公众号记得加星标⭐️，第一时间看推送不会错过。 近日，博通宣布其Tomahawk 6 - Davisson（或TH6-Davisson）产品现已上市。这意义重大， 因为这是一款共封装光纤 (CPO) 以太网交换机，可提供惊人的 102.4 Tb/s。而且，随着它的出 现，它将共同封装的光器件带入了下一代网络。 CPO 通过将光转换引擎与交换机 ASIC 并排嵌入，避免了传统可插拔光模块的缺点，信号无需通过 长距离电气线路传输，而是几乎立即耦合到光纤中。因此，电气损耗降低至 4 分贝，每端口功耗降至 9W。这种布局省去了众多可能出现故障的组件，并大大简化了光互连的实施。 在博通看来，共封装光学器件 (CPO) 是一种先进的异构集成技术，将光学器件和硅片集成在单个封 装基板上，旨在应对下一代带宽、功耗和成本挑战。CPO 融合了光纤、数字信号处理 (DSP)、ASIC 以及先进的封装和测试技术，为支持横向扩展和纵向扩展网络的数据中心互连提供颠覆性的系统价 值。 博通进一步指出，如今，可插拔光模块内部需要高功率 DSP，以补偿信号从 ASIC 传输过程中产生 的路径互连损耗。随着 SerDes 技术扩展到 212 G ...

Core Viewpoint - The article discusses advancements in optical chip technology presented at the Hot Chips 2025 conference, highlighting innovations from companies like NVIDIA, Celestial AI, and Ayar Labs, which aim to enhance data center performance and efficiency through silicon photonics and optical interconnects [2][75][236]. Group 1: NVIDIA's Innovations - NVIDIA emphasizes the need for co-packaged optics to significantly enhance AI factory scalability, noting that AI factories consume approximately 17 times more optical power than traditional cloud data centers [3][5]. - The Spectrum-X Ethernet technology aims to reduce the cost of network photonics, which constitutes about 10% of the total computing power cost in AI factories [3][16]. - Spectrum-X Ethernet photonics is claimed to be the first to implement 200 G/channel SerDes technology, improving signal integrity and reducing power consumption by decreasing the number of lasers required for high-speed links [16][34]. Group 2: Celestial AI's Photonic Fabric - Celestial AI showcases its Photonic Fabric module, which utilizes optical connections to link next-generation GPUs and accelerators, moving away from traditional electrical connections [75][78]. - The company is developing its own optical MAC (OMAC) to achieve high energy efficiency by matching SerDes with channels [80][92]. - Celestial AI's technology aims to optimize the integration of optical I/O within chip designs, enhancing performance and reducing power consumption [92][94]. Group 3: Ayar Labs' Optical I/O Solutions - Ayar Labs introduces a UCIe optical I/O retimer, designed to facilitate the integration of optical I/O into chip packaging, achieving data rates of up to 8 Tbps [108][122]. - The company addresses the challenge of scaling interconnects for large clusters of chips, which is critical for AI applications [111][117]. - Ayar Labs' technology aims to decouple optical and electrical signal transmission, allowing for larger systems compared to traditional electrical I/O [162][169]. Group 4: Lightmatter's Passage M1000 - Lightmatter presents the Passage M1000, which integrates co-packaged optics and silicon photonics, promising bandwidths of up to 114 Tbps [170][184]. - The design focuses on compact optical I/O, utilizing silicon micro-rings for efficient modulation [191][196]. - Lightmatter aims to enable optical circuit switching for redundancy and scalability in high-performance computing environments [211][224]. Group 5: OpenLight's Photonic Solutions - OpenLight Photonics aims to accelerate the transition to silicon photonics with its proprietary photonic application-specific integrated circuits (PASIC), designed for high-density, low-power optical interconnects [234][236]. - The company emphasizes customization in its PASIC offerings, allowing clients to tailor components to specific applications, enhancing flexibility in rapidly evolving markets [237][239]. - OpenLight's technology is positioned to address the growing demand for faster data transmission in AI data centers, with plans to expand its component library and production capabilities [238][239].

Core Viewpoint - The increasing demand for communication between AI GPU clusters is driving the shift towards optical communication for inter-network layer communication, with Nvidia planning to implement silicon photonic interconnect technology and co-packaged optics (CPO) in its next-generation AI platform by 2026 [2][9]. Group 1: Nvidia's CPO Technology - Nvidia's CPO technology integrates optical engines directly with switch ASICs, significantly reducing electrical losses and power consumption, achieving a reduction in electrical loss to 4 dB and power per port to 9W [8]. - The efficiency of CPO is highlighted by improvements such as a 3.5 times increase in power efficiency, a 64 times enhancement in signal integrity, and a 10 times increase in resilience due to fewer active components [8][13]. - Nvidia's Quantum-X InfiniBand switch will offer 115 Tb/s throughput with 144 ports, each supporting 800 Gb/s, and will feature liquid cooling [9]. Group 2: TSMC's COUPE Roadmap - TSMC's COUPE roadmap consists of three phases, starting with an optical engine for OSFP connectors capable of 1.6 Tb/s, which is twice the speed of current top copper Ethernet solutions [15][18]. - The second generation will integrate COUPE into CoWoS packaging, achieving 6.4 Tb/s at the motherboard level, while the third generation aims for 12.8 Tb/s within processor packaging [18]. Group 3: Industry Implications - The adoption of CPO technology is positioned as a structural necessity for future AI data centers, providing a competitive edge over rivals like AMD [13]. - The shift to optical communication is essential for managing the complexities of large-scale AI clusters, which require low-latency and high-bandwidth interconnects [4][6].

Core Viewpoint - The article discusses Nvidia's announcement of a co-packaged optics (CPO) switch aimed at significantly reducing power consumption in AI data centers, marking a potential breakthrough in optical networking technology [1][5][18]. Group 1: Technology Overview - The CPO switch integrates optical and electronic components to enhance bandwidth and reduce power consumption by minimizing the distance electronic signals must travel [2][4]. - Nvidia's CPO technology claims to reduce power consumption by 70%, from 30W per 1.6T pluggable transceiver to just 9W per CPO port [5][17]. - The CPO switch is designed to handle data rates of 1.6 Tb/s, utilizing micro-ring modulators (MRM) for improved power efficiency [13][17]. Group 2: Market Implications - The introduction of CPO technology is seen as a significant advancement that could lead to a reduction in the number of lasers required in AI data centers by 75%, thus saving substantial energy [18][19]. - Nvidia's CPO switch is expected to enhance the reliability of data transmission by 63% and improve the ability to withstand network interruptions by 10 times [18]. - The company plans to launch two types of switches, Spectrum-X and Quantum-X, with Quantum-X expected to be available later this year [19]. Group 3: Competitive Landscape - Other companies, such as Broadcom, are also developing CPO switches, but Nvidia's approach with micro-ring modulators differs fundamentally from Broadcom's use of Mach-Zehnder modulators [20][24]. - Micas Networks has announced a 51.2T product based on Broadcom's CPO platform, which offers a 50% reduction in power consumption [22][23]. - The competition in the CPO market is intensifying, with various companies exploring different optical technologies to meet the growing demands of data centers [20][22]. Group 4: Future Developments - Nvidia is actively researching new optical technologies to enhance the scalability of its networking solutions, with plans for future integration of optical interconnects into GPUs [28][29]. - The company is collaborating with multiple partners, including TSMC and Coherent, to optimize the technology for AI data center needs [19][14]. - The ongoing development of CPO technology is expected to lead to further innovations in optical networking, potentially transforming data center architectures [26][28].