证券研究报告 行业研究 / 行业点评 2025 年 12 月 19 日 强于大市 投资要点： 一年内行业指数与沪深 300 指数对比走势： 资料来源：聚源数据，爱建证券研究所 相关研究 《电子行业周报：NVIDIA H200 芯片放松出口 限制》2025-12-16 《爱建电子专题报告：iPhone 折叠屏有望带来 产业发展拐点》2025-12-15 《电子行业周报：字节跳动发布豆包手机助手》 2025-12-08 《"朱雀三号"首飞在即，推动卫星通信行业 发展》2025-12-01 《电子行业周报：Google 发布第三代旗舰人工 智能模型 Gemini 3》2025-11-25 许亮 S0820525010002 0755-83562506 xuliang@ajzq.com 电子 ——人工智能月度跟踪 风险提示：国际贸易摩擦加剧、下游需求不及预期、技术升级进度滞后。 事件：12 月 3 日，摩尔线程披露公告称，经上海证券交易所审核同意，公司发行的人民币 普通股股票已于 2025 年 12 月 5 日在科创板正式挂牌上市；12 月 17 日，沐曦股份成为继 摩尔线程之后第二家实现 A 股上市的国产 GPU ...

Why are we interested in the stocks that hedge funds pile into? The reason is simple: our research has shown that we can outperform the market by imitating the top stock picks of the best hedge funds. Our quarterly newsletter’s strategy selects 14 small-cap and large-cap stocks every quarter and has returned 427.7% since May 2014, beating its benchmark by 264 percentage points ( see more details here ).To identify the 10 most undervalued stocks to invest in, we first screened all semiconductor companies wit ...

Group 1: EUV Lithography and Market Impact - The existence and capabilities of a new prototype in EUV lithography remain unverified, but market reactions indicate investor seriousness [1] - ASML currently produces EUV machines, and export restrictions have prevented China from acquiring them, leading to Chinese manufacturers developing their own [2] - China's potential entry into the EUV lithography market could disrupt existing chip manufacturers and EUV machine suppliers, allowing China to sell machines and chips at significantly lower prices [2] Group 2: Economic Conditions in China - China's economic headwinds challenge its 5% GDP growth target for 2025, with a notable decline in domestic demand observed in November [5] - The housing crisis continues to impact consumer sentiment, with house prices falling 2.4% year-on-year in November, following a 2.2% drop in October [7] - Lawmakers have pledged fiscal and monetary support to stabilize the labor market and boost domestic demand, emphasizing the need for effective measures to achieve GDP growth targets [6][8] Group 3: Market Performance and Outlook - The Hang Seng Index has decreased by 5.44% in Q4 but remains up 26.54% year-to-date, while the CSI 300 has gained 15.64% year-to-date despite a modest decline in the current quarter [10] - The Nasdaq Composite Index has seen a slight increase of 0.15% in Q4, remaining close to its all-time high [11] - Technical indicators suggest a bearish near-term outlook for the CSI 300, with critical support levels identified at 4,500 and 4,365 [12][13]

Core Insights - RLWRLD, a Physical AI startup, won first place in the Foundation Models, Robot Brains and Runtime category at the 2025 Nebius Robotics & Physical AI Awards, highlighting its advancements in next-generation robotics and Physical AI [2][3][5] Group 1: Awards and Recognition - The Nebius Robotics & Physical AI Awards, hosted by Nebius and supported by NVIDIA, awarded $1.5 million in AI cloud compute and inference credits to startups from over 250 applicants across 60+ countries [3][5] - RLWRLD received $150,000 in Nebius AI Cloud compute and inference credits as the first-place winner, emphasizing its innovative contributions to the field [5] Group 2: Company Focus and Technology - RLWRLD specializes in developing robot foundation models for high-difficulty manipulation and reliable industrial deployment, focusing on practical applications rather than general-purpose robotics [6][8] - The company's core platform, RLDX, is a proprietary robotics foundation model that controls a 15-degree-of-freedom robotic hand, enabling tasks that require human-level dexterity [6] Group 3: Commercial Readiness and Partnerships - Judges recognized RLWRLD's use of high-fidelity, multimodal industrial data and its collaboration with industry partners as indicators of commercial readiness, particularly in manufacturing and logistics [7] - Following the award, RLWRLD plans to expand global partnerships and accelerate the commercial deployment of its foundation models, focusing on manufacturing and logistics environments in the U.S. and Asia [8] Group 4: Company Background - Founded in 2024, RLWRLD develops robot foundation models that enable machines to operate with human-like dexterity and cognition, utilizing high-precision 4D+ multimodal industrial data [9] - The company operates in the United States, South Korea, and Japan, and is working with major manufacturing partners to establish itself as a global leader in industrial robotics AI [9]

