Financial Data and Key Metrics Changes - In Q3, the company reported revenue of $425.2 million and non-GAAP EPS of $0.57, both exceeding the high end of guidance [20] - Non-GAAP gross margin was 35.2%, up 290 basis points sequentially and 650 basis points year on year, attributed to better manufacturing utilization and a favorable product mix [20][21] - Non-GAAP operating margin improved to 10.8%, up 290 basis points sequentially and 1,100 basis points year on year [20][21] Business Line Data and Key Metrics Changes - Cloud and Networking segment revenue reached $365.2 million, increasing 8% sequentially and 16% year on year, driven by strong demand from hyperscale cloud customers [9][22] - Industrial Tech segment revenue was $60 million, down 5% sequentially but up 14% year on year, with ultrafast laser shipments remaining steady [14][15][22] Market Data and Key Metrics Changes - The company anticipates strong sequential growth in the Cloud and Networking segment in Q4, driven by new capacity and increasing demand from network equipment manufacturers [14] - The Industrial Tech segment is expected to see a sequential decline in revenue due to ongoing macroeconomic headwinds and typical seasonal trends [16] Company Strategy and Development Direction - The company aims to achieve quarterly revenue of $750 million, gross margins above 40%, and operating margins greater than 20% [6][18] - Focus on high-growth areas such as cloud and AI, with investments directed towards expanding manufacturing capacity and product offerings [17][18] Management's Comments on Operating Environment and Future Outlook - Management expressed confidence in achieving medium and long-term financial targets despite macroeconomic uncertainties and tariff impacts [18] - The company is building resilience through a diversified manufacturing footprint and flexible supply chain [18] Other Important Information - The company is ramping production of CW lasers and EMLs, with expectations to double EML shipments by the end of calendar 2025 [10][11] - The company is also focusing on co-packaged optics (CPO) as a future growth area, although significant adoption is expected to take several years [118] Q&A Session Summary Question: How is the company tracking towards the $500 million revenue target? - Management confirmed they are still on track for the $500 million target by the end of the year, emphasizing confidence in current performance [30] Question: Update on datacom chip business performance? - The company is outperforming initial expectations in terms of volume and ASP, with a focus on increasing capacity for EMLs and CW lasers [32][34] Question: Clarification on tariff headwinds? - The company quantified a 100 basis point headwind to gross margins due to increased component costs and tariffs, with plans to move more production to Thailand to mitigate this [41][42] Question: Contributions from co-packaged optics? - Management indicated that while CPO is a long-term prospect, they are looking to add components to the CPO ecosystem over the coming years [48][118] Question: Insights on telecom performance and supply issues? - Telecom performance met expectations, with improvements anticipated in supply for pumps and tunables moving into Q4 [97][99] Question: Guidance caution related to macro or tariffs? - Management acknowledged macro uncertainty but expressed confidence in the guidance provided, highlighting strong demand across product lines [100][102]

如果您希望可以时常见面，欢迎标星收藏哦~ 2025年，AI大模型赛道竞争白热化。以ChatGPT、Grok、DeepSeek与Gemini为代表的全球 四大主流通用语言模型，正经历一场前所未有的技术迭代与商业格局重塑。 ChatGPT依托封闭的MoE（Mixture of Experts）架构构建出领先的多模态生态体系，其成熟的 API服务体系已占营收的70%，展现出极强的变现能力；Grok则以动态推理网络为核心，主打实时 数据响应，并通过与X平台深度捆绑实现商业化突破；DeepSeek以开源模型为基础，针对中文场 景深度优化，广泛部署于政企私有化系统中，快速拓展本地市场；Gemini则依托Google Pathways 系统，具备强大的算力基座，且已深度集成至Workspace生态，打通了办公与AI应用的边界。 在这场大模型技术与生态竞速的背后，是AI巨头们在2025年持续加码的资本投入与算力争夺。而 AI应用对高速、低延迟数据交互的需求，也正成为推动光通信技术快速演进的关键驱动力。 4月23日，在第三届九峰山论坛暨化合物半导体产业博览会的下午论坛上， 华工正源光子技术有限 公司总经理胡长飞 发表了 《光速 ...

