英伟达（NVDA.O） 2026 年 02 月 28 日 投资评级：买入（维持） ——美股公司信息更新报告 | 日期 | 2026/2/28 | | --- | --- | | 当前股价(美元) | 177.19 | | 一年最高最低(美元) | 212.18/86.60 | | 总市值(亿美元) | 43057 | | 流通市值(亿美元) | 43057 | | 总股本(亿股) | 243 | | 流通股本(亿股) | 243 | | 近 3 个月换手率(%) | 44.40% | 股价走势图 -40% -30% -20% -10% 0% 10% 20% 30% 40% 50% 60% 英伟达 纳斯达克指数 数据来源：聚源 相关研究报告 | 初敏（分析师） | | 杨哲（分析师） | | | | | --- | --- | --- | --- | --- | --- | | chumin@kysec.cn | | yangzhe@kysec.cn | | | | | 证书编号：S0790522080008 | | 证书编号：S0790524100001 | | | | |  | Agent | 行情带动需 ...

Group 1: Industry Insights - Nvidia (NVDA.US) is expected to release its next-generation Rubin series products in the second half of next year, utilizing M9 materials and orthogonal backplanes, which significantly increases processing difficulty and leads to a rise in both quantity and price of PCB drill bits [1] - The global PCB drill bit industry is currently facing a shortage, with the top five companies (CR5) holding a combined market share of 75.2% as of the first half of 2025, with DingTai High-Tech leading with a sales volume of 500 million units and a market share of 28.9% [2] - The demand for high-performance drill bits is expected to increase significantly due to the higher requirements for drill bit length, breakage rates, and processing performance associated with advanced materials [1][2] Group 2: Company Developments - Minexplosion Optoelectronics (301362.SZ) announced plans to acquire 100% of Xiamen Maida's stake in Xiamen Precision, focusing on the production of PCB, FPC, IC substrates, and AI PCB tungsten micro-drills, with a product size range of 0.09mm to 0.35mm [1] - DingTai High-Tech (301377.SZ) is set to benefit from the technological iteration in the PCB drill bit market, leveraging its scale and self-manufacturing capabilities to maximize profits amid the current shortage [2] - Wald (688028.SH) has made progress in internal verification of PCD drill bits, focusing on high-hardness and brittle materials for PCB and semiconductor applications, with the ability to process over 8,000 holes in M9 boards without breakage [3] - Xiamen Precision, established in 1995, has a strong technical foundation and plans to expand its production capacity from the current 15 million units per month to over 50 million units in the next 2-3 years [3]

Core Viewpoint - The A-share market opened positively on January 5, with the Shanghai Composite Index rising over 1% and surpassing 4000 points, driven by a rebound in the hard technology sector, particularly the robotics sector [1] Group 1: Market Performance - The Robot ETF Fund (159213) increased by 1.63%, recovering from the previous day's decline, with a trading volume exceeding 35 million yuan [1] - Continuous capital inflow into the robotics sector, with the Robot ETF Fund attracting over 24 million yuan during the trading session and accumulating over 120 million yuan in five consecutive days [1] Group 2: Component Stocks Performance - Most component stocks of the Robot ETF Fund saw significant gains, with iFlytek rising over 6%, Huichuan Technology over 4%, and Dazhong Laser over 3% [3] - The top ten component stocks of the Robot ETF Fund include: - iFlytek (6.76% increase, 9.95% estimated weight) - Huichuan Technology (4.10% increase, 9.94% estimated weight) - Dazhong Laser (3.74% increase, 4.08% estimated weight) [3] Group 3: Industry Developments - Yushun Technology released a training video of its humanoid robot H2, showcasing advanced movements, and clarified that recent reports regarding its IPO status were misleading [4] - The humanoid robot industry is approaching mass production, with significant changes expected in 2026, including the release of Tesla's V3 robot and the establishment of domestic standards for humanoid robots [5] - The industry is experiencing a convergence of application, computing power, and capital, with expectations for a significant leap in humanoid robotics by 2026 [6]

