据产业链调研反馈，谷歌TPUv8p有望引入内存池化，用以拓展内存容量，该方案下，每张TPU有望额外增加数百 GB的内存拓展；内存池化有望逐步迎来规模商业化、放量可期。 【长江电子】#澜起科技第四call：在产业趋势上做配置 #rambus自身因封测品质因素指引不及预期、但不改产业趋势 rambus因自身封测品质因素影响一季度产品交付，Q1业绩指引不及预期至盘后大跌。 但公司#对服务器CPU内存条未来增长指引、MRDIMM等产品放量均给予正面回应、产业趋势持续向上。 #谷歌TPUv8p或引入内存池化、CXL放量可期 据产业链调 【长江电子】#澜起科技第四call：在产业趋势上做配置 #rambus自身因封测品质因素指引不及预期、但不改产业趋势 rambus因自身封测品质因素影响一季度产品交付，Q1业绩指引不及预期至盘后大跌。 但公司#对服务器CPU内存条未来增长指引、MRDIMM等产品放量均给予正面回应、产业趋势持续向上。 #谷歌TPUv8p或引入内存池化、CXL放量可期 伴随着CXL拓展芯片、MRCD/MDB、PCIe Retimer、PCIe Switch的规模起量；我们认为公司远期利润有望达100- 15 ...

Summary of Conference Call Records Companies and Industries Involved - **Companies**: Haiguang Information, Lianqi Technology, Wangsu Technology - **Industries**: AI computing, CDN (Content Delivery Network), semiconductor technology Key Points and Arguments Haiguang Information - Haiguang Information's market capitalization increased by over 90 billion RMB, leading the A-share market in January 2026 [2] - The company’s Deep Computing 3 has entered mass production, supporting FP8/FP4 precision, while Deep Computing 4 is expected to double performance, potentially becoming the strongest AI chip in China [3][7] - The estimated valuation for Haiguang's CPU is 900 billion RMB and for its GPU is 1.3 trillion RMB [3][7] - The company is projected to reach a market capitalization of over 2 trillion RMB by 2028, with a target of 1.2 trillion RMB for 2026 [8] Lianqi Technology - Lianqi Technology benefits from the growth in AI inference and supernode industries, particularly in memory interconnect chips, PCIe Retimer Switch, and CXL chips [1][2] - The company has made significant progress in the CXL field, with its products expected to be adopted by Google's next-generation TPU, creating a substantial incremental market [10] - Lianqi's revenue breakdown includes 90% from memory interconnect, 5% from PCIe CXL, and 5% from CPU and server-related products [10] Wangsu Technology - Wangsu Technology is the largest third-party neutral CDN company in China, with CDN business accounting for 60-70% of its revenue [11] - The company is benefiting from a near doubling of CDN prices in North America due to Google Cloud's price increase, indicating a reversal in the CDN and cloud computing price war [1][2][12] - Wangsu is expected to achieve a net profit of 1 billion RMB in 2026, with significant profit elasticity due to price increases, suggesting over 50% growth potential in its valuation [12] Capital Expenditure Trends - North America's top five CSPs are projected to have capital expenditures nearing 700 billion USD in 2026, a 50% increase from 400 billion USD in 2025, driven by Meta and Microsoft's unexpected capital spending [4] - Domestic internet capital expenditure in China is expected to reach 570-600 billion RMB in 2026, with growth anticipated to surpass that of overseas markets by 2027 due to advancements in self-developed chips and easing of restrictions [4] AI Inference Demand - The emergence of applications like MudBot is driving exponential growth in data and computing power consumption, shifting traffic from human-driven to robot-driven, enabling 24/7 usage [5] Supply-Side Technological Advances - Future server architectures are expected to adopt supernode technology, which will enhance cluster efficiency through memory pooling and high-speed interconnects [6] Other Notable Companies - Additional companies to watch include DingTong Technology, Zhongke Shuguang, Shuguang Shuchuang, Feirongda, Yingweike, and application vendors like Shuiyou Co. and Keda Xunfei, all of which show promising development prospects [13]

Group 1 - Wells Fargo indicates that Marvell Technology (MRVL.US) plans to acquire XConn for $540 million, which is deemed crucial for memory pooling and expected to enhance company earnings in the near term [1] - Analyst Aaron Rakes emphasizes that the acquisition further validates the importance of memory pooling technology in high-performance and competitive hardware solutions, particularly in supporting larger models and improving inference performance [1] - The proposed transaction, with 60% cash and 40% stock payment, is expected to contribute to revenue starting from the second half of the current fiscal year, potentially reaching $100 million by fiscal year 2028 [1] Group 2 - In the AI 2.0 era, the core contradiction in computing power development has shifted from merely "not fast enough" to "data handling cannot keep up" [2] - The emergence of CXL (Compute Express Link) technology represents a significant transformation of traditional computing models at the physical architecture level, enhancing AI computing power through memory decoupling, capacity expansion, and communication collaboration [2] - CXL technology is not just about increasing bandwidth; it reconstructs fragmented data centers into a cohesive working whole through resource pooling, capacity decoupling, and consistent communication, serving as a foundational technology for the transition from "single performance competition" to "cluster efficiency game" in AI computing [2]

