在人工智能的狂飙突进中，市场的目光往往聚焦于 GPU 本身的算力迭代，却容易忽视支撑这 一庞大算力网络的"神经系统"。 当所有人还在为 H100、B200 的产能分配争论不休时，英伟达创始人黄仁勋已经将视线投向 了更深层的基础设施——光互联。 近日，市场传出 NVIDIA 向光通信巨头 Lumentum 与 Coherent 各注资 20 亿美元，并签下 巨额采购承诺与未来产能权益。 这一动作看似是财务投资，实则是算力基础设施的"战略围堵"。 这标志着 AI 产业链的竞争逻辑正在发生根本性转变：从单纯的芯片性能竞赛，演变为对上游 核心供应链的资本绑定与产能锁定。 光模块，尤其是共封装光学（CPO）技术，正站在大爆发的前夜，一场属于光互联的"军备竞 赛"已经悄然打响。 不是投资，是锁仓：英伟达 在为"光模块缺货"做预案 【如需和我们交流可扫码添加进社群】 要理解英伟达这一举动的深意，必须回顾过去两年 AI 服务器利润曲线被 HBM（高带宽内存） 反向"卡脖子"的教训。 在 AI 算力爆发的初期，市场普遍认为瓶颈在于 GPU 的制造产能。然而，随着需求激增，真 正的制约因素迅速转移到了 HBM 上。 存储三巨头（ ...

英伟达分别向光子技术巨头Lumentum和Coherent各注资20亿美元，并锁定多十亿美元采购承诺及未来产能权益。此举剑指共封装光学（CPO）核心 供应链，旨在为下一代"吉瓦级AI工厂"筑牢光互联底座。受此消息影响，Lumentum收盘上涨近12%，Coherent涨幅逾15%。英伟达当日上涨近3%。 英伟达宣布向两家光子技术公司合计投资40亿美元，将对AI数据中心网络架构的战略布局推向新高度。 周一，英伟达宣布分别向Lumentum和Coherent各投资20亿美元， 两笔交易均包含数十亿美元的采购承诺及对先进激光组件的未来产能权益。 消息公布后，两家光巨头股价拉升，Lumentum收盘上涨近12%，Coherent涨幅逾15%。英伟达当日上涨近3%。 此次投资直指英伟达正在大力推进的共封装光学（CPO）技术路线。CPO将可插拔收发器直接集成至交换机内部，可降低数据中心运营商的硬件采购 成本，并显著压缩网络能耗。两家被投公司将把所获资金用于研发投入及美国本土制造产能扩张。 战略意图：锁定CPO供应链核心环节 英伟达此次出手，与其去年在数据中心交换机产品线中引入CPO技术的战略部署高度吻合。 据悉，英伟达于 ...

Group 1 - The core viewpoint of the articles highlights the significant rise in optical communication concept stocks in the U.S. market, driven by advancements in Co-Packaged Optics (CPO) technology and supportive government policies [1][2]. - Companies such as Credo Technology, Corning, Ciena Technologies, and Lumentum have shown notable stock price increases, indicating strong market interest in optical communication technologies [1]. - NVIDIA is advancing the scale deployment of CPO technology to address the growing power consumption and reliability challenges of AI supercomputers, integrating optical engines into the same package as switching chips [1]. Group 2 - In the A-share market, as of February 6, ten CPO concept stocks have received net financing purchases exceeding 100 million yuan, with Tianfu Communication and Zhongji Xuchuang leading at 2.542 billion yuan and 2.21 billion yuan respectively [2]. - Tianfu Communication has reported that its 1.6T optical engine products have entered normal delivery stages, and the company is coordinating supply chain resources to enhance production capacity [2]. - New Yisong expects to achieve a net profit attributable to shareholders of 9.4 billion to 9.9 billion yuan by 2025, representing a year-on-year growth of 231.24% to 248.86%, and has successfully mass-produced and delivered 1.6T optical module products [2].

Group 1: Market Performance - The optical communication concept stocks continued to perform strongly in the US stock market, with Credo Technology rising by 10.78%, Corning by 7.54%, Ciena Technology by 6.78%, and Lumentum by 4.56% [1][8] Group 2: Technological Developments - NVIDIA is advancing the scale deployment of Co-Packaged Optics (CPO) technology to address the increasing power consumption and reliability challenges of AI supercomputers. CPO integrates optical engines into the same package as switching chips, significantly shortening electrical channel lengths and reducing signal conversion steps [2][9] - NVIDIA has developed CPO-supported switches for its Spectrum-X Ethernet platform and Quantum-X InfiniBand platform, collaborating with ecosystem partners to enhance packaging processes, fiber connection methods, and liquid cooling designs [2][9] Group 3: Industry Insights - China Galaxy indicates that the optical communication sector is currently undervalued, with demand continuing to grow. Capacity constraints are expected to be a major bottleneck for the development of optical communication until 2026. The market share of Chinese optical module manufacturers is leading globally and is expected to continue increasing [3][10] - The transition from "Made in China" to "Intelligent Manufacturing in China" is anticipated to be prominently reflected in the optical communication sector, with ample opportunities for upstream penetration during the 14th Five-Year Plan period [3][10] Group 4: Financing Activities - As of February 6, a total of 10 CPO concept stocks had net financing purchases exceeding 100 million yuan, with Tianfu Communication and Zhongji Xuchuang having net financing purchases of 2.542 billion yuan and 2.21 billion yuan, respectively [4][11] - Tianfu Communication has disclosed that its 1.6T optical engine product has entered normal delivery stages and is actively coordinating supply chain resources to enhance capacity [4][11] Group 5: Company Performance - NewEase is benefiting from the continuous growth in computing power investments, with a significant increase in sales revenue and net profit compared to the same period last year. The company expects to achieve a net profit attributable to shareholders of 9.4 billion to 9.9 billion yuan in 2025, representing a year-on-year growth of 231.24% to 248.86% [6][13] - NewEase has established a technology system covering various interconnection forms, including traditional pluggable optical modules, LPO, NPO, and CPO. The company has the capability for mass production and delivery of optical modules at 800G and above, successfully producing and delivering the latest generation of 1.6T optical module products while actively advancing the development of next-generation products [7][13]

风险提示：提及个股仅用于行业事件分析，不构成任何个股推荐或投资建议。指数等短期涨跌仅供参 考，不代表其未来表现，亦不构成对基金业绩的承诺或保证。观点可能随市场环境变化而调整，不构成 投资建议或承诺。提及基金风险收益特征各不相同，敬请投资者仔细阅读基金法律文件，充分了解产品 要素、风险等级及收益分配原则，选择与自身风险承受能力匹配的产品，谨慎投资。 （文章来源：每日经济新闻） 半导体设备ETF（159516）跟踪的是半导体材料设备指数（931743），该指数聚焦于半导体产业链中的 材料与设备环节，成分股涵盖半导体制造、封装测试等过程中所需的材料和设备供应商，以反映半导体 产业上游关键领域的市场表现。 华鑫证券指出，AI算力需求持续旺盛，推动电子与半导体行业相关环节发展。为满足Meta、谷歌、微 软等科技巨头持续增长的需求，光纤光缆作为算力网络的物理底座，其需求刚性与成长持续性获双重验 证。在AI模型变大、推理实时化背景下，共封装光学（CPO）技术成为AI算力规模化扩张的核心支 撑。从PCB行业看，下游AI算力需求旺盛带动AI-PCB需求提升，行业从2024年初开始逐步复苏，2025 年整体实现了较为稳定的增长。 ...

Core Insights - The year 2025 is marked by significant technological advancements, particularly in artificial intelligence and commercial space, reshaping economic and social development [2] - The technological wave is driving structural changes in capital markets, with emerging trends providing key investment directions [2] Group 1: Annual Technology Buzzwords - Each buzzword corresponds to a rising industry chain, reflecting market trends [4] Group 2: AI Agents - The strategic open-sourcing by models like DeepSeek has lowered barriers for AI development, leading to a surge in innovation and a shift in market expectations for Chinese tech [6] Group 3: Supernodes - Supernode architecture integrates hundreds of chips into efficient computing units, providing a differentiated development path in global computing power competition [8] Group 4: PCB Industry - The high-end PCB industry is evolving towards higher layers, density, and reliability, becoming a cornerstone for electronic information innovation [10] Group 5: AI Computing Power - AI computing power is now a strategic digital foundation in international tech competition, with China's "East Data West Computing" strategy accelerating [12] Group 6: Co-Packaged Optics (CPO) - CPO technology addresses challenges in data center expansion, becoming crucial for enhancing computing efficiency and reshaping value distribution in the optical communication industry [14] Group 7: Digital Economy - The digital economy in China is transitioning to a strategic phase, with AI and data driving efficiency and value creation across various sectors [16] Group 8: Controlled Nuclear Fusion - Significant breakthroughs in controlled nuclear fusion are anticipated, with China making strides in this area, attracting capital market attention [18] Group 9: Commercial Space - China's commercial space sector is entering a phase of rapid development, with a focus on satellite internet and a burgeoning trillion-dollar space economy [20] Group 10: Humanoid Robots - The humanoid robot industry is reaching a critical point of mass production, with significant events indicating a shift towards large-scale applications [22] Group 11: Quantum Computing - Quantum computing is accelerating from research to engineering exploration, with major economies prioritizing it in their tech strategies [24] Group 12: 2026 Outlook - The technological wave is expected to continue, with a focus on AI integration, independent innovation, and future industries [26] - Investment opportunities are anticipated in areas such as AI applications, commercial space, and defense industries [28][30]