Investment Rating - The report maintains a "Buy" rating for Nvidia (NVDA.O) [7][11]. Core Insights - Nvidia's Q3 performance and strong guidance counter the concerns regarding an AI bubble, indicating genuine resilience in AI demand [3][11]. - The company has adjusted its revenue forecasts for FY2026E-FY2028E to $213.7 billion, $341 billion, and $412.5 billion respectively, with corresponding Non-GAAP net profits of $114.2 billion, $193.1 billion, and $234.5 billion [11][12]. Financial Summary - Revenue projections for Nvidia are as follows (in million USD): - FY2024: $60,922 (125.9% YoY growth) - FY2025: $130,497 (114.2% YoY growth) - FY2026E: $213,695 (63.8% YoY growth) - FY2027E: $341,044 (59.6% YoY growth) - FY2028E: $412,504 (21.0% YoY growth) [5][11]. - Non-GAAP net profit projections are: - FY2024: $32,312 (286.3% YoY growth) - FY2025: $74,266 (129.8% YoY growth) - FY2026E: $114,187 (53.8% YoY growth) - FY2027E: $193,110 (69.1% YoY growth) - FY2028E: $234,537 (21.5% YoY growth) [5][11]. Market Data - Current stock price is $180.64, with a 52-week price range of $94.31 to $207.04 [7][8]. - The current market capitalization is approximately $4.39 trillion [8]. Earnings Forecast and Investment Recommendations - The report suggests a target price of $252 for FY2027, based on a PE ratio of 32X, maintaining the "Buy" rating [11][12]. - The strong Q3 results included a revenue of $57 billion, exceeding Bloomberg's consensus estimate of $55.1 billion, with data center revenue growing by 66% YoY to $51.2 billion [11].

Core Insights - Nvidia's valuation has reached $5 trillion due to its significant share in the artificial intelligence spending boom [2] - The new benchmark for advanced data centers is measured in gigawatts of computing power, shifting the focus from physical size or server count [2] - The cost of 1 gigawatt (GW) of AI data center capacity is approximately $35 billion, representing a new economic foundation for the AI industry [3] Cost Structure of AI Data Centers - Approximately 39% of total spending in AI data centers is allocated to GPUs, with Nvidia's products dominating this segment [6] - Nvidia captures nearly 30% of the profits from AI data center expenditures due to its 70% gross margin [6] - Each gigawatt of power can support over 1 million GPU chips, generating $1.3 billion in revenue for Nvidia's manufacturing partner, TSMC [6] Networking Equipment - 13% of data center costs are attributed to networking equipment, benefiting companies like Arista Networks, Broadcom, and Marvell [7] - Component manufacturers such as Amphenol and Luxshare Precision will also gain from cables and connectors [7] Power and Cooling Infrastructure - Physical infrastructure, including power distribution, accounts for nearly 10% of the costs of a 1 GW AI data center [9] - Major players in this sector include Eaton, Schneider Electric, ABB, and Vertiv, with Vertiv also having opportunities in thermal management [9] Real Estate and Labor Costs - Land and buildings represent about 10% of upfront costs, while operational costs are relatively low, with annual electricity costs for a 1 GW data center around $1.3 billion [11] - Large data centers typically employ only 8 to 10 staff members, with salaries ranging from $30,000 to $80,000 [11] - The bottleneck is shifting to power supply, with companies like Siemens Energy and GE Vernova reporting increased orders for turbine and grid infrastructure [11]

Core Insights - Nvidia has become the first company globally to reach a market capitalization of $5 trillion, driven by its significant share in AI investment spending [1] - The measurement of data centers has shifted from physical size to computing power measured in gigawatts, with Wall Street now evaluating data centers based on "cost per gigawatt" [1] - The construction cost for a 1 gigawatt AI data center is estimated at $35 billion, representing a new economic foundation for the AI industry [2] Cost Structure of AI Data Centers - The largest single cost in AI data centers is attributed to GPUs, accounting for approximately 39% of total expenditures, with Nvidia's chips being the primary contributors [3] - Each 1 gigawatt of computing power requires over 1 million GPU chips, highlighting the central role of GPUs in the AI industry [3] - Networking equipment follows GPUs in cost, comprising about 13% of data center expenses, benefiting companies like Arista Networks and Broadcom [4] Infrastructure and Operational Costs - Power and cooling infrastructure, including generators and transformers, represent nearly 10% of total costs for a 1 gigawatt AI data center [6] - The operational costs of running a 1 gigawatt AI data center are relatively low, with annual electricity costs estimated at $1.3 billion and minimal personnel requirements [7] - The industry is facing challenges related to power supply, with major companies like Siemens Energy and GE Vernova reporting a surge in orders for turbines and grid infrastructure [7]

Core Viewpoint - NVIDIA's CEO Jensen Huang expressed optimism about the revenue potential of the Blackwell and Rubin AI product lines, projecting significant sales growth compared to the previous Hopper series, although subsequent clarifications adjusted these figures [2][4]. Revenue Projections - The initial claim of $500 billion in revenue from Blackwell and Rubin products over the next five quarters was clarified to represent cumulative shipments from 2025 to 2026, including revenues from NVIDIA's InfiniBand and NVLink products [4]. - It is estimated that 30% of the expected demand has already been shipped, contributing $100 billion in revenue from Blackwell products this month [4]. - The revised revenue expectation for the next five quarters is $307 billion, lower than the initial projection made by Huang [4]. Product Performance - The Blackwell series is noted for its superior performance and energy efficiency, which has made it popular among customers, marking a significant achievement for NVIDIA [4]. - The Rubin product line is anticipated to be crucial for NVIDIA's expansion in computing capabilities, with the introduction of the Vera Rubin superchip, which integrates ARM-based Vera CPU and Rubin chipsets [4]. Market Position and Growth - NVIDIA's market capitalization reached a record $5 trillion, indicating strong momentum in the AI sector [5]. - The company has transitioned from a focus on consumer GPUs to establishing a foundational role in providing necessary computing power for large tech firms, significantly increasing its market share [5]. - NVIDIA's recent announcements at the GTC 2025 conference have contributed to its soaring market value and reinforced its leadership in AI computing [5][6]. Strategic Developments - The GTC 2025 conference showcased NVIDIA's collaborations with notable companies like Nokia and Palantir, highlighting untapped applications of AI technology [6]. - There is potential for NVIDIA to re-enter the Chinese AI market, which could yield substantial additional revenue, as discussions about Blackwell AI chips with Chinese leaders are anticipated [6].

Market Performance - As of October 23, 2025, the Zhongzheng Semiconductor Industry Index decreased by 1.52%, with major stocks like ShenGong Co. leading the decline at 5.10% [2] - The Sci-Tech Innovation Board Chip Index fell by 2.00%, with stocks such as TuoJing Technology rising by 3.06% while ShengKe Tongxin dropped by 4.98% [4] - The Sci-Tech Innovation Board New Materials Index declined by 1.32%, with TianCheng Technology increasing by 5.98% and ShenGong Co. again leading the decline [7] ETF Performance - The Semiconductor Industry ETF (159582) decreased by 1.37%, with a recent price of 2.09 yuan, but showed a 3.87% increase over the past month [2] - The Sci-Tech Chip ETF (588990) also fell by 1.94%, with a recent price of 2.43 yuan, and a 0.98% increase over the past month [4] - The Sci-Tech New Materials ETF (588010) dropped by 1.37%, with a recent price of 0.79 yuan, and a 0.75% increase over the past month [7] Liquidity and Trading Volume - The Semiconductor Industry ETF had a turnover rate of 4.43% with a trading volume of 17.2464 million yuan, averaging 77.5828 million yuan daily over the past month [2] - The Sci-Tech Chip ETF had a turnover rate of 8.66% with a trading volume of 58.0718 million yuan, averaging 161 million yuan daily over the past month [4] - The Sci-Tech New Materials ETF had a turnover rate of 8.03% with a trading volume of 19.9337 million yuan, averaging 32.1769 million yuan daily over the past week [7] Industry Developments - Yangtze Memory Technologies is considering an IPO in mainland China, potentially valued over 40 billion USD, which could be one of the largest listings in the domestic semiconductor sector [8] - Tesla has added a manufacturing engineer position for its Optimus robot, planning to launch several products next year, indicating a growing interest in humanoid robots [8] - NVIDIA's new Rubin series will utilize M9 materials, suggesting an increase in material chain focus and potential technological upgrades [8] Institutional Insights - The expectation of Yangtze Memory's IPO strengthens the logic of domestic storage autonomy, potentially boosting valuations in chip equipment and material chains [9] - The AI and semiconductor sectors are experiencing renewed interest, with a focus on upstream components like sensors and drive systems due to Tesla and other developments [9] - The peak production of global advanced packaging lines is expected between 2025 and 2028, with equipment procurement cycles starting in late 2024, indicating a significant demand for related equipment [9] ETF Overview - The Semiconductor Industry ETF closely tracks the Zhongzheng Semiconductor Industry Index, covering key sectors in semiconductor materials, equipment, and applications [10] - The Sci-Tech Chip ETF tracks the Sci-Tech Innovation Board Chip Index, focusing on semiconductor-related companies with a maximum of 50 constituents [10] - The Sci-Tech New Materials ETF follows the Sci-Tech Innovation Board New Materials Index, concentrating on advanced materials relevant to the semiconductor industry [10] Weighting and Composition - As of September 30, 2025, the top ten stocks in the Zhongzheng Semiconductor Industry Index accounted for 78.18% of the index [11] - The top ten stocks in the Sci-Tech Chip ETF represented 59.69% of the index [11] - The top ten stocks in the Sci-Tech New Materials Index made up 51.58% of the index [12]

Group 1: Nvidia's New Product and Market Impact - Nvidia has confirmed the use of M9 materials in its next-generation product Rubin, with the CPX and midplane PCBs utilizing M9 CCL, amid a shortage of Q cloth [1] - The Rubin CPX is designed for ultra-long context AI inference tasks, featuring a decoupled inference architecture and significant hardware changes, including a wireless cable architecture [1] - The market potential for CPX, midplane, and orthogonal backplane is nearly 100 billion [1] - By 2027, Nvidia's AI PCB market is projected to reach $6.96 billion, a 142% increase from 2026 [1] Group 2: PCB Industry Growth Driven by AI - The global PCB market is expected to grow from $62 billion in 2020 to $75 billion by 2024, with a compound annual growth rate of 4.9% [2] - The value of PCBs in AI servers is significantly higher than in traditional servers, leading to a substantial increase in demand for high-performance PCBs [3] - Low dielectric constant and low loss factor materials are deemed most suitable for high-performance applications, with Q cloth outperforming second-generation cloth [3] Group 3: Recommendations for M9 and Related PCB Manufacturers - Strong recommendations have been made for M9 upstream and related PCB manufacturers, with companies like 建滔积层板 reporting a revenue increase of 11% year-on-year [4] - 建滔集团 is expanding its production capabilities for AI-related products, with a projected investment of approximately 800 to 1,000 million RMB for a new production line [4] - The demand for copper-clad laminates and printed circuit boards is expected to rise significantly due to the rapid development of AI technologies [4]

Group 1: Nvidia's New Product and Market Impact - Nvidia is set to launch its new product series, Rubin, in the second half of next year, utilizing M9 materials for its CPX and midplane PCBs due to a shortage of quartz fabric [1] - The Rubin CPX is designed specifically for long-context AI inference tasks, featuring a decoupled inference architecture and significant hardware innovations, which are expected to expand its market size [1] - According to CICC, the AI PCB market for Nvidia is projected to reach $6.96 billion by 2027, representing a 142% increase from 2026 [1] Group 2: PCB Industry Growth Driven by AI - The global PCB market is expected to grow from $62 billion in 2020 to $75 billion by 2024, with a compound annual growth rate (CAGR) of 4.9% [2] - The demand for high-performance PCBs is anticipated to rise significantly due to the higher value of AI server PCBs compared to traditional servers, with low dielectric constant materials being preferred [3] Group 3: Recommendations for M9 and Related PCB Manufacturers - Strong recommendations have been made for M9 upstream and related PCB manufacturers, with companies like Kintor achieving a revenue of HKD 9.588 billion in the first half of the year, a year-on-year increase of 11% [4] - Kintor Group is actively developing high-frequency and high-speed products for AI server GPU motherboards and plans to establish an AI PCB production line in Guangdong with an investment of approximately RMB 800 million to 1 billion [4]