← 返回全部时间线
TheElec(韩国电子产业媒体)原文发布 10-01 20:50Radar 收录 10-02 00:46查看原文 ↗

村田公布整合光学、电气与热技术的CPO路线图

Murata Unveils CPO Road Map Integrating Optical, Electrical and Thermal Technologies
中文摘要

村田制作所(Murata Manufacturing Co.)公布面向人工智能数据中心的共封装光学(CPO)技术路线图,说明其CPO解决方案的开发方向,该方案此前已首次推出。路线图整合光学、电气与热管理技术。

原文深度解读

本文介绍村田制作所(Murata Manufacturing Co.)公布的面向人工智能数据中心的共封装光学(CPO)技术路线图,涉及时间节点为2026年3月OFC展会首次推出CPO解决方案、2028年量产目标、2030年与2032年技术演进规划。关键数字包括:电光带宽超过100 GHz、单通道传输速度超过400 Gbps、2030年提升至800 Gbps、LTCC基板样品尺寸125 mm、2030年扩展至200 mm。与电子元器件供应链的关系在于:村田作为MLCC、硅电容、电感等被动元件供应商,正将高频设计、阻抗匹配、陶瓷与有机基板技术延伸至光通信组件,并规划TFLN光调制器、LTCC封装基板、LCP柔性布线等产品的开发与量产节奏,可能影响CPO相关光学引擎、封装基板与高频材料的供应格局。

  • 村田在2026年3月OFC光通信展首次推出CPO解决方案,并公布面向AI数据中心的技术路线图,计划到2032年推进整合电气、光学与热管理功能的光互连产品,聚焦光调制器、封装基板与布线技术三个领域。
  • CPO将光学引擎放置在交换机ASIC等芯片附近,缩短电信号传输距离,用光缆替代铜缆以实现高速大容量数据传输并降低信号损耗与功耗;光学引擎从机架外移至主板并靠近芯片,要求高速光调制以及高密度封装与布线技术。
  • 村田已为光通信模块等应用供应MLCC、硅电容(Si-Caps)和电感等被动元件,并将移动射频滤波器与射频模块业务中积累的高频设计、阻抗匹配以及陶瓷和有机基板技术应用于光通信组件。
  • TFLN光调制器基于村田智能手机SAW滤波器业务中积累的材料与工艺技术,目标电光带宽超过100 GHz、单通道传输速度超过400 Gbps,可支持单激光多通道结构;计划今年第四季度完成首个样品开发,2028年开始量产,2030年将单通道传输速度提升至800 Gbps,并拓展至相干通信、WDM和量子通信。
  • 村田选择陶瓷作为集成PIC与EIC的封装基板材料,因其热导率高于有机基板、吸湿性更低,在温湿度变化下可靠性更高,且高频信号损耗低、可在基板内部形成3D腔体;LTCC基板由多层陶瓷与金属布线堆叠而成,已产出125 mm样品,支持单通道400 Gbps高速电信号,原型生产约需三个月。
  • 村田计划2028年推出嵌入被动元件与冷却功能的LTCC,并开发在所需区域嵌入RDL的结构;2030年计划在基板上形成光波导并将面板尺寸扩展至200 mm,目标是将电信号、供电、热管理与光信号传输集成到单一陶瓷封装中。

供应链影响

  • 若村田按路线图在2028年实现TFLN光调制器量产、2028年推出带被动元件与冷却功能的LTCC、2030年扩展至200 mm面板并加入光波导,CPO相关的高速光调制器、陶瓷封装基板与柔性布线材料供应可能逐步增加,但实际节奏取决于样品验证与客户导入进度。
  • 村田将MLCC、硅电容、电感等被动元件以及高频设计、阻抗匹配、陶瓷与有机基板技术延伸至光通信组件,可能使光通信模块与CPO封装对高频被动元件和基板材料的需求结构发生变化,具体影响取决于CPO在AI数据中心的采用速度。
  • LTCC与LCP基板均支持单通道400 Gbps电信号,且村田规划2030年将光波导、光耦合器、光间距转换器集成到LCP基板,可能推动封装基板与布线环节向光电热一体化方向演进,但电光共载集成布线仍处概念设计阶段,短期供应格局取决于技术成熟度。
  • CPO将光学引擎与ASIC集成在单一封装中导致维修更换比传统可拆卸光模块更困难,可能影响数据中心运营商对可维护性与备件策略的考量,进而影响CPO与传统可插拔光模块的采用比例,但最终取决于村田等供应商提出的结构与解决方案能否满足客户要求。
  • 村田当前聚焦理解超大规模客户等终端客户的系统规格,并据此开发兼顾带宽、光学性能、电源完整性与热管理的设计,可能使其CPO相关产品的规格定义与量产时间表受终端客户需求牵引,供应链上游材料与设备准备节奏也取决于这些规格的明确时点。

系统解读,仅作为判断线索,不构成备货、出货、涨价或投资建议。

正文内容 · AI翻译

村田的CPO解决方案。(照片:THE ELEC)

村田制作所(Murata Manufacturing Co.)公布了面向人工智能数据中心的共封装光学(CPO)技术路线图,详细说明了其CPO解决方案的开发方向,该方案于今年3月在OFC 2026光通信展上首次推出。该公司计划到2032年推进整合电气、光学与热管理功能的光互连产品,聚焦三个领域:光调制器、封装基板和布线技术。

CPO将光学引擎放置在交换机专用集成电路(ASIC)等芯片附近,缩短电信号需要传输的距离。通过使用光缆代替铜缆,该技术可实现大量数据的高速传输,同时降低信号损耗和功耗。随着此前位于机架外部的光学引擎移至主板上并靠近芯片,需要高速光调制以及高密度封装和布线技术。

村田已为光通信模块等应用供应一系列被动元件,包括多层陶瓷电容器(MLCC)、硅电容器(Si-Caps)和电感。该公司正将移动射频(RF)滤波器和射频模块业务中积累的高频设计、阻抗匹配以及陶瓷和有机基板技术应用于光通信组件。

TFLN光调制器目标第四季度出样,2028年量产

村田的TFLN光调制器。(照片:THE ELEC)

村田正在开发一种光调制器,用于在光子集成电路(PIC)内部将电信号转换为光信号。该器件将数据加载到光学引擎中激光器产生的光上,从而生成光信号。它使用薄膜铌酸锂(TFLN),即将铌酸锂($\text{LiNbO}_3$)加工成薄膜形态。

该技术基于村田智能手机声表面波(SAW)滤波器业务中积累的材料和工艺技术。SAW滤波器允许选定频段通过,同时滤除其他信号。$\text{LiNbO}_3$适合高速光通信,因为其光学特性可随电信号快速变化。村田的目标是电光(EO)带宽超过100吉赫兹(GHz),单通道传输速度超过400吉比特每秒(Gbps)。还可实现能够从单个激光器跨多个通道生成光信号的多通道结构。

村田计划在今年第四季度完成其首个TFLN光调制器样品的开发。该公司还计划获得绝缘体上铌酸锂(LNOI)晶圆的自有技术,该晶圆具有薄$\text{LiNbO}_3$层,以及器件制造工艺。该公司目标是在2028年开始量产该产品,并在2030年将单通道传输速度提升至800 Gbps。应用还将从CPO扩展到用于长距离传输的相干通信、波分复用(WDM)和量子通信。

耐热耐湿陶瓷封装基板,LTCC

村田的LTCC基板。(照片:THE ELEC)

村田已选择陶瓷作为集成PIC和电子集成电路(EIC)的封装基板材料。与有机基板相比,陶瓷具有更高的热导率和更低的吸湿性,在温度和湿度变化下提供更高的可靠性。它还具有低高频信号损耗,并允许在基板内部形成三维(3D)腔体。这有助于为高密度封装创造空间。

低温共烧陶瓷(LTCC)基板通过堆叠多层陶瓷和金属布线制成。村田已生产出125毫米(mm)基板样品,目前专注于电信号传输。它支持每通道400 Gbps的高速电信号。原型生产大约需要三个月。

2028年,村田计划推出嵌入被动元件和冷却功能的LTCC。该公司还在开发一种在所需区域嵌入再分布层(RDL)的结构。这将为迄今专注于电信号传输的LTCC增加电源稳定和散热功能。2030年,该公司计划在基板上形成光波导,并将面板尺寸扩展至200 mm。其目标是将电信号、供电、热管理和光信号传输集成到单一陶瓷封装中。

柔性布线结构克服空间限制

村田使用LCP的光电集成布线结构。(照片:THE ELEC)

