英特尔以18A制程开启全新AI计算时代，其首个基于18A制程的Panther Lake架构性能较前代提升50%，能耗降低约30%。Panther Lake芯片将于 年底量产，2026年上市。对于英特尔来说，这不仅是一款新芯片的亮相，也打响了其重夺先进制程芯片竞赛主动权的「关键一战」。 刚刚，英特尔在官网公布了其下一代客户端处理器——Intel Core Ultra 3（代号Panther Lake）的架构细节。 Panther Lake是全球首个基于Intel 18A（约1.8nm）工艺的AI PC平台。 英特尔计划今年启动Panther Lake量产，首批产品将于年底出货，预计2026年1月全面上市。 18A（约1.8纳米）是英特尔首个2纳米级别制程节点，它代表了半导体行业的两项重大创新：全环绕栅极晶体管（Gate-All-Around）和背面供电网络 （Backside Power Delivery Network）。 与Intel 3制程工艺相比，其每瓦性能提升达15%，芯片密度提升约30%。 据英特尔官方数据，与Lunar Lake相比，新一代芯片在相同功耗下性能提升可达 50%；其功耗在性能一致情况 ...

一、行情 9月30日，存储概念再度大涨，德明利、诚邦股份、深科技等涨停，江波龙、天山电子等涨幅超10%。 1）隔夜美股存储个股继续大涨，闪迪大涨16.87%，西部数据涨9.23%，希捷涨5.35%，美光涨4.22%。 2）摩根大通认为，这一轮周期的核心驱动力来自AI计算对高性能内存的巨大需求，其影响已从高端的高带宽内存 （HBM）迅速扩展至传统DRAM和NAND闪存。市场动态显示，供应商在未来12个月内可能难以满足全部需求，从而 支撑价格持续走强。 在AI计算对高性能内存的巨大需求的推动下，"内存饥渴"趋势正在推动整个行业进入结构性增长阶段，DRAM市场正 迎来一个从2024年持续至2027年的"前所未有的四年定价上行周期"，预计到2027年，全球存储市场规模将达到近3000 亿美元。 3）CFM闪存市场近日发布2025年Q4存储市场展望报告。报告指出，预计四季度，服务器eSSD涨幅将达到10%以上， DDR5 RDIMM价格涨幅约10%~15%。 4）据证券时报援引台湾工商时报消息，继闪迪、美光、三星及西部数据宣布涨价之后，存储模组大厂威刚在日前宣 布，自29日起停止DDR4报价，DDR5与NAND闪存优先 ...

Core Insights - Google has begun selling its Tensor Processing Units (TPUs) to cloud service providers, aiming to compete directly with NVIDIA in the AI computing market, which is projected to be worth trillions of dollars [4][6][7] - The competition between Google and NVIDIA is intensifying, with analysts predicting a significant decline in NVIDIA's GPU sales due to the rise of TPUs [7][19] - Google's TPUs are designed specifically for AI computing, offering a cost-effective and energy-efficient alternative to traditional GPUs, with reported costs being one-fifth of those for GPUs used by OpenAI [11][12] Google TPU Development - Google initiated discussions about deploying specialized hardware in its data centers as early as 2006, but the project gained momentum in 2013 due to increasing computational demands [9][10] - The TPU architecture focuses on high matrix multiplication throughput and energy efficiency, utilizing a "Systolic Array" design to optimize data flow and processing speed [10][11] - Over the years, Google has released multiple generations of TPUs, with the latest, Ironwood, achieving peak performance of 4614 TFLOPs and supporting advanced computing formats [15][16] Market Position and Future Outlook - By 2025, Google is expected to ship 2.5 million TPUs, with a significant portion being the v5 series, indicating strong market demand [15] - Analysts suggest that Google's TPUs could become a viable alternative to NVIDIA's offerings, with a notable increase in developer activity around Google Cloud TPUs [19] - The competitive landscape is evolving, with other companies like Meta and Microsoft also developing their own ASIC chips, further challenging NVIDIA's dominance in the market [23][25]

Group 1 - The core viewpoint of the news is that Moole Technology's IPO application has been approved by the Shanghai Stock Exchange, marking a significant step towards its listing on the A-share market [1] - Moole Technology plans to raise 8 billion RMB through its IPO, which is the largest fundraising project in the A-share market this year and also the largest IPO in the semiconductor sector [1] - The funds raised will be allocated to the development of new AI training and inference chips, new graphics chips, and new AI SoC chips, as well as to supplement working capital [1] Group 2 - Moole Technology, established in 2020, is a domestic GPU company that focuses on providing accelerated computing infrastructure and solutions for digital transformation across various industries [1] - The company is recognized as a "unicorn" in the industry and is often referred to as the "Chinese version of NVIDIA" due to its capabilities in both graphics rendering and AI computing [1] - The news of Moole Technology's IPO has led to a surge in the stock prices of related companies, particularly Heertai, which has seen its stock price rise significantly this year [2]

