieving efficient connections, benefiting upstream components like Scale up switches, switch chips, interconnect chips, PCBs, and ODMs [2]