以下文章来源于芯东西 ，作者程 茜 芯东西1月7日消息， 今日凌晨，美国无线、数据处理和存储芯片制造商 Marvell 宣布，将斥 资 5.4亿美元（约合人民币37.71亿元） 收购提供先进PCIe和CXL交换芯片的供应商XConn Technologiesm，该交易将于2026年初完成。 | MARVELL | | PRODUCTS 产品 | COMPANY 公司 | SUPPORT 支持 | | છ | ( | | --- | --- | --- | --- | --- | --- | --- | --- | | OVERVIEW | NEWS & | COMPANY INFORMATION 公 | FINANCIAL INFORMATION 财 | STOCK DATA 股 | SEC FILINGS SEC | GOVERNANCE | | | 概述 | EVENTS | 司信息 | 务信息 | 票数据 | 文件 | 治理 | | | | 新闻与活动 | | | | | | | 芯东西 . 芯东西专注报道芯片、半导体产业创新，尤其是以芯片设计创新引领的计算新革命和国产替代浪潮；我 们是一群追"芯"人， ...

Marvell Technology has agreed the acquisition of XConn Technologies for approximately $540m, with the transaction comprising approximately 60% cash and 40% stock. The acquisition is set to strengthen Marvell’s portfolio in PCIe and CXL switching products, with the integration of XConn’s technology and engineering team expected to support the company’s data centre connectivity strategy, particularly in AI infrastructure. Marvell will issue roughly 2.5 million common shares as part of the stock component, ...

1)PCIe协议与交换机是传统的计算机扩展总线标准，随技术迭代速率提升，但CPU、GPU等各设备间通 信速度仍存瓶颈，因此CXL协议应运而生。此外，众多厂商亦使用各自互联协议，其中NVLink处于领 先地位。 2)NVLink实现了Scale-Up中GPU与GPU间高速互联;NVSwitch则是多GPU互联推理的硬件支持，其延迟 低、通道数多、带宽高、功耗大;NVLink C2C则实现了Scale- Up中实现CPU与CPU、CPU与GPU间的高 速互联，在速率上第五代NVLink单通道的带宽为200Gbps，而PCIe Gen5为32Gbps。 3)华为灵衢(UB)提供百ns同步内存语义访问时延、2-5us异步内存语义访问时延，提供组件间TB/s级带 宽，UBProcessing Unit是支持UB协议栈的处理单元，其内嵌UBSwitch，实现多级UBSwitch扩展组网， 并支持通过UBoE与以太Switch融合组网。 4)UAlink利用以太网基础设施实现Scale-Up，UALink1.0规范支持每通道最高200GT/s的数据传输速率， 每四条物理通道组合构成一个UALink基本单元组，在发送(TX) ...

Flash | 22 Dec 2025 06:22:55 ET │ 13 pages Marvell Technology Inc (MRVL.O) Adding a Positive Catalyst Watch into CES CITI'S TAKE We hosted Marvell's SVP of Finance Steve Vermillion and SVP of Investor Relations Ashish Saran for investor meetings in Asia last week. Meetings reinforced positive conviction in our core thesis of accelerating Y/Y sales growth in 2026/27 driven by AI data centers. Fundamentally, we believe investor concerns around XPU competition at two hyperscale customers and optical/DSP opport ...

Investment Rating - The report maintains a "Recommendation" rating for the industry, indicating a positive outlook for investment opportunities in the sector [6]. Core Insights - The transition from Scale-out to Scale-up architecture in AI data centers is expected to drive significant growth in the interconnect technology market, particularly benefiting the switch chip market [3][4]. - Major technology companies are collaborating to establish new open standards like UALink to challenge Nvidia's dominance in AI data centers, which is anticipated to enhance interconnect capabilities [2][24]. - The domestic market for PCIe switch chips in China is projected to grow significantly, with estimates suggesting a market size of approximately 3.8 billion yuan in 2024, potentially reaching 17 billion yuan by 2029 [5][44]. Summary by Sections 1. Scale-up as an Upgrade Direction - Scale-up interconnect technology is becoming the preferred upgrade path for supernode solutions, addressing the limitations of traditional Scale-out architectures [1][15]. - The shift to Scale-up architecture aims to enhance performance by increasing the number of GPUs, CPUs, and storage resources within a single node, utilizing high-speed interconnect technologies like PCIe and NVLink [1][12]. 2. Growth of Switch Chip Market - The switch chip market is expected to benefit from the growth of Scale-up interconnect technology, with both PCIe switch chips and Ethernet switch chips poised for significant demand increases [3][26]. - The report highlights that the demand for AI large models will continue to drive the growth of Scale-up interconnects, positively impacting the switch chip market [3][26]. 3. Domestic Replacement Opportunities - The report identifies a substantial domestic replacement opportunity in the switch chip market, particularly for PCIe switch chips, as China is the largest market globally [5][44]. - The domestic PCIe switch chip market is expected to grow from 3.8 billion yuan in 2024 to 17 billion yuan by 2029, driven by the increasing demand for AI servers [5][44]. 4. Investment Recommendations - The report suggests focusing on companies with PCIe design capabilities, such as ZTE, Montage Technology, and others, as they are well-positioned to benefit from the growth in the switch chip market [5][54]. - Additionally, companies involved in Ethernet switch chip production, like ZTE and Sanken Communication, are also recommended for investment consideration [5][54].

