"21 世纪关键技术 " 关注科技未来发展趋势，研究 21 世纪前沿科技关键技术的需求，和影响。将不定期推荐和发布世界范围重要关键技术研究进展和未来趋势 研究。 2026年3月20日，法国兴业银行集团（Societe Generale Group）旗下的伯恩斯坦（Bernstein）发布了一份题为《科技的未来：太空科 技——全球机遇》（Future of Tech: Space Tech- global opportunities）的深度行业报告 。 这份由Venugopal Garre等人撰写的报告明确指出，参与太空经济的竞赛已经从遥不可及的雄心壮志，彻底转变为当下切实的投资主题 。 在过去的大部分历史中，太空探索主要被视为超级大国之间的面子工程和虚荣心竞赛，但如今，随着通信、地球观测和国防需求的重新点燃，从 日本、印度到法国的各国政府已经打开了商业化的大门，太空已不再是单纯的国家威望项目，而是一种核心的工业战略 。 根据美国太空基金会（Space Foundation）的估算，当前全球太空经济的规模约为6150亿美元，其中78%由私营部门驱动 。伴随着商业 发射成本的断崖式下跌、近地轨道（LEO）星座的激 ...

量子科技正从实验室走向产业主战场。中泰证券研究所近期发布的《2026量子科技产业发展前景、产业链布局、中美竞争现状及相关标的分析报 告》指出，全球量子产业市场规模预计到2035年将达970亿美元，到2040年更可能突破1980亿美元。在这一进程中，量子计算被认为是远期价值 最大的细分赛道，而量子通信与量子测量则率先进入商用落地阶段。报告同时深入分析了中美两国在这场"第二次量子革命"中的竞争格局，认为 中国正从量子通信的"单项冠军"向"全方面发展"转型，而"十五五"规划的政策加持或将推动中国量子产业步入黄金发展期。 三大赛道分化明显，量子计算被视为终极引擎 这份报告系统梳理了量子科技的三大核心领域——量子计算、量子通信和量子精密测量——各自的技术逻辑、产业链结构和市场前景，所呈现的 图景既充满想象空间，也包含清醒的不确定性判断。 量子计算被视为最具远景价值的方向。根据麦肯锡的预测数据，报告指出，到2035年量子计算创造的全球价值可能在280亿至720亿美元之间。光 子盒研究院的更为乐观的预测认为，产业规模将从2024年的约50亿美元跃升至2030年的近2200亿美元，到2035年或达到约8078亿美元。报告对此 ...

原文创作：半导体行业观察 近日，英伟达、谷歌和美国多家多学研究人员写了一篇名为《10-Year Roadmap for AI + Hardware》的文章。在文章中，他们披 露了包括芯片在内的AI硬件预期。 以下为文章正文： 人工智能 (AI) 和硬件 (HW) 正以前所未有的速度发展，但它们的轨迹已密不可分。大型 AI 模型和数据密集型应用的指数级增长对 更强大、更高效的硬件加速提出了更高的要求，而从 GPU、FPGA 和 TPU 到新兴的 NPU、模拟 AI 芯片、光子系统和神经形态处 理器等专用计算平台的突破，正在重新定义智能系统的极限。 这种良性循环正在改变计算格局，但也暴露出一个关键的差距：尽管两者协同演进迅速，但全球研究界缺乏一个统一的、长远的战 略愿景来协调 AI 和硬件的发展。今天的算法是围绕昨天的系统设计的，而明天的芯片是针对今天的工作负载优化的。这种碎片化 限制了构建能够在云端、边缘和物理环境中高效学习、推理和运行的整体性、可持续和自适应 AI 系统的进程。 与此同时，人工智能的能源消耗已达到环境和经济上不可持续的水平。训练一个前沿模型所需的能源相当于数百个家庭的用电量， 而人工智能数据中 ...

