前言 ②2026年或成中国商业火箭可回收发射元年，机构前瞻商业航天投资机遇：重视火箭核心"铲子 股"及3D打印新技术路线标的； ③上海低空经济布局再升级，eVTOL产业链风口已至； ④马斯克按下"量产快进键"，脑机接口产业迎来关键拐点。 点击付费阅读，解锁市场最强音，把握投资机会！ 券商研报信息复杂？机构调研数据过时？屡屡错失投资机会？那是你不会挖！想知道哪份研报有用？什 么时候该看？《研报金选》满足你！每日拆解热门产业链或核心公司，快市场一步的投研思维+严苛的 研报选择标准+几近偏执的超预期挖掘，游资私募都在用！ 【今日速览】 ①无光不AI！新一轮算力革命的关键推手，引爆10倍市场空间，产业链利润正向这个环节集中！硅 光产业投资全梳理：从Fab厂到光模块，这些公司卡位算力新基建； ...

证券研究报告 晨会 资料来源：聚源，华源证券研究所，截至2025年12月24日 华源晨会精粹 20251224 公用环保 算力革命与能源革命共振，美国缺电背景下的电力投资机遇：OpenAI 已 将其截至 2033 年算力投资规模上调至 250GW，2030 年美国最高用电负荷有望突破 1000GW（目前在 820GW 上下），需求大幅上调导致美国可能出现缺电问题。电源 侧可能通过建设气电、核电、储能、SOFC 等形式解决，同时电网建设也有望同步。 此外衍生的问题是：（1）如果美国算力投资继续上调或倾尽全力也无法满足算力需 求，行业发展将如何演化？（2）中美科技竞争背景下，算力投资对中国电力供需有 何影响。发电侧：气电、核电是主力电源，储能、SOFC 作为应急手段。我们测算 在 2030 年 1000GW 的最高预测下，考虑/不考虑既定机组退役时，美国电力缺口为 182/89GW。电网侧：电网投资上调，国内企业或迎出口机遇。电源设备：800VDC 是演化趋势，SST 或将是长期解决方案。国内篇：考虑中国 AIDC 投资与美国相当， 中性情景下 AIDC 和充电桩将拉动十五五用电量复合增长率约 1.1pct、乐观 ...

Core Viewpoint - OpenAI has raised its computing power investment target to 250GW by 2033, while the peak electricity load in the U.S. is expected to exceed 1000GW by 2030, leading to potential electricity shortages in the U.S. due to significantly increased demand [1] Group 1: Power Supply and Demand - OpenAI's computing power investment has been significantly increased, indicating a notable shortfall in U.S. electricity supply and demand [1] - The projected electricity gap in the U.S. by 2030 is estimated to be 182GW or 89GW, depending on whether existing units are retired [2] - Gas and nuclear power are expected to be the main sources of electricity, with gas power being a stable and cost-effective option, while nuclear power's capacity is set to increase from approximately 100GW to 400GW by 2050 [2] Group 2: Grid Investment and Export Opportunities - U.S. grid investment is projected to rise significantly, with expectations of surpassing $30 billion in 2024, driven by the need to maintain grid reliability [3] - China is expected to see a substantial increase in transformer exports to non-U.S. regions, with growth rates of 41% in Q1-Q3 of 2025, indicating a potential opportunity for Chinese exports due to U.S. electricity investment growth [3] Group 3: Power Equipment Trends - The trend towards 800VDC power architecture is emerging, with NVIDIA proposing this as a solution to increasing power demands from GPUs, which could enhance power efficiency by approximately 157% compared to 415V AC [4] - Solid-state transformers (SST) are anticipated to become a long-term solution for converting medium-voltage AC to 800V DC [4] Group 4: Domestic Implications - The competition between China and the U.S. in technology may lead to increased computing power investments in China, potentially resulting in a tighter domestic electricity supply-demand balance [5] - If China's AIDC investment aligns with that of the U.S., it could drive a compound growth rate of approximately 1.1% to 1.5% in electricity consumption during the 14th Five-Year Plan [5]