村田正在将液晶聚合物(LCP)材料应用于连接光学引擎与周边组件的布线。LCP是一种高性能塑料材料,可加工成薄膜。它可高速传输电信号,同时可自由弯曲。该材料还能保持弯曲形状,允许在空间受限环境中根据组件布局安排布线。另一个特点是可以无需粘合剂地堆叠多层。

村田已生产出使用LCP进行电信号传输的样品基板。它支持每通道400 Gbps的高速电信号。基板的柔性特性允许根据光源等组件的位置布置布线。其高耐热性也使其可用于回流焊等高温工艺。然而,能够在单一基板上同时承载电信号和光信号的集成布线技术仍处于概念设计阶段。

2030年,村田计划为其LCP基板增加光布线功能。该公司计划引入承载光的光波导、连接光信号的光耦合器以及调整光信号间距的光间距转换器。到2032年,它计划将电布线和光布线集成到单一基板中。根据传输距离,可选择适当的传输方式——电或光。村田还设想了一种在多层之间路由多个光信号的结构。

通过光电热协同设计实现系统优化

村田的目标是将光调制器、封装基板和布线技术集成到单一光通信系统中。通过“光电热协同设计”,它旨在优化整个系统的性能,而非单个组件。目前正在研究利用包括TFLN、LTCC和LCP在内的自有技术,以及通过与外部公司合作,在系统层面整合光学、电气和热技术。

“AI数据中心容量越来越大、速度越来越快、密度越来越高,仅提升单个光学器件的性能已经不够,”韩国村田电子高级专家Kim Sa-yeol表示。“随着光学器件性能提升,还必须考虑电信号的高频特性,而随着集成密度上升,热管理变得越来越重要。”他补充说,“最终,村田光通信解决方案的目标是通过光电热协同设计优化整个系统的性能。”

维护和维修也仍是挑战。由于光学引擎和ASIC集成在单一封装中,发生故障时更换和维修组件比传统可拆卸光模块更困难。村田正在研究各种结构和解决方案以解决该问题。

该公司目前专注于理解终端客户(包括超大规模客户)所需的系统规格。基于这些要求,它正致力于开发不仅考虑CPO所需带宽和光学性能,还考虑电源完整性和热管理的设计。

Murata’s CPO solution. (Photo: THE ELEC)

Murata Manufacturing Co. has unveiled a technology road map for co-packaged optics (CPO) for artificial intelligence data centers, detailing the development direction of its CPO solution first introduced at the OFC 2026 optical communications exhibition in March. The company plans to advance optical interconnect products integrating electrical, optical and thermal functions through 2032, focusing on three areas: optical modulators, package substrates and wiring technologies.

CPO places optical engines near chips such as switch application-specific integrated circuits (ASICs), reducing the distance electrical signals need to travel. By using optical cables instead of copper, the technology can enable high-speed transmission of large volumes of data while reducing signal loss and power consumption. As optical engines that were previously positioned outside the rack move onto the main board and closer to chips, technologies for high-speed optical modulation as well as high-density packaging and wiring are required.

Murata has supplied a range of passive components, including multilayer ceramic capacitors (MLCCs), silicon capacitors (Si-Caps) and inductors, for optical communications modules and other applications. It is applying high-frequency design, impedance-matching and ceramic and organic substrate technologies developed through its mobile radio-frequency (RF) filter and RF module businesses to optical communications components.

TFLN Optical Modulator Targets Samples in Fourth Quarter, Mass Production in 2028

Murata’s TFLN optical modulator. (Photo: THE ELEC)

Murata is developing an optical modulator that converts electrical signals into optical signals inside a photonic integrated circuit (PIC). The device carries data on light generated by the laser in an optical engine to create an optical signal. It uses thin-film lithium niobate (TFLN), in which lithium niobate ($\text{LiNbO}_3$) is processed into a thin-film form.

The technology builds on materials and process technologies developed through Murata's surface acoustic wave (SAW) filter business for smartphones. SAW filters allow selected frequency bands to pass while filtering out other signals. $\text{LiNbO}_3$ is suited to high-speed optical communications because its optical properties can be rapidly changed in response to electrical signals. Murata is targeting an electro-optic (EO) bandwidth of more than 100 gigahertz (GHz) and transmission speeds of more than 400 gigabits per second (Gbps) per channel. A multichannel structure capable of generating optical signals across multiple channels from a single laser is also possible.