Group 1 - Astera Labs, Inc. has been downgraded by Barclays' analyst Tom O'Malley from Overweight to Equal Weight, maintaining a price target of $155 [1] - The downgrade is attributed to the lack of adoption of Astera's UALink product beyond Amazon's Trainium 3 [1][2] - Barclays also downgraded other companies like Lumentum and Marvell Technology, indicating they appear less attractive compared to AI-focused companies [1] Group 2 - The firm does not see signs of Astera's adoption or momentum with United Airlines, noting a shift towards Ethernet technology [2] - Astera Labs specializes in designing, manufacturing, and selling semiconductor-based connectivity solutions for cloud and AI infrastructure [2]

Group 1 - Barclays has downgraded the ratings of Lumentum (LITE.US), Astera Labs (ALAB.US), and Marvell Technology (MRVL.US) from "Overweight" to "Hold" due to decreased attractiveness compared to other AI-focused companies [1] - Lumentum's stock price has increased by 60% over the past three months, significantly outperforming the S&P 500's 5% increase, but its risk-reward profile is no longer favorable compared to peers [1] - Astera Labs faces market penetration issues with its UALink product, lacking significant advantages outside of the Amazon Trainium 3 project, and risks being marginalized as the industry shifts towards Ethernet, ESUN, and SUE-T solutions [1] Group 2 - Marvell Technology's rating was lowered due to its specialized integrated circuit business underperforming initial expectations and additional market risks facing its data center business [1] - Barclays is not bearish on the semiconductor sector overall, having upgraded the rating of KLA Corporation (KLAC.US) to "Overweight," citing optimism about long-term growth prospects and KLA's leading position in process control [2] - KLA is expected to experience strong growth as revenue from markets outside of China is anticipated to increase with relatively low entry barriers [2]

Group 1 - Barclays has downgraded the ratings of Lumentum (LITE.US), Astera Labs (ALAB.US), and Marvell Technology (MRVL.US) from "Overweight" to "Hold" due to decreased attractiveness compared to other AI-focused companies [1] - Lumentum's stock price has increased by 60% over the past three months, significantly outperforming the S&P 500's 5% increase, but its risk-reward profile is no longer favorable compared to peers [1] - Astera Labs faces market penetration challenges with its UALink product, lacking significant advantages outside of Amazon's Trainium 3 project, and risks being marginalized as the industry shifts towards Ethernet, ESUN, and SUE-T solutions [1] - Marvell Technology's rating was lowered due to its specialized integrated circuit business underperforming initial expectations and additional market risks facing its data center business [1] Group 2 - Barclays has upgraded the rating of KLA Corporation (KLAC.US) to "Overweight," expressing optimism about the long-term prospects of process control strength and KLA's leading position in the industry [2] - The company is expected to experience strong growth cycles as revenue growth thresholds outside of China are relatively low [2]

Core Viewpoint - The article discusses the comparison of Scale Up solutions in AI servers, focusing on the UALink technology promoted by Marvell and the current mainstream Scale Up approaches in the international market [1][3]. Comparison of Scale Up Solutions - Scale Up refers to high-speed communication networks between GPUs within the same server or rack, allowing them to operate collaboratively as a large supercomputer [3]. - The market for Scale Up networks is projected to reach $4 billion in 2024, with a compound annual growth rate (CAGR) of 34%, potentially growing to $17 billion by 2029 [5][7]. Key Players and Technologies - NVIDIA's NVLink technology is currently dominant in the Scale Up market, enabling GPU interconnection and communication within server configurations [11][12]. - AMD is developing UALink, which is based on its Infinity Fabric technology, and aims to transition to a complete UALink solution once native switches are available [12][17]. - Google utilizes inter-chip interconnect (ICI) technology for TPU Scale Up, while Amazon employs NeuronLink for its Trainium chips [13][14]. Challenges in the Ecosystem - The current ecosystem for Scale Up solutions is fragmented, with various proprietary technologies leading to compatibility issues among different manufacturers [10][22]. - Domestic GPU manufacturers face challenges in developing their own interconnect protocols due to system complexity and resource constraints [9]. Future Trends - The article suggests that as the market matures, there will be a shift from proprietary Scale Up networks to open solutions like UAL and SUE, which are expected to gain traction by 2027-2028 [22]. - The choice between copper and optical connections for Scale Up networks is influenced by cost and performance, with copper currently being the preferred option for short distances [20][21].

Core Insights - Astera Labs is leading the development of semiconductor-based connectivity solutions for rack-scale AI infrastructure, emphasizing the shift towards unified computing platforms rather than individual servers [1][2] - The company is showcasing its ecosystem collaborations at the 2025 OCP Global Summit, highlighting the importance of open standards for AI Infrastructure 2.0 [1][2] Industry Trends - The AI infrastructure landscape is transitioning from server-level architectures to rack-scale systems, driven by significant investments from hyperscalers [2] - Open standards are essential for integrating diverse accelerators, interconnects, and management tools, enabling optimized solutions for specialized AI workloads [2] Ecosystem Collaborations - Astera Labs is collaborating with various industry leaders, including AMD, Arm, and Molex, to enhance AI infrastructure through high-performance connectivity solutions [3][4][9] - These partnerships focus on delivering reliable, high-speed cable solutions and ensuring robust signal integrity across rack-scale distances [3][9] Technical Innovations - The company is presenting technical sessions on UALink deployment strategies and PCIe 6 security considerations at the OCP Global Summit [2] - Astera Labs' Intelligent Connectivity Platform integrates multiple semiconductor-based technologies, including CXL, Ethernet, PCIe, and UALink, to create cohesive systems [13] Market Position - Astera Labs positions itself as a key player in the AI infrastructure market by providing purpose-built connectivity solutions grounded in open standards [13] - The company's collaborations aim to accelerate the adoption of open rack architectures, enhancing performance, interoperability, and scalability for customers [10][12]