Core Viewpoint - The "memory hunger" trend driven by AI computing is pushing the entire industry into a structural growth phase, leading to an unprecedented four-year pricing upcycle in the DRAM market from 2024 to 2027, with the global storage market expected to reach nearly $300 billion by 2027 [1][2]. Group 1: Market Dynamics - The core driver of the current cycle is the immense demand for high-performance memory due to AI computing, which has rapidly expanded from high-bandwidth memory (HBM) to traditional DRAM and NAND flash [2][3]. - Suppliers may struggle to meet the entire demand in the next 12 months, supporting sustained price increases [2]. - The NAND flash market is also experiencing a strong rebound, partly due to a severe shortage of hard disk drives (HDD), pushing customers towards enterprise solid-state drives (eSSD) [2][9]. Group 2: DRAM Market Outlook - The DRAM market is entering an unprecedented four-year pricing upcycle from 2024 to 2027, breaking the previous pattern of volatility [3]. - By 2027, HBM is expected to account for 43% of the total value in the DRAM market, stabilizing traditional DRAM price fluctuations and enhancing overall market profitability [3]. - Even with potential price drops for HBM3E products in 2026 due to increased supply, the next-generation HBM4 is anticipated to maintain a premium of about 35%, keeping the average selling price (ASP) stable [3]. Group 3: AI-Driven Demand - The rise of AI inference, AI services, and edge AI applications necessitates upgrades in memory configurations for general servers to achieve lower latency and better power efficiency [8]. - By 2027, AI-related applications are projected to account for 53% of the DRAM market's total addressable market (TAM) [8]. Group 4: NAND Market Recovery - The NAND flash market is experiencing a strong pricing recovery after two years of underinvestment, driven by a surge in demand for eSSD [9]. - The shortage of traditional HDDs, with delivery cycles extending up to 52 weeks, is prompting a significant shift towards eSSD in nearline storage markets [9]. - The structural importance of NAND is increasing as AI models transition from training to inference applications, raising demands for data read speeds and latency [10].

2025 年 09 月 25 日 研 究 报 告 英伟达（NVDA.O）：与英特尔合作并投资，巩固 AI 计算领域核心地位 推荐(维持) 事件 | 分析师：任春阳 | S1050521110006 | | --- | --- | | rency@cfsc.com.cn | | | 联系人：谢孟津 | S1050123110012 | | xiemj@cfsc.com.cn | | 行业相对表现 | 表现 | 1M | 3M | 12M | | --- | --- | --- | --- | | 计算机(申万) | -2.7 | 17.9 | 78.5 | | 沪深 300 | 2.2 | 15.3 | 34.2 | 市场表现 -40 -20 0 20 40 60 80 (%) 计算机 沪深300 资料来源：Wind，华鑫证券研究 相关研究 1 、 《 计 算 机 行 业 点 评 报 告 ： Waymo：运营扩张提速，技术迭代领 先，政策环境助力规模化落地》 2025-09-24 2、《计算机行业点评报告：文远知 行（WRD.0）：产品矩阵与全球战略 协同》2025-09-24 3、《计算机行业点评报告：英伟达 ...

Core Viewpoint - Nvidia announced a $5 billion acquisition of Intel common stock at $23.28 per share, making it one of Intel's largest shareholders with an expected ownership of over 4% [2] - Following the announcement, Intel's stock price surged over 30% in pre-market trading, marking its largest single-day increase in recent years [2] - Nvidia and Intel will collaborate to develop customized CPUs for data center platforms and integrate Nvidia's GPUs with Intel's CPUs, signaling a significant partnership in the AI and accelerated computing sectors [2] Group 1: Nvidia and Intel Collaboration - The partnership is described as a fusion of two world-class platforms, emphasizing that it is not merely a financial investment but a strategic alliance [2] - Nvidia's CEO highlighted the importance of this collaboration in driving advancements in AI and accelerated computing [2] - The integration of Nvidia's AI technology with Intel's x86 ecosystem is expected to open new application scenarios for Nvidia's CUDA platform [5] Group 2: Market Context - Intel has historically dominated the GPU market with over 60% market share, primarily due to its integrated GPU chips in desktop and laptop computers [3] - However, in the independent GPU market, Nvidia and AMD have captured most of the market share, leaving Intel with a minimal presence [5] - The collaboration is viewed as a pivotal moment for Intel's "second startup," potentially allowing it to catch up in the AI chip market [5] Group 3: Industry Trends - The global GPU chip industry is recognized as a vital and strategic sector, with applications in gaming, AI, data centers, and autonomous driving [2] - The semiconductor industry is shifting from a "arms race" to "ecosystem building," as evidenced by alliances like the Nvidia-Intel partnership and TSMC's wafer foundry collaborations [5]