Market Overview - The main board saw significant gains with companies like Duolun Technology (+10.04%), Aerospace Information (+10.02%), and Datang Telecom (+10.00%) leading the charge [1] - The ChiNext board also performed well, with Wanjie Technology (+20.02%) and Jiuzhiyang (+20.00%) showing remarkable increases [1] - The Sci-Tech Innovation board was led by Jin Chengzi (+7.62%) and Zhenlei Technology (+6.73%) [1] - The SW vertical application software sector showed a slight increase of +0.11% [1] Domestic News - SuKOS Semiconductor successfully shipped advanced coating equipment for 4.5G glass-based TGV, significantly increasing chip production per run and reducing unit costs, marking a key step towards large-scale commercialization [1] - Jingjia Micro reported that its subsidiary Chengheng Micro has completed key stages for its AI SoC chip CH37 series, achieving a significant project milestone [1] - The Ministry of Industry and Information Technology granted conditional approval for L3-level autonomous vehicles from Changan Automobile and BAIC Blue Valley Magna [1] - Winbond Electronics announced a cumulative order of NT$2.332 billion for semiconductor equipment from ASML, to be used in production processes [1] Company Announcements - Xiechuang Data plans to purchase servers from multiple suppliers, with a total procurement amount not exceeding RMB 9 billion [3] - Aisino Security received a government subsidy of RMB 3 million to support its daily operations or specific projects [3] - Shanghai Steel Union's subsidiary received a government subsidy of RMB 26.1235 million [3] - Wolong Nuclear Materials announced plans to invest up to RMB 1 billion in overseas production bases in Vietnam and Malaysia to enhance global capacity and supply chain resilience [3] International News - India launched the DHRUV64, a fully indigenous microprocessor developed under the MDP framework, suitable for 5G infrastructure and IoT applications [1] - Counterpoint Research predicts a 2.1% decline in global smartphone shipments by 2026 due to rising component costs [1] - NVIDIA acquired SchedMD, a leading developer of the Slurm workload management system, to strengthen its open-source software ecosystem [1] - Samsung Display will supply OLED screens to Ferrari, enhancing the luxury car experience with customizable displays [1]

Investor Michael Burry has doubled down on his claims regarding GPU depreciation at leading hyperscalers and AI companies, citing recent remarks by Microsoft Corp. (NASDAQ:MSFT) CEO Satya Nadella, who highlighted the complex economics of large-scale data center investments. Burry Regrets Not Being More Vocal In 2008On Monday, in a post on X, Burry, who rose to prominence for shorting the housing market ahead of the 2008 financial crisis, said that he regrets not warning others more effectively about the ris ...

KUALA LUMPUR, Malaysia, Dec. 15, 2025 /PRNewswire/ -- GIBO Holdings Ltd. (NASDAQ: GIBO) today announced a strategic collaboration with Malaysia-based technology and infrastructure solutions provider E Total Technology Sdn Bhd, under which the parties will work together on the planning, site sourcing, feasibility studies, and deployment of next-generation artificial intelligence compute centers in Malaysia. The project is expected to incorporate NVIDIA's most advanced high-performance AI chips and computing ...

福莱新材表示，Apex Sensing的柔性传感器能够为机器人提供接近人类皮肤的细腻触觉感知，是实现高精度、高触觉能力"灵巧手"的关键核心元件。通过 此次峰会，公司成功建立了高质量潜在客户联系，锁定了未来战略合作项目，获得了第一手市场反馈。(厉平) 2025年12月11日至12日，硅谷知名人形机器人峰会Humanoids Summit在硅谷举行。作为全球人形机器人领域的顶级盛会，本次峰会汇聚了Agility Robotics、Apptronik、Boston Dynamics、Sanctuary等全球领先的开发商，并吸引了来自Meta、Amazon、NVIDIA、Google等"科技七巨头"的代表，参会专 业人士近2000名。 福莱新材（605488）全资拥有的美国公司Apex Sensing，作为本次峰会上唯一一家柔性传感器供应商参展。Apex Sensing充分利用此次宝贵的市场机遇， 不仅接待了众多的北美潜在客户，更与多家人形机器人产业链的领先供应商进行了深入的技术探讨及产品演示。公司吸引的客户代表覆盖范围极广，包 括：北美领先的人形机器人制造商，多个来自"科技七巨头"的代表，涵盖仿生手、工业自动化、动 ...

Global Economy & Central Banks - US and European futures are gaining ahead of central bank decisions [1][49] - The market expects the Bank of England to cut interest rates by 25 basis points [3][96] - The Bank of Japan is expected to move rates higher to 075% [26][96] - The US 10-year Treasury yield is at 417%, and the Euro-dollar is at 117% [4][50] - The market is pricing in slightly over two rate cuts by 2026 [92] Geopolitics & Security - Australia is considering tougher gun laws after a terrorist attack that killed 16 people [1][5][48][65] - Ukraine may drop its bid for NATO membership in return for Western security guarantees [1][48] - The EU is considering seizing or monetizing Russian assets to fund Ukraine [18] Market & Industry Trends - Asian stocks are under pressure, tracking losses from the US session, with concerns around AI investment [20][21] - China's retail sales are up 13% year-over-year, below estimates, indicating economic weakness [22] - China has built seven large petrochemical hubs, raising fears of oversupply and pressure on other nations [40] - European defense sector is interesting due to potential pullback on Russia-Ukraine discussions, but NATO demands higher defense spending [63][64] Artificial Intelligence - PolyAI raised $86 million in Series D funding for its conversational AI voice agents [70] - PolyAI's voice AI agents can handle tasks equivalent to over 1000 full-time employees for individual customers [73] - PolyAI claims to deliver approximately $10 million of value per client, totaling almost $15 billion for 250 customers [75][76] - PolyAI's annual recurring revenue was just over $20 million last year, with a goal to double this year [83]

Summary of Key Points from the Webinar on AI Data Center Memory Demand Industry Overview - The discussion centers around the U.S. IT Hardware industry, specifically focusing on AI data centers and their memory requirements [1][12]. - The webinar featured Gunjan Shah, a former Senior Cloud Engineer at Google, who provided insights into memory demand for AI workloads [1][12]. Core Insights Memory Demand in AI - Training AI models requires significantly more memory than inference, with medium-sized models consuming approximately 1TB of memory during training compared to much lower demands during inference [2][15]. - The rapid adoption of AI has led to a sharp increase in memory demand and prices, particularly for components like HBM (High Bandwidth Memory) and DRAM [3][21]. - Innovations in model architectures and memory technologies are expected to help manage memory demand sustainably in the long term [3][18]. Shift from HDDs to SSDs - Due to HDD shortages, many hyperscalers are transitioning to SSDs, which are 5 to 10 times more expensive but offer superior performance and lower operational costs [4][38]. - SSDs provide benefits such as reduced power consumption and minimal cooling requirements, contributing to a lower total cost of ownership (TCO) [4][40]. Emerging Memory Technologies - High Bandwidth Flash (HBF) is an emerging technology that aims to provide fast, non-volatile memory, potentially lowering energy consumption and cooling costs for AI inference workloads [5][18]. Investment Implications - Companies such as Seagate Technology (STX), Western Digital (WDC), SanDisk (SNDK), Samsung, SK Hynix, and Micron have been rated with specific price targets based on their performance in the memory market [7][8][9][10][11]. - STX is rated Outperform with a price target of $370, while WDC is rated Market-Perform with a target of $170 [8][9]. Additional Insights Memory Usage Breakdown - The memory footprint for training is heavily reliant on model weights, activations, and gradients, while inference requires only temporary tensors and KV caches [15][16]. - The demand for storage during training is significantly higher, with requirements ranging from terabytes to petabytes depending on the model size [24][25]. Market Dynamics - The demand for memory is outpacing supply, leading to increased prices for HBM, DRAM, and SSDs [21][29]. - Hyperscalers are signing multi-year purchase agreements and vertically integrating into chip design to secure memory supplies [29][36]. Comparison of AI Models - Gemini 3.0 is currently outperforming ChatGPT 5.0 in various benchmarks, attributed to its optimized training and architecture [33][34]. - The U.S. is leading in AI model development compared to China, with significant differences in performance and resource availability [35][36]. Cost Considerations - Despite the higher initial costs of SSDs, their lower operational costs and performance benefits make them more economical for performance-critical tasks over time [40][42]. - The TCO for SSDs is favorable due to lower power consumption, reduced cooling needs, and higher reliability compared to HDDs [40][42]. Conclusion - The AI data center memory landscape is evolving rapidly, driven by increasing model sizes and the need for efficient memory solutions. The shift from HDDs to SSDs and the emergence of new memory technologies are key trends to watch in this sector.