证券研究报告 [Table_Title] 通信行业 GTC&OFC 前瞻 前次评级 买入 报告日期 2026-03-09 [Table_Summary] 核心观点： [Table_PicQuote] 相对市场表现 [Table_Page] 投资策略周报|通信 识别风险，发现价值 请务必阅读末页的免责声明 1 / 15 [Table_Grade] 行业评级 买入 [Table_Contacts] -20% 3% 26% 48% 71% 94% 03/25 05/25 08/25 10/25 12/25 03/26 通信 沪深300 | [分析师： Table_Author]韩东 | | | --- | --- | | | SAC 执证号：S0260523050005 | | 021-38003776 | | | | gfhandong@gf.com.cn | | 分析师： | 王昊 | | | SAC 执证号：S0260525030001 | | 021-38003541 | | | | shwanghao@gf.com.cn | | 分析师： | 王亮 | | | SAC 执证号：S0260519060001 ...

Billionaire Leo KoGuan doubled his stake in Nvidia to 2 million shares, boosting his investment in the chipmaker at a time when the war in the Middle East has sparked a global selloff of everything from bonds to equities https://t.co/T371FYcwQs ...

Core Viewpoint - NVIDIA plans to restart production of the GeForce RTX 3060 graphics card to address the current shortage in consumer-grade graphics cards, balancing its AI and consumer business needs [1][4]. Group 1: Reasons for Restarting Production - The primary reason for the restart is the global DRAM memory shortage and the intense competition for semiconductor resources from the AI industry [1]. - The DRAM shortage has significantly impacted the GPU industry, disrupting product release schedules and worsening the supply of popular graphics card models [1]. Group 2: Role of Samsung - Samsung, as NVIDIA's core foundry partner for Ampere architecture graphics cards, is prepared to support the restart of RTX 3060 production with its 8nm production line [4]. - Samsung has a history of providing 8nm process support for NVIDIA, which facilitates a smooth ramp-up of production without technical obstacles [4]. Group 3: Production Strategy and Technology - NVIDIA may adopt a "soft restart" strategy for the RTX 3060, gradually releasing products into the market without a public announcement [4]. - The RTX 3060 is expected to incorporate the latest DLSS 4.5 technology, enhancing gaming performance by optimizing graphics quality and reducing ghosting [4]. Group 4: Future Product Plans - In addition to reviving the RTX 3060, NVIDIA is rumored to be planning the release of a new entry-level RTX 5050 graphics card with 9GB of memory to diversify its product lineup [5]. - These initiatives are seen as pragmatic choices to quickly alleviate supply shortages while maintaining market share amid increasing competition for AI resources [5].

格隆汇3月9日｜据台湾经济日报，英伟达首席执行官黄仁勋表示，内存供给短缺，对英伟达是极好的消 息，可以让客户直接选择性能最强的解决方案。他还表示，存储器厂产能扩多少，英伟达就会用多少。 ...

Nvidia (NVDA 2.94%) has been the staple of artificial intelligence (AI) investing since it became mainstream in 2023. However, its stock has remained relatively dormant since August 2025. The stock is up about 5% since then, and there have been several other AI investments that have taken off in that same time frame. One of the best stocks to bet on since Nvidia quit rising is Micron Technology (MU 6.68%). In the same time frame that Nvidia rose a mere 5%, Micron's stock is up nearly 300%. With returns like ...

3月9日，据《韩国经济日报》独家报道，英伟达下一代AI加速器Vera Rubin的核心供应链悬念正式揭 晓。作为这款重磅产品的"算力基石"，第六代高带宽内存HBM4的供应商名单尘埃落定，三星电子与SK 海力士成功突围，携手拿下核心供应资格，而全球第三大内存厂商美光则意外暂别顶级供应链，半导体 行业的高端内存争霸赛迎来关键转折点。 早在3月8日，半导体行业就已传出核心消息：三星与SK海力士已正式跻身Vera Rubin核心零部件供应商 阵营，成为HBM4的暂定供应商。对英伟达而言，HBM4的重要性不言而喻——自去年起，该公司便将 其定义为Vera Rubin能否成功的核心关键，毕竟HBM作为AI芯片的"数据高速公路"，直接决定GPU算力 的释放效率，没有高带宽支撑，再强大的芯片也会陷入"数据饥饿"的尴尬境地。为此，英伟达提出了远 超行业标准的严苛要求，不仅将运行速度门槛提升至10Gb/s以上，大幅超越JEDEC规范的8Gb/s标准， 更在容量上实现突破，最终确定Vera Rubin将搭载16颗HBM4，总容量高达576GB，较AMD下一代AI加 速器MI450的432GB领先33%，尽显其在AI算力领域的绝对竞 ...