Murata plans to complete development of its first TFLN optical-modulator sample in the fourth quarter of this year. It also plans to secure proprietary technology for lithium niobate on insulator (LNOI) wafers, which have a thin $\text{LiNbO}_3$ layer, as well as the device manufacturing process. The company aims to begin mass production of the product in 2028 and raise transmission speed to 800 Gbps per channel in 2030. Applications will also expand beyond CPO to coherent communications used for long-distance transmission, wavelength-division multiplexing (WDM) and quantum communications.

Heat- and Humidity-Resistant Ceramic Package Substrate, LTCC

Murata’s LTCC substrate. (Photo: THE ELEC)

Murata has selected ceramic as the package-substrate material for integrating PICs and electronic integrated circuits (EICs). Ceramic has higher thermal conductivity and lower moisture absorption than organic substrates, providing greater reliability amid changes in temperature and humidity. It also offers low high-frequency signal loss and allows three-dimensional (3D) cavities to be formed inside the substrate. This can help create space for high-density packaging.

Low-temperature co-fired ceramic (LTCC) substrates are made by stacking multiple layers of ceramic and metal wiring. Murata has produced a 125-millimeter (mm) substrate sample currently focused on electrical signal transmission. It supports high-speed electrical signals of 400 Gbps per channel. Prototype production takes about three months.

In 2028, Murata plans to unveil LTCC with embedded passive components and cooling functions. It is also developing a structure in which redistribution layers (RDLs) are embedded in required areas. This would add power stabilization and heat dissipation to LTCC, which has so far focused on electrical signal transmission. In 2030, the company plans to form optical waveguides on the substrate and expand the panel size to 200 mm. Its goal is to integrate electrical signaling, power delivery, thermal management and optical signal transmission into a single ceramic package.

Flexible Wiring Structure to Overcome Space Constraints

Murata’s opto-electrical integrated wiring structure using LCP. (Photo: THE ELEC)

Murata is applying liquid-crystal polymer (LCP) material to wiring that connects optical engines with surrounding components. LCP is a high-performance plastic material that can be processed into thin films. It can transmit electrical signals at high speed while being freely bent. The material can also retain its bent shape, allowing wiring to be arranged according to component placement in space-constrained environments. Another feature is that multiple layers can be stacked without adhesive.

Murata has produced a sample substrate using LCP for electrical signal transmission. It supports high-speed electrical signals of 400 Gbps per channel. The flexible characteristics of the substrate allow wiring to be positioned according to the location of components such as light sources. Its high heat resistance also enables use in high-temperature processes such as reflow. Integrated wiring technology capable of carrying both electrical and optical signals on a single substrate, however, remains at the conceptual-design stage.

In 2030, Murata plans to add optical wiring functions to its LCP substrate. The company plans to incorporate waveguides that carry light, optical couplers that connect optical signals, and optical pitch converters that adjust the spacing between optical signals. By 2032, it plans to integrate electrical and optical wiring into a single substrate. Depending on transmission distance, the appropriate transmission method—electrical or optical—could be selected. Murata is also envisioning a structure in which multiple optical signals are routed across multiple layers.

System Optimization Through Opto-Electro-Thermal Co-Design

Murata's goal is to integrate optical modulators, package substrates and wiring technologies into a single optical communications system. Through "Opto-Electro-Thermal Co-design," it aims to optimize the performance of the entire system rather than individual components. Research is underway to integrate optical, electrical and thermal technologies at the system level using proprietary technologies including TFLN, LTCC and LCP, as well as through cooperation with external companies.

"AI data centers are becoming larger in capacity, faster and denser, making it insufficient to simply improve the performance of individual optical devices," said Kim Sa-yeol, senior professional at Korea Murata Electronics. "As the performance of optical devices increases, the high-frequency characteristics of electrical signals must also be considered, and as integration density rises, thermal management becomes increasingly important." He added that "ultimately, Murata’s goal for optical communications solutions is to optimize the performance of the entire system through opto-electro-thermal co-design."

Maintenance and repair also remain challenges. Because the optical engine and ASIC are integrated into a single package, replacing and repairing components in the event of a failure is more difficult than with conventional detachable optical modules. Murata is reviewing various structures and solutions to address the issue.

The company is currently focused on understanding the system specifications required by end customers, including hyperscalers. Based on those requirements, it is working to develop designs that consider not only the bandwidth and optical performance required for CPO, but also power integrity and thermal management.

↑