Core Viewpoint - The article emphasizes the significance of the latest CXL 4.0 specification in enhancing memory connectivity and performance for high-performance computing, particularly in artificial intelligence applications [2][13]. Group 1: CXL 4.0 Specification Features - CXL 4.0 doubles the bandwidth to 128GTs without additional latency, enhancing data transfer speeds between connected devices [4][11]. - It supports high-speed data transfer between CXL devices, improving overall system performance [7]. - The specification retains full backward compatibility with CXL 3.x, 2.0, 1.1, and 1.0 versions, ensuring a smoother transition for existing deployments [12]. Group 2: Importance of CXL for AI - CXL addresses memory bottlenecks in AI workloads by enabling memory pooling, allowing all processors to access a unified shared memory space, thus improving memory utilization [15][17]. - It facilitates large-scale inference by providing quick access to large datasets without the need for memory duplication across GPUs [18]. - CXL is designed to meet the growing performance and scalability demands of modern workloads, particularly in AI and high-performance computing [19]. Group 3: Future Implications of CXL - The introduction of CXL is seen as a fundamental shift from static, isolated architectures to flexible, network-based computing, paving the way for next-generation AI and data-intensive systems [20]. - CXL enables a unified, flexible AI architecture across server racks, crucial for training large language models efficiently [21]. - Major industry players, including Intel, AMD, and Samsung, are beginning to pilot CXL deployments, indicating its growing importance in the semiconductor landscape [21].

Core Insights - Intel and Alibaba Cloud announced a series of cloud instances and storage solutions based on the new generation Xeon® 6 processors, addressing the challenges posed by AI scalability on cloud infrastructure [1][9] - High performance, high elasticity, and low total cost of ownership (TCO) are becoming key competitive indicators for global cloud providers [1] Group 1: Cloud Infrastructure Innovations - The introduction of "memory pooling" and flexible architecture is transforming cloud infrastructure, allowing dynamic allocation of resources based on demand [2][6] - Alibaba Cloud has deployed a unified hardware architecture across 29 global regions and 91 availability zones, enabling rapid resource allocation in response to sudden computing demands [4][9] Group 2: AI and Heterogeneous Computing - AI-driven heterogeneous computing is redefining the role of CPUs as central coordinators, with Intel integrating AMX matrix acceleration instruction sets to support various precision calculations [7] - The Xeon® 6 processors can efficiently handle large AI models, demonstrating significant performance improvements in various applications, such as data preprocessing for autonomous driving [7][8] Group 3: Collaboration and Competitive Edge - The stability and engineering support of the collaboration between Intel and Alibaba Cloud are highlighted as foundational elements for their long-term partnership [8] - The optimization of both hardware and software is becoming a key differentiator in the market, with Alibaba Cloud leveraging CXL 2.0 memory pooling technology for enhanced performance [8][9] Group 4: Future Directions - The shift from cloud adoption to intelligent cloud solutions is seen as an inevitable development path, with AI moving into a phase of large-scale application [9][10] - The collaboration between Intel and Alibaba Cloud aims to provide scalable and sustainable pathways for various industries through enhanced hardware performance and optimized software stacks [9][10]

Core Viewpoint - The launch of the CXL3.1 memory expansion controller (M88MX6852) by 澜起科技 marks a significant advancement in memory architecture, aimed at enhancing bandwidth and reducing latency for next-generation data center servers [1][2]. Group 1: Product Features - The M88MX6852 chip supports CXL.mem and CXL.io protocols, providing high bandwidth and low latency memory expansion and pooling solutions [1]. - It utilizes a PCIe 6.2 physical layer interface with a maximum transmission rate of 64 GT/s (x8 channels) and features dual-channel DDR5 memory controller supporting speeds up to 8000 MT/s [1]. - The chip integrates dual RISC-V microprocessors for dynamic resource configuration and hardware-level security management, along with multiple interfaces for system integration [1]. Group 2: Market Demand and Applications - The demand for cloud computing resource pooling is increasing, making traditional memory architectures a performance bottleneck [2]. - The CXL3.1 memory expansion controller enables elastic allocation and efficient utilization of memory resources, thereby reducing total cost of ownership (TCO) [2]. - The chip is compatible with EDSFF (E3.S) and PCIe add-in card (AIC) formats, making it suitable for various deployment environments including servers and edge computing [2]. Group 3: Industry Feedback - Stephen Tai, the company president, highlighted that the chip represents a breakthrough in CXL technology, enhancing memory expansion performance and energy efficiency [2]. - Feedback from industry leaders like Samsung and AMD indicates strong support for the CXL3.1 controller, emphasizing its role in advancing memory decoupling architecture and reducing TCO in data centers [2][3].