Summary of Key Points from the Conference Call Industry and Company Overview - The conference focused on the technology hardware industry, particularly memory storage, CPU developments, and AI-related hardware such as optical modules and liquid cooling systems [1][2][9]. Core Insights and Arguments Memory Storage Industry - Recent fluctuations in memory prices have been influenced by geopolitical factors, particularly affecting South Korean memory stocks [2][10]. - Despite recent volatility, the memory shortage is expected to persist until 2028, with HBM4 technology led by Samsung maintaining a strong market position [2][10]. - Contract prices in the memory sector are anticipated to increase by 30%-50% quarter-over-quarter in Q2, with demand visibility extended through long-term agreements (LTA) until 2028 [2][10][11]. - Samsung Electronics is recommended as a top stock pick due to its competitive advantage in the HBM market and potential for capacity expansion [6][12]. CPU and Optical Engine Developments - The CPU market has gained significant attention, with NVIDIA and AMD being key players. NVIDIA's SpectrumX cabinets are expected to dominate data centers this year [2][7]. - Production of optical engines is projected to reach between 600,000 to 1,000,000 units, with demand for optical engines expected to hit 2 million units by 2027 [2][7][12]. - NVIDIA is set to release a new version of its Context Switch at the GTC event, with market growth for CLCPL switches projected to increase by 144% by 2030 [2][7][12]. AI Infrastructure and Cooling Technologies - The discussion highlighted advancements in AI infrastructure, particularly in liquid cooling and power technologies, with companies like Delta Electronics and Lite-On actively expanding to meet future demand [5][17]. - The introduction of high-voltage direct current (DC) power racks is anticipated to occur earlier than expected, with small-scale shipments of 800V DC products starting this year [5][17]. - Cooling stocks have regained market focus due to increased demand from AI applications, with expectations for significant growth in water cooling board usage [5][18]. Additional Important Insights - The relationship between CPUs and optical modules is evolving, with the market now recognizing the potential risks associated with CPU impacts on optical module investments [3][14]. - The optical module market is projected to grow from $18 billion to $50 billion between 2025 and 2028, indicating substantial growth potential [5][15]. - Investment recommendations include companies within the optical module supply chain such as Coherent in North America and Mark and VPEC in Taiwan, as well as TSMC in the CPU sector [5][16]. Conclusion - The conference provided a comprehensive analysis of the technology hardware sector, emphasizing the interconnectedness of memory storage, CPU advancements, and AI infrastructure. The insights shared indicate a positive outlook for specific stocks and sectors, driven by ongoing technological advancements and market demand.

最近半年的PC DIY市场只能用"惨淡"来形容，在闪存价格暴涨的影响下，不管是内存还是显卡都在涨价，很多朋友估计都和小雷一样暂缓或取消了升级电 脑的计划。 电脑升级是别想了，那么玩家们缺的游戏性能这块谁来给补上呢？没关系，微软出手了——玩家们苦等的DirectX 12 SER终于完成技术验证，正式加入DXR 1.2规范的大家庭。 性能最高暴涨90%，SER到底是什么？ SER，全名Shader Execution Reordering（着色器执行重排序），这项技术最初由英伟达研发，随后被微软相中并纳入DirectX 12的开放规范中，其主要作用 是通过对光追模式下的GPU计算线程进行重新排列，归纳相似的工作负载，让GPU的并行计算能力得到充分释放。 图源：微软 图源：Gemini 简单来说，就是在原料下产线前加了个分拣员，把零件归纳整理并交给对应的产线，让GPU又回到自己熟悉且高效的运行状态中。事实上，英伟达在RTX 20系列上更新光追功能后，很快就意识到光追的瓶颈在于调度问题，所以才花了几年时间折腾出个SER。 然后，微软就把SER"招安"了，毕竟这么好的东西还是不要藏着捏着了，分享给大家不好吗？彼时的英伟 ...

来源：环球网 【环球网科技综合报道】3月9日消息，据 外媒Wccftech报道，英伟达计划复产旧款显卡GeForce RTX 3060，三星晶圆厂正筹备重启该型号的8纳米生产线，以此缓解消费级显卡供应短缺问题，兼顾AI与消 费级业务平衡。 据悉，三星曾是英伟达安培架构显卡的核心代工伙伴，2021年便为RTX 3060提供8纳米工艺支持。目前 三星仍在推进8纳米产能，此前已承接英伟达为任天堂Switch 2定制的芯片订单，复产RTX 3060的产能 衔接无压力。 英伟达计划以低调方式重启该型号，通过搭载DLSS 4.5技术提升竞争力——该技术可借助第二代 Transformer模型优化画质、减少残影，大幅提升游戏流畅度，以此打动刚需玩家。此外，有传闻称英伟 达还计划推出9GB内存的RTX 5050，而非发布全新显卡系列。 外媒报道称，此次复产背后，是AI行业抢占大量DRAM内存和半导体资源，英伟达难以兼顾全新消费 级显卡研发与供应，复产旧款成为高效解决方案，既缓解短缺，也契合当前市场需求。（旺旺） ...

Core Viewpoint - Rising oil prices, potentially reaching $150 per barrel, may not directly impact Nvidia's business fundamentals but could affect its stock valuation due to inflation fears and multiple compression [1] Group 1: Company Overview - Nvidia trades at $177.82 with a trailing earnings multiple of 36x and an analyst target price of $265.18 [1] - The company has provided Q1 FY2027 revenue guidance of $78 billion, driven by strong demand for AI compute from hyperscalers [1] - Nvidia's capital expenditures for FY2026 were $6.04 billion against $215.94 billion in revenue, indicating a capital intensity ratio favorable compared to many industrial companies [1] Group 2: Impact of Oil Prices - While oil prices do not directly affect Nvidia's operations, a significant rise to $150 could lead to inflation, prompting the Federal Reserve to maintain higher interest rates, which would compress valuations of high-growth stocks like Nvidia [1] - The prediction markets currently assign a 46.5% probability that Nvidia's stock will close above $180 by the end of March, reflecting market uncertainty rather than fundamental deterioration [1] Group 3: Business Model Insights - Nvidia operates as a fabless semiconductor company, outsourcing manufacturing to TSMC, with primary costs focused on R&D and compensation rather than energy inputs [1] - Gross margins have improved from 71.3% in Q1 FY2026 to 75.2% in Q4 FY2026, indicating a business model more akin to a software platform than a traditional manufacturing entity [1] - The demand for AI infrastructure is described as structural, largely unaffected by fluctuations in crude oil prices [1]