Investment Rating - The industry investment rating is optimistic (maintained) [1] Core Insights - The report highlights a significant increase in computing power investment, with OpenAI raising its projected capacity investment to 250GW by 2033, leading to a potential electricity shortage in the U.S. as demand is expected to exceed 1000GW by 2030 [3][12] - The power generation side will rely on gas and nuclear power as primary sources, with storage and solid oxide fuel cells (SOFC) serving as emergency measures [3][29] - The report anticipates a substantial increase in U.S. grid investment, with projections indicating investments will exceed $30 billion in 2024 and continue to rise in subsequent years, presenting export opportunities for domestic companies [3][29] - The evolution towards 800VDC power systems is noted, with SST (solid-state transformers) expected to be a long-term solution for power supply challenges [3][29] - The domestic market is expected to experience a tightening of electricity supply due to AI investments, suggesting potential investment opportunities in domestic power and grid equipment manufacturers [3][29] Summary by Sections 1. Introduction - The report discusses the upward adjustment of computing power investments and the resulting significant electricity supply-demand imbalance in the U.S. [6] 2. Power Generation Side - Gas and nuclear power are identified as the main power sources, while storage and SOFC are positioned as emergency solutions [6][29] - The projected electricity gap by 2030 is estimated at 182GW, considering the retirement of existing power plants [3][29] 3. Grid Investment - U.S. grid investment is expected to increase significantly, with forecasts indicating investments reaching $37.8 billion by 2027 [3][29] - Domestic companies are likely to benefit from increased exports due to rising U.S. grid investments [3][29] 4. Power Equipment - The transition to 800VDC systems is highlighted as a trend, with SST potentially becoming a long-term solution for power supply issues [3][29] 5. Domestic Market - AI investments are projected to lead to a tightening of electricity supply in China, creating new investment opportunities in power and grid equipment sectors [3][29] 6. Investment Analysis - Detailed investment analysis and recommendations are provided in Chapter 6 of the report [3]

Core Viewpoint - Siemens has launched its upgraded data center solution 5.0 at the 20th China IDC Industry Annual Conference, showcasing new products aimed at future "AI factories," including the SENTRON 3VA series circuit breakers for 800/1000 volt DC applications, marking their debut in the Chinese market [1][6]. Group 1: Data Center Solutions - The new data center solution 5.0 integrates virtual simulation software, intelligent hardware, and AI applications, supporting the entire lifecycle from planning to operation and maintenance [3]. - Siemens aims to leverage its technological advantages and decades of industry experience to provide comprehensive lifecycle technical support for data centers, enhancing productivity in the AI era [3]. Group 2: Simulation and Optimization Tools - Simcenter software allows for simulation design of complex interactions from chips to cooling systems, enabling precise thermal flow analysis across the entire data center chain [4]. - PSS®SINCAL software enhances energy efficiency and stability in power systems during operation through simulation, analysis, and calculation [4]. Group 3: Green and Efficient Solutions - Siemens' green low-carbon solutions focus on energy management, integrating various digital products to achieve power supply greening, energy transparency, and efficiency [4]. - The WSCO solution optimizes cooling system operations using AI algorithms, achieving up to 40% energy savings in cooling costs with a payback period of less than three years [5]. Group 4: Safety and Reliability - Siemens provides high-reliability intelligent air-insulated switchgear and environmentally friendly gas-insulated switchgear for uninterrupted data center operations [5]. - The introduction of the SENTRON 3VA series circuit breakers meets both IEC and UL standards, facilitating customer access to overseas markets [6].

Group 1 - The WISE2025 conference highlighted the evolution of Chinese commercial ecology from mere visibility to deeper insights and practices, focusing on brand mentality, technological empowerment, and future scenarios [2][11][64] - The event featured interactive experiences showcasing advanced technologies, including various robots and smart devices, emphasizing the concept of human-machine coexistence [4][8][28] - The conference discussed the importance of brand recognition and emotional value in a changing market, with insights from industry leaders on how to build lasting brands [12][16][20] Group 2 - The release of the "Future Super Brand List" by 36Kr and Nielsen IQ emphasized the significance of cultural resonance and user loyalty in brand development amidst economic fluctuations [14][16] - Insights from various speakers, including Zhao Yan and Wang Zhiguo, highlighted the shift from industrial manufacturing to biological manufacturing and the role of AI in enhancing productivity and organizational efficiency [18][21][22] - The discussions on globalization underscored the need for companies to adapt to local markets and build value networks, moving from a mindset of mere expansion to one of symbiosis [38][39][46] Group 3 - The conference concluded with a consensus that Chinese business is transitioning from model innovation to value reconstruction, with a focus on the deeper logic of commercial practices [64][63] - Key trends identified included the importance of understanding generational motivations and the role of founders in shaping brand identity in a trust-scarce environment [55][56] - The event showcased the potential of AI and technology in redefining marketing and consumer engagement, emphasizing the need for businesses to leverage these tools for future growth [29][40][48]

Core Insights - The blockchain industry has been trapped in a cycle of "computational power involution," where PoW mechanisms consume vast amounts of energy without generating economic value, while PoS mechanisms exacerbate capital monopolies, making it difficult for ordinary users to participate [1][3] - AW3C introduces a revolutionary shift in computational power, transforming it from "ineffective consumption" to "value creation" [1][3] Group 1: AW3C's Innovations - AW3C's core breakthrough lies in reconstructing the value logic of computational power through its unique PoI consensus mechanism, which eliminates meaningless computational competitions and redefines computational resources as "intelligent labor" [3] - Instead of solving mathematical problems, computational power is now utilized for valuable AI tasks such as global commodity price comparisons, supply chain trend forecasting, and market demand analysis, providing users with decision-making support and companies with market insights [3] Group 2: Accessibility and Participation - AW3C breaks down industry entry barriers, allowing ordinary users to participate in computational mining without the need for specialized equipment or significant capital, as basic computer devices are sufficient to engage in lightweight AI tasks and earn rewards [3] - This democratization of computational power transforms it into a "value production tool" accessible to all, rather than being exclusive to a few industry giants [3] Group 3: Real-World Impact - AW3C enables blockchain technology to serve the real economy by generating pricing data and market predictions that empower consumer markets and supply chain industries, addressing issues like information asymmetry and decision-making efficiency [3] - The revolution led by AW3C is helping blockchain shed its image of being disconnected from the real economy, positioning it as a core driver of efficient economic operations [3]

Core Viewpoint - Nvidia's Q3 earnings report of $57 billion has effectively countered skepticism regarding the AI bubble, showcasing robust growth in its data center business and affirming the company's strong position in the AI market [3][4][5]. Financial Performance - Nvidia reported total revenue of $57 billion for Q3, a 22% increase from the previous quarter and a 62% increase year-over-year [4][5]. - The data center revenue reached $51.2 billion, reflecting a 25% quarter-over-quarter growth and a 66% year-over-year increase, significantly surpassing market expectations [4][5]. - The company's earnings per share (EPS) was $1.30, slightly above the consensus estimate of $1.26 [5]. Market Dynamics - Despite concerns about an AI investment peak, Nvidia's performance indicates a divergence from the broader market trends, where many AI application companies are struggling with inflated valuations and insufficient revenue [3][4]. - Nvidia's market capitalization briefly surpassed $5 trillion, highlighting its pivotal role in the global economy and technology sector [7]. AI Market Trends - Nvidia's CEO, Jensen Huang, emphasized that the current AI wave is not a bubble but a fundamental shift in computing paradigms, moving from general-purpose CPUs to accelerated GPU computing [10][12]. - Key trends driving this shift include the saturation of Moore's Law, the AI-driven upgrade of recommendation systems, and the emergence of intelligent agents that significantly increase demand for computational power [10][12][13]. Profitability and Market Position - Nvidia's gross margin is projected to be around 69.8% for the last 12 months, with expectations to reach 72.42% in the upcoming quarter, showcasing its superior profitability compared to typical hardware manufacturers [16][19]. - The company's strategic product roadmap, including the introduction of Blackwell and Rubin chips, is designed to maintain its competitive edge and create high barriers to entry for competitors [19][22]. Future Outlook - Global investments in AI infrastructure are expected to reach $3 to $4 trillion by 2030, positioning Nvidia as a primary beneficiary of this growth [25]. - Despite potential challenges in the AI application layer, the underlying demand for computational power remains strong, ensuring Nvidia's continued relevance and profitability [25][27]. Industry Implications - Nvidia's dominance in the GPU market has raised concerns about industry monopolization, as the company controls critical technology that underpins the AI revolution [29]. - The concentration of technological power in a single entity poses risks to competitive balance and the diversity of the global tech ecosystem, necessitating careful consideration by governments and businesses [29].

Core Insights - The "14th Five-Year Plan" emphasizes computing power as a core element of productivity in the digital economy, aiming to achieve the world's largest computing power scale by 2030 [1] - The "East Data West Computing" project has established a comprehensive computing power network covering eight national hub nodes and ten data center clusters, with the "super node" architecture emerging as a key technology for enhancing computing efficiency [1][2] Industry Trends - The demand for AI model training is growing exponentially, leading to a bottleneck in traditional data center architectures, with China's data centers consuming over 2% of the total electricity [2] - The "East Data West Computing" initiative aims to create a national integrated computing power network, focusing on efficient scheduling and green low-carbon operations [2] Technological Developments - The super node technology, characterized by high-density cabinet design and integration of heterogeneous computing resources, achieves a Power Usage Effectiveness (PUE) of below 1.05, significantly improving energy efficiency [2][3] - Super nodes have demonstrated a 40% increase in AI training efficiency and a 35% reduction in total ownership costs during tests at Alibaba Cloud's Zhangbei Super Data Center [3] Global Landscape - The global computing power infrastructure investment is expected to exceed $520 billion by 2025, with a year-on-year growth of 55% [4] - The U.S. maintains a lead through a "business-led + government-enabled" model, while China is rapidly advancing in intelligent computing and regional hub layouts under national strategies [4] Structural Challenges - The computing power industry faces structural issues such as supply-demand mismatches, high costs, and energy consumption pressures [4] - The existing challenges include an imbalance between supply and demand in eastern and western regions, and inefficiencies in resource utilization due to a lack of hardware-software synergy [4] Opportunities and Innovations - Liquid cooling technologies are gaining traction as a solution to the high energy consumption of computing facilities, potentially lowering PUE to very low levels [5] - Super nodes enhance effective computing resource utilization by over 50%, addressing the issue of idle computing resources in traditional clusters [6] Ecosystem Transformation - The strategic significance of super node technology extends beyond mere technical innovation, facilitating the pooling and service-oriented transformation of computing resources [7] - The first commercial intelligent computing super node was launched in May, significantly improving model training efficiency and performance [7] Future Prospects - The super node architecture supports the "East Data West Training" model, connecting real-time computing needs in the east with storage-type resources in the west through low-latency networks [8] - As computing power becomes a new productivity driver, super nodes are expected to evolve towards nanosecond latency and exabyte-level computing capabilities, forming the foundation for general artificial intelligence [8]

Core Insights - Major investments in AI infrastructure by global tech giants are driven by the urgent need for energy infrastructure upgrades and the new challenges posed by high-density computing on power supply [1] - The rapid development of renewable energy is outpacing the construction of grid infrastructure, necessitating accelerated grid development to keep up with power generation [1] - The rise of AI is providing new growth momentum for the power equipment industry, with data centers (AIDC) being central to AI infrastructure and their stable operation relying on energy supply [1] Group 1 - The traditional 415V AC systems are becoming inadequate due to increasing rack power density, leading to a potential shift towards 800V DC distribution systems [2] - The concentration of AC-DC conversion equipment is expected to rise, with the use of Solid State Transformers (SST) simplifying systems and aiding in carbon reduction for data centers [2] - The white paper from NVIDIA highlights significant fluctuations in rack power due to GPU power increases, presenting challenges for power supply and grid stability [2] Group 2 - Solutions proposed for managing power fluctuations include software optimization, energy storage systems, and limiting GPU performance, which opens up application spaces for supercapacitors and energy storage [2] - The development of new power systems, such as virtual power plants, is suggested to enhance system stability by matching electricity consumption with generation [2] - Companies with strong systemic solution capabilities are expected to gain a competitive advantage in this evolving landscape [2]