Core Viewpoint - The unexpected collaboration between Nvidia and Intel marks a significant shift in the semiconductor industry, highlighting Nvidia's dominance in AI computing and Intel's need for strategic partnerships to recover from its declining market position [5][11][14]. Group 1: Nvidia's Investment in Intel - Nvidia announced an investment of $5.9 billion to acquire approximately 4% of Intel's shares at a price of $23.28 per share, pending regulatory approval [5]. - Following the announcement, Intel's stock price surged by 30%, significantly increasing the value of shares held by the U.S. government by $4.9 billion, bringing the total value to approximately $14 billion [5]. - The collaboration aims to integrate Nvidia's AI and accelerated computing capabilities with Intel's CPU technology, creating advanced solutions for data centers and personal computing [6]. Group 2: Impact on Competitors - AMD's stock fell by 5.9% in response to the news, as the collaboration poses a competitive threat in both the PC and data center markets [8][12]. - Analysts expressed skepticism about the true intentions behind the Nvidia-Intel partnership, questioning whether it is a superficial gesture or the beginning of a more extensive collaboration [8]. - AMD remains committed to its strategy of capturing market share from Intel, emphasizing AI as a primary focus [8]. Group 3: Nvidia's Industry Leadership - Nvidia's rise to prominence is attributed to its early recognition of the need for new chips and software in AI computing, positioning itself ahead of competitors [13]. - Nvidia's market capitalization has surpassed $4 trillion, while Intel's stands at $116 billion, illustrating the dramatic shift in industry dynamics [13]. - The rapid change in market positions highlights the volatility and competitive nature of the semiconductor industry, with Nvidia's sales projected to reach $200 billion this year [13]. Group 4: Intel's Strategic Shift - Intel's current predicament is a result of its failure to capitalize on the AI boom and a loss of manufacturing advantages, now relying on TSMC for top-tier chip production [18]. - The appointment of a new CEO, who advocates for an open strategy and collaboration, marks a pivotal change in Intel's approach to regain competitiveness [15][17]. - Intel's ongoing challenges include reliance on external investments and asset sales to fund operations amid declining market share [17].

来源：中国证券报-中证网 9月19日，上交所网站显示，上交所上市审核委员会定于9月26日召开2025年第40次上市审核委员会审议会议， 审议摩尔线程智能科技（北京）股份有限公司的首发事项。 摩尔线程本次科创板上市拟募集资金80亿元。 2025年6月18日，证监会网站披露了摩尔线程首次公开发行股票并上市辅导工作完成报告。6月30日，摩尔线程 科创板IPO申请获上交所受理。9月5日和9月18日，摩尔线程关于首轮和第二轮审核问询函的回复挂网。 资料显示，摩尔线程成立于2020年6月，实际控制人张建中控制公司44.07%股份。摩尔线程成立后曾获得五源 资本、红杉中国、联想创投、腾讯、深创投、字节跳动、小马智行、中国移动旗下产业基金等机构投资。 净利润方面，摩尔线程在2022年至2024年分别录得归母净亏损约18.4亿元、16.73亿元、14.92亿元，2025年上 半年归母净亏损约2.71亿元。摩尔线程表示，根据公司预测，公司最早可于2027年实现合并报表盈利。公司营 业收入持续增长，2022年至2024年，公司营业收入由0.46亿元增长至4.38亿元，复合增长率为208.44%；2025 年上半年，公司营业收入为7 ...

| 序号 | 项目名称 | 项目实 | 募投项目 | 募集资金 拟投入金 | 项目备案编号 | 项目环保 | | --- | --- | --- | --- | --- | --- | --- | | | | 施单位 | 投资总额 | | | 批文号 | | | | | | 额 | | | | | 摩尔线程新一 代自主可控 AI 训推一体芯片 | 摩尔线 程 | 250.957.98 | 250.957.98 | 京朝科信局备 (2025) 17 号 | 不适用 | | | 研发项目 | | | | | | | 2 | 摩尔线程新一 代自主可控图 形芯片研发项 | 壓尔线 程 | 250.233.23 | 250.233.23 | 京朝科信局备 (2025) 16 号 | 不适用 | | | 目 | | | | | | | 3 | 摩尔线程新一 代自主可控 AI SoC 芯片研发 | 无锡壓 尔 | 198.180.33 | 198.180.33 | 惠数投备 (2025) 255 | 不适用 | | | | | | | 름 | | | | 项目 | | | | | | | 4 | 补充流动资金 | l | ...