Core Insights - Marvell's Q1 FY26 financial performance showed total revenue of $1.82 billion, exceeding consensus estimates by 0.89% [1] - The company reported a net income of $180 million, reflecting a year-over-year increase of 151.0% [1] - The gross margin for Q1 FY26 was 50.6%, slightly above the expected 48.3% [1] Financial Performance - Total Revenue: $1.82 billion, up 4.3% quarter-over-quarter and 19.9% year-over-year [1] - Gross Profit: $950 million with a gross margin of 50.6% [1] - R&D expenses were $510 million, accounting for 26.9% of revenue [1] - SG&A expenses were $190 million, representing 10.8% of revenue [1] - Net Income: $180 million, with a net profit margin of 9.4% [1] Q2 Guidance - Revenue is expected to be around $2 billion, with a variance of ±5% [2] - GAAP gross margin is projected to be between 50% and 51%, while Non-GAAP gross margin is expected to be between 59% and 60% [3] - GAAP diluted EPS is forecasted to be between $0.16 and $0.26, and Non-GAAP diluted EPS is expected to be between $0.62 and $0.72 [3] Business Dynamics - Marvell plans to sell its automotive Ethernet business to Infineon for $2.5 billion, expected to close in 2025, enhancing capital allocation flexibility [6] - The data center market showed strong performance with Q1 revenue of $1.44 billion, a 5% increase quarter-over-quarter and a 76% increase year-over-year [7] - The company is optimistic about the long-term potential of the data center business, driven by capital expenditures from hyperscale enterprises and sovereign data demands [7] Technology Developments - Marvell is focusing on scaling AI chip production and optical product shipments [8] - The company is developing custom HBM and Co-Packaged Optics technologies to optimize AI accelerator performance [9] - Collaboration with NVIDIA to introduce NVLink Fusion technology is underway, supporting custom XPU projects [10] Market Performance - The enterprise networking and carrier infrastructure market reported combined revenue of $306 million in Q1, with a 14% quarter-over-quarter increase [11] - The automotive and industrial market had Q1 revenue of $76 million, down 12% quarter-over-quarter [14] Future Outlook - AI has become a significant portion of new business in the data center segment, with custom projects like XPU expected to drive revenue growth [16] - A custom AI investor event is scheduled for June 17, focusing on market opportunities and technology platforms [17]

Core Viewpoint - Nvidia's innovations in AI inference technologies, including the introduction of inference Token expansion, inference stack, Dynamo technology, and Co-Packaged Optics (CPO), are expected to significantly reduce the total cost of ownership for AI systems, thereby solidifying Nvidia's leading position in the global AI ecosystem [2][4][68]. Group 1: Inference Token Expansion - The rapid advancement of AI models has accelerated, with improvements in the last six months surpassing those of the previous six months. This trend is driven by three expansion laws: pre-training, post-training, and inference-time expansion [8]. - Nvidia aims to achieve a 35-fold improvement in inference cost efficiency, supporting model training and deployment [10]. - As AI costs decrease, the demand for AI capabilities is expected to increase, demonstrating the classic example of Jevons Paradox [10][11]. Group 2: Innovations in Hardware and Software - Nvidia's new mathematical rules introduced by CEO Jensen Huang include metrics for FLOPs sparsity, bidirectional bandwidth measurement, and a new method for counting GPU chips based on the number of chips in a package [15][16]. - The Blackwell Ultra B300 and Rubin series showcase significant performance improvements, with the B300 achieving over 50% enhancement in FP4 FLOPs density and maintaining an 8 TB/s bandwidth [20][26]. - The introduction of the inference stack and Dynamo technology is expected to greatly enhance inference throughput and efficiency, with improvements in smart routing, GPU planning, and communication algorithms [53][56]. Group 3: Co-Packaged Optics (CPO) Technology - CPO technology is anticipated to significantly lower power consumption and improve network scalability by allowing for a flatter network structure, which can lead to up to 12% power savings in large deployments [75][76]. - Nvidia's CPO solutions are expected to enhance the number of GPUs that can be interconnected, paving the way for networks exceeding 576 GPUs [77]. Group 4: Cost Reduction and Market Position - Nvidia's advancements have led to a performance increase of 68 times and a cost reduction of 87% compared to previous generations, with the Rubin series projected to achieve a 900-fold performance increase and a 99.97% cost reduction [69]. - The overall trend indicates that as Nvidia continues to innovate, it will maintain a competitive edge over rivals, reinforcing its position as a leader in the AI hardware market [80].