电力设备 证券研究报告 行业研究 / 行业点评 2026 年 01 月 21 日 行业及产业 台积电 2025Q4 营收同比增长，DRAM 价格持续上涨 ——数据中心供配电设备行业跟踪 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 《锂电行业跟踪：2025 年 12 月国内电池产量 和装车量同比高增，六氟磷酸锂价格下降》 2026-01-21 《光储行业跟踪：光伏出口退税将取消，电池 片价格持续上涨》2026-01-21 《数据中心供配电设备行业跟踪：台积电 2025Q4 营收创新高，DRAM 涨价趋势延续》 2026-01-14 《锂电行业跟踪：2025 年 11 月新能源车销量 高景气延续，碳酸锂价格快速上行》 2026-01-13 《光储行业跟踪：光伏出口退税取消，硅料价 格小幅上涨》2026-01-13 朱攀 S0820525070001 021-32229888-25527 zhupan@ajzq.com 陆嘉怡 S0820124120008 021-32229888-25521 lujiayi@ajzq.com 强于大市 投资要点： 数据中心已成为电力 ...

Market Performance - A-shares major indices showed strong performance in 2025, with the CSI 300 index rising by 17.66%, the ChiNext index by 49.57%, and the SSE 50 index by 12.9% [1] - The North Stock 50 index increased by 38.8%, and the Sci-Tech 50 index rose by 35.92%, with the Shanghai Composite Index surpassing 4000 points for the first time in nearly a decade on October 28 [1] - In comparison, the S&P 500 index in the US recorded a 16.39% annual increase, while the Nasdaq managed a 20.36% rise, and the Dow Jones increased by 12.97% [2] Comparative Analysis - A-shares outperformed US benchmarks in 2025, with the CSI 300 index slightly exceeding the S&P 500 index, while indices focused on technology and growth, such as ChiNext and Sci-Tech 50, significantly outperformed the Nasdaq [5] - The A-share market exhibited a more balanced and diversified growth pattern, driven by multiple sectors including materials (+57.67%), information technology (+47.18%), and industrials (+30.03%) [8] Sector Dynamics - The A-share market's growth is characterized by a dual-engine structure focusing on advanced manufacturing and critical resource supply chains, contrasting with the US market's heavy reliance on a few tech giants [9] - The US market's growth is primarily driven by major tech companies, with the "Magnificent Seven" accounting for nearly 40% of the S&P 500's total market capitalization, indicating a concentrated market structure that poses risks [6] Notable Stocks - In 2025, two stocks in the A-share market achieved over tenfold increases, with Shangwei New Materials leading at 1689.13% and Tianpu Co. at 1645.4% [12] - Shangwei New Materials' surge was attributed to its transformation from traditional chemicals to humanoid robots, supported by a change in control to Zhiyuan Robotics [13] US Market Trends - The US tech sector saw significant differentiation among the "Magnificent Seven," with Google and Nvidia outperforming the market, while other giants like Microsoft and Meta lagged behind [21] - Data storage companies emerged as major winners in the S&P 500, with top performers like Western Digital and Micron benefiting from increased AI infrastructure investments [18] Investment Themes - The investment narrative in the US shifted from potential AI concepts to profit realization capabilities, impacting the valuation of tech companies based on their ability to demonstrate revenue growth and cost advantages [22] - Nvidia's strong performance is attributed to its dominant position in the AI computing supply chain, with consistent quarterly earnings exceeding market expectations [23]

Investment Rating - The report assigns an "Outperform" rating for the power equipment industry, indicating a positive outlook for investment opportunities in this sector [3]. Core Insights - The data center industry has become a core incremental application scenario for the power equipment sector, directly driving demand growth and technological iteration in power equipment [6]. - The report emphasizes the need to incorporate multi-dimensional indicators from the AI industry to accurately assess the demand for power distribution equipment, given the capital expenditure characteristics of the data center industry [3][6]. - Key indicators are constructed from three aspects: demand side (capital expenditure trends of leading cloud vendors), supply chain (GPU supply capabilities), and AI application development [3][6]. Summary by Sections Demand Side - Capital expenditure from overseas cloud vendors reached $99.617 billion in Q3 2025, a year-on-year increase of 80.39% and a quarter-on-quarter increase of 9.54% [8]. - Alibaba's capital expenditure was 31.5 billion yuan in Q3 2025, up 80.10% year-on-year but down 18.55% quarter-on-quarter [13]. - Tencent's capital expenditure was 13 billion yuan, down 24.05% year-on-year and 32.05% quarter-on-quarter [13]. Supply Chain - Nvidia's revenue in Q3 2025 reached 362.571 billion yuan, a historical peak with a quarter-on-quarter growth of 24.63% and a year-on-year growth of 62.49% [18]. - TSMC's revenue in November 2025 was 343.614 billion New Taiwan dollars, a year-on-year increase of 24.5% [24]. - The CPU price index in October 2025 was 98.20, showing a slight recovery from 96.15 in September [27]. Application Side - The number of AI models is steadily increasing, with a notable growth in application deployments [29]. - The token usage on OpenRouter from December 16 to December 22 was 5.70T, reflecting a 2.56% decrease [35]. - The price of tokens for models scoring over 40 on the Artificial Analysis Intelligence Index dropped by over 50% in Q3 2025 [46].

Core Insights - The article discusses the investment logic required to capture opportunities in the AI chip sector, highlighting the recent success of companies like Moer Thread and Muxi, which have seen significant market capitalization increases [2][3] - Lenovo Capital stands out as a unique investor that successfully backed multiple AI chip giants, demonstrating a forward-looking investment strategy [5][6] Investment Strategy - Lenovo Capital's early investments in companies like Cambricon and its simultaneous backing of Moer Thread and Muxi during their A-round financing showcase its proactive approach [5][6] - The firm’s investment logic is rooted in a long-term vision, aiming to identify and support innovative directions for Lenovo Group over the next 5-10 years [6][7] Market Positioning - Lenovo Capital has built a complementary landscape of AI computing capabilities through its investments in various companies, each addressing different aspects of computing needs [7][9] - The firm emphasizes deep industry research, dedicating significant time to understanding market trends before making investment decisions [8][9] Ecosystem and Value Creation - The article highlights the dual benefits of Lenovo Capital's investments, where the firm not only provides resources but also gains from the technological advancements of its portfolio companies [12][13] - Collaborations between Lenovo Group and its portfolio companies have led to integrated solutions that enhance value beyond mere financial returns [13][14] Future Outlook - Lenovo Capital is not only focused on current AI chip technologies but is also exploring next-generation computing paradigms, including quantum computing and brain-like computing [14][15]

Core Insights - The article discusses the recent IPOs of domestic GPU companies, Moer Thread and Muxi, highlighting the investment strategy of Lenovo Capital in the AI computing sector [2][3]. Group 1: Investment Strategy - Lenovo Capital has strategically invested in four key AI computing companies: Cambricon, Haiguang, Moer Thread, and Muxi, becoming the only investment institution to hold stakes in all four [3]. - The firm began its investment in the AI computing space around 2016, well before it became a hot topic, by backing Cambricon in its Series A round and continuing to invest through its IPO [2][4]. Group 2: Technology and Market Trends - The four companies complement each other in terms of technology paths and product positioning, aligning with the diversified demand for AI computing [5]. - Lenovo Capital's approach reflects a systematic coverage of the AI computing landscape, moving from centralized and generalized computing to diversified, specialized, and ubiquitous computing [5]. Group 3: Ecosystem Development - The role of capital in China's semiconductor industry is evolving from merely providing financial support to becoming a co-builder of ecosystems, as exemplified by Lenovo Capital [6]. - Lenovo Capital is also investing in next-generation computing paradigms, including integrated storage and computing, quantum computing, RISC-V, and brain-like computing technologies [6].

Core Insights - Tech stocks, particularly in the Nasdaq, have shown resilience in 2025 with a gain of approximately 20% despite early-year volatility [1][2] - Historical trends suggest that bull markets lasting three years often extend to an average of eight years, indicating potential for continued growth in the Nasdaq [2] - Major technology companies have reported strong quarterly results, reinforcing the sustainability of the Nasdaq's rally [2] Company Analysis: Nvidia - Nvidia's stock has increased by 34% in 2025, currently trading around $180, driven by high demand for its data center GPUs [4][5] - The company has a backlog of booked orders worth $307 billion, significantly exceeding its past four quarters' revenue of $187 billion, leading to heightened growth expectations for the next fiscal year [7][10] - Analysts forecast a 58% increase in Nvidia's earnings for fiscal 2027 to $7.43 per share, with the stock trading at a lower price-to-earnings ratio compared to the Nasdaq 100 index [10] Company Analysis: Applied Digital - Applied Digital's stock has surged by 211% in 2025, currently priced around $24, despite a recent decline of about 33% from its October peak [12][13] - The company is expanding its data center capacity, with existing leases representing potential revenue of $16 billion and plans for 700 MW of capacity by 2027 [14] - With $5 billion in funding secured for future development, Applied Digital's growth is expected to accelerate, potentially increasing its market cap by 27% based on projected revenue [16][17]

Group 1 - The global data center GPU shipments saw a significant quarter-on-quarter increase of 145% in Q3, while PC GPU shipments experienced a modest growth of 2.5% quarter-on-quarter and 4% year-on-year [2] - LG Chem announced the development of core technology for commercializing all-solid-state batteries, which improves battery capacity by approximately 15% and high-rate discharge capacity by about 50% compared to existing electrolyte batteries [2] - A new low-altitude super factory for ZG-ONE aircraft was launched in Shenzhen, marking the entry of the first generation of smart electric vertical take-off and landing aircraft into the market, alongside the establishment of an innovation center for low-altitude economy [2] - Samsung has received a patent for an "augmented reality wearable device," which includes an innovative hinge design aimed at enhancing the overall comfort of future AR glasses [2]

Group 1: Gold Jewelry Price Increase - Multiple domestic gold jewelry brands have raised their gold jewelry prices, with Chow Sang Sang and Lao Feng Xiang both quoting 1315 CNY per gram, while Chow Tai Fook quoted 1312 CNY per gram [1] Group 2: Fund Risk Level Adjustment - China Construction Bank has announced an increase in the risk levels of 87 public fund products, with 32 products moving from "medium-low risk" to "medium risk" and 55 products from "medium risk" to "medium-high risk" [1] Group 3: Charging Battery Regulations - The new national standard for mobile power supplies is expected to increase industry costs by 20%-30%, primarily due to higher quality battery procurement and hardware upgrades [2] - The overall GPU shipments for data centers surged by 145% quarter-on-quarter in Q3, while PC GPU shipments saw a modest increase of 2.5% quarter-on-quarter and 4% year-on-year [2] Group 4: Nvidia Short Selling - Notable short-seller Michael Burry has reiterated his bearish stance on Nvidia, labeling the current AI hype as a "magnificent absurdity" and suggesting that Nvidia is a key indicator of a potential bubble in the AI sector [4] Group 5: Novartis China Team Restructuring - Novartis has announced a change in its operational model for county-level business in China, stating that it will continue to support county-level healthcare despite the restructuring [5] Group 6: Vanke Bond Performance - Vanke's 2027 dollar bond has experienced a record drop, with a decrease of 12 cents to 43.5 cents per dollar, marking the largest single-day decline since its issuance in 2017 [6] Group 7: Regulatory Actions - Hengbang Co. received an administrative regulatory decision from the Shandong Securities Regulatory Bureau due to non-compliance with independent director replacement rules [7] - Coris received a regulatory decision from the Guangdong Securities Regulatory Bureau for violations related to accounts receivable, related party transactions, and insider information management [11] Group 8: Stock Market Performance - The ChiNext Index saw a nearly 2% rebound, with significant trading volume and numerous stocks hitting the daily limit up, particularly in AI application and computing hardware sectors [17]

Core Insights - Intel's CEO, Lip-Bu Tan, acknowledged significant changes in the AI leadership team following the departure of Sachin Katti, who was appointed as Chief Technology and AI Officer earlier this year [2][3] - Katti is set to join OpenAI to help build computing infrastructure for its general AI goals, highlighting a shift in Intel's AI strategy [2][3] - Tan emphasized the importance of AI as a top priority for Intel, aiming to establish the company as a preferred computing platform for next-generation AI workloads [3] Leadership Changes - The announcement of Katti's departure coincided with the exit of Saurabh Kulkarni, a senior executive in the data center AI division, who has joined AMD [4][5] - Anil Nanduri has taken over Kulkarni's responsibilities in AI product management [5] - Tan has been actively restructuring Intel's organization to reduce complexity and enhance innovation, marking his second major adjustment in two months [6][8] AI Strategy and Market Position - Intel's AI strategy aims to challenge Nvidia's dominance in the AI infrastructure market, with a focus on high-performance GPUs and cost-effective systems [2][10] - Katti previously outlined a new strategy involving heterogeneous systems designed to optimize performance for different stages of AI workloads, claiming a 70% cost-performance improvement compared to existing homogeneous systems [10][11] - The company is committed to an open software approach that supports multiple infrastructure vendors without requiring developers to change their usage habits [11]

Core Insights - Qualcomm is challenging Nvidia's dominance in the AI computing sector by launching two AI chips, AI200 and AI250, expected to be commercially available in 2026 and 2027 respectively [1][2] - Qualcomm's stock surged over 20% on October 27, marking its largest single-day gain since 2019 following the announcement of these AI chips [1] - Nvidia has a stronghold in the AI computing chip market, particularly in data center GPUs, with established technology and ecosystem barriers [1] Group 1: Product Launch and Strategy - Qualcomm's AI chips focus on rack-level performance and high memory for rapid generative AI inference, emphasizing advantages in energy efficiency and total cost of ownership [1] - The first AI rack solutions are set to be deployed in 2026 at the data center of Middle Eastern client "HUMAIN," marking a significant milestone for Qualcomm's AI chip commercialization [1] Group 2: Market Position and Acquisitions - Qualcomm's acquisition of Alphawave Semi for approximately $2.4 billion aims to enhance its semiconductor technology for data centers, indicating a strategic move to strengthen its AI product portfolio [2] - Qualcomm's revenue is primarily driven by its semiconductor business, which consists of mobile, automotive, and IoT segments, with mobile chips contributing about 70% of revenue [2] Group 3: Competitive Landscape and Challenges - Despite Qualcomm's technical advantages from its experience in mobile chip design, Nvidia's market and ecosystem barriers remain high, and specific performance metrics for AI200 and AI250 are yet to be disclosed [2] - The lengthy procurement cycles and complex validation processes in data center equipment mean that Qualcomm may require time to achieve scalable revenue from its new AI products [2]