Core Viewpoint - Intel aims to control capital expenditures at $18 billion for Q1 2025, adopting aggressive strategies to enhance returns on invested capital while reducing spending [1][11]. Group 1: Capital Expenditure and Financial Strategy - The company plans to maintain capital expenditures at $18 billion for 2025, with a focus on improving returns on invested capital [1][11]. - Long-term capital intensity under the Integrated Device Manufacturer (IDM) model is estimated at around 25% [1]. - The company is optimistic about the data center business for the remainder of the year, driven by product launches like Granite Rapids and Xeon 6 series [1][10]. Group 2: Product Development and Market Position - The transition to Lunar Lake in the client product portfolio may pressure profit margins due to memory and packaging issues [3]. - The Panther Lake product is expected to have a competitive market presence, with initial products launching by the end of the year [6][13]. - The company is focusing on enhancing internal production capabilities, aiming for a 70% internal production ratio for Panther Lake [16]. Group 3: Market Dynamics and Customer Demand - There is strong demand for older generation products like Raptor Lake, driven by macroeconomic factors and customer risk management strategies [12]. - The data center business has shown higher-than-expected growth, primarily from large-scale customers, with a focus on stabilizing market share and increasing average selling prices [10][14]. - The company is exploring new architectures and disruptive platforms in the AI space, emphasizing workload understanding and collaboration with industry leaders [9][18]. Group 4: Operational Efficiency and Cost Management - The company is working to balance internal and external manufacturing to optimize capital intensity and avoid over-investment [20]. - Operational expenses for 2025 and 2026 include costs related to Altera, which will be adjusted upon the completion of its divestiture [21]. - The company is committed to improving yield efficiency in advanced packaging and testing operations [20].

如果您希望可以时常见面，欢迎标星收藏哦~ 来源：内容编译自IEEE，谢谢。 人们期待已久的新兴计算机网络组件可能终于迎来了它的时代。在上周于圣何塞举行的Nvidia GTC 活动上，该公司宣布将生产一种光纤网络交换机，旨在大幅降低 AI数据中心的功耗。该系统称为共 封装光学器件（CPO）交换机，每秒可将数十兆比特的数据从一个机架中的计算机路由到另一个机 架中的计算机。 如今，在数据中心中，计算机机架中的网络交换机由专用芯片组成，这些芯片通过电气方式与插入 系统的光收发器相连（机架内的连接是电气的，但有几家初创公司希望改变这一现状）。可插拔收 发器结合了激光器、光电路、数字信号处理器和其他电子设备。它们与交换机建立电气连接，并在 交换机端的电子比特和沿光纤穿过数据中心的光子之间转换数据。 共封装光学器件是一种通过将光/电数据转换尽可能靠近交换芯片来提高带宽并降低功耗的方法。这 简化了设置，并通过减少所需的独立组件数量和电子信号必须传输的距离来节省电力。先进的封装 技术使芯片制造商能够用多个硅光收发器芯片包围网络芯片。光纤直接连接到封装上。因此，除激 光器外，所有组件都集成到一个封装中，激光器保持外部，因为它们是 ...

如果您希望可以时常见面，欢迎标星收藏哦~ 黄仁勋在演讲中将Spectrum-X称为"全垒打式创新"，旨在将以太网提升至InfiniBand的性能水 平。他表示Spectrum-X将在Rubin平台时代实现"数十万GPU"的互联。尽管英伟达未透露细节， 但Spectrum-X Photonics平台采用了与Quantum-X Photonic不同的设计架构，可能代表第二代技 术演进。 无论如何，英伟达的入局将为CPO生态注入强心剂。自2021年初起，该技术主要由博通推动。预 计到2027年，两家巨头都将推出基于单通道200G的CPO交换机，届时产业生态将趋于成熟。 下图展示了该技术涉及的多元组件生态：由台积电制造的电子芯片与光子芯片通过3D堆叠封装集 成，台积电的紧凑型通用光子引擎（COUPE）技术整合了用于光纤阵列表面耦合的微透镜。在 Quantum-X Photonic平台中，光学引擎组件通过中介层与交换机ASIC连接。虽然英伟达未披露各 环节的具体分工，但其CPO合作伙伴名单已涵盖Browave、Coherent、康宁、Fabrinet、富士康、 Lumentum、Senko、SPIL、住友、天孚通信和台 ...