在SCSP评估的12个技术领域中，中国在先进电池、先进制造、商业无人机和5G基础设施四个领域保持明确领先，且置信度均为"高"。这 四个领域有一个共同特征：都是资本密集、依赖规模化生产与国家协调投入的"重基础设施"赛道，而恰恰是中国工业体制最擅长运作的 领域。 电池领域的数字最为直观。2023年，中国电池制造产能达到1,705吉瓦时，而美国仅为93吉瓦时，差距接近18倍。中国控制着全球80%的 锂离子电池组件出货量，在全球电动车电池市场占据约60%的份额。CATL和比亚迪两家中国企业合计超过全球市场份额的50%，而美国 没有任何一家企业进入全球电动车电池制造商前十。这种垄断性优势并非偶然，而是中国对锂、石墨等关键矿物精炼环节长期战略布局 的结果——中国精炼全球65%的锂、77%的钴和约96%的石墨。 "21 世纪关键技术 " 关注科技未来发展趋势，研究 21 世纪前沿科技关键技术的需求，和影响。将不定期推荐和发布世界范围重要关键技术研究进展和未来趋势 研究。 美国特别竞争研究项目（Special Competitive Studies Project，SCSP）发布年度旗舰报告《进入竞技场：中美技术竞争中谁领先 ...

Core Viewpoint - The article explores the possibility of creating a "periodic table" for intelligence, akin to Mendeleev's periodic table for chemical elements, to unify various forms of intelligence, including biological, artificial, and theoretical constructs [2][4][10]. Group 1: Background and Framework - The concept of "agents" is central to artificial intelligence but lacks a unified formal definition, leading to a fragmented understanding across different disciplines [3][17]. - The article proposes a "Minimal Complete Architecture" (MCA) for agents, defining them as open information processing systems with five fundamental functions: Input, Memory, Generation, Control, and Output [20][22]. - This architecture serves as the foundation for constructing an "intelligence capability periodic table," categorizing 243 types of agents based on their capabilities [23][26]. Group 2: Significance of the Intelligence Capability Periodic Table - The intelligence capability periodic table aims to unify various forms of existence, from inanimate objects to complex biological and artificial systems, into a single theoretical framework [39]. - It provides insights into the underlying reasons for differences in classical mechanics, relativity, and quantum mechanics by interpreting the distribution of observer capabilities within the periodic table [40]. - The table also serves as a predictive tool for unknown forms of intelligence, similar to how Mendeleev's table predicted undiscovered elements [43][46]. Group 3: Classification of Agent Types - The 243 types of agents are categorized into four groups based on their capability distributions, ranging from "absolute void" to "omniscient and omnipotent" [33][35]. - The Alpha group consists of a single member with zero capabilities, representing the most basic forms of matter [35]. - The Limited Agents group includes 31 types with at least one non-zero capability, representing common real-world systems [36]. - The Super-Limited Agents group contains 210 types with at least one infinite capability, while the Omega group represents the theoretical limit of intelligence with all capabilities being infinite [37][38].

Core Insights - The report from Contrary Research highlights that artificial intelligence (AI) is evolving from a singular technological issue to a comprehensive restructuring force affecting energy, manufacturing, defense, and human relationships [2] AI Model Competition - The report emphasizes the rapid advancements in AI foundational models, predicting that by 2030, top AI systems in software engineering, biology, and mathematics will achieve near-perfect accuracy on their respective benchmarks [3] - Key players in the AI model landscape include Google, Meta, Microsoft, and OpenAI, which have dominated foundational model releases from 2014 to 2024, with academic institutions following closely [4] Trust Issues in Evaluation Systems - The report addresses a crisis of trust in current evaluation systems, citing instances of undisclosed participation in benchmark tests and allegations of artificially inflated scores [5] - It presents a paradox in computational economics, where training power consumption has doubled every six months since 2010, while the actual computational power required for equivalent performance has significantly decreased [5] AI Commercial Penetration - As of May 2025, approximately 10% of U.S. companies have integrated AI into their products or services, while around 44.8% have subscribed to some form of AI model or platform [6] - The revenue from enterprise AI is projected to grow from $1.7 billion in 2022 to $37 billion in 2024, reflecting a growth rate exceeding three times [6] Infrastructure Competition - The demand for computational power is driving an unprecedented scale of infrastructure development, with major cloud service providers expected to spend nearly $100 billion quarterly by Q2 2025 [8] - Global capital expenditures on data centers, cloud computing, and AI-specific infrastructure are projected to reach $1.3 trillion by 2027, potentially amounting to 17% of U.S. GDP from 2025 to 2030 [8] Energy Consumption Concerns - U.S. data centers consumed 183 terawatt-hours of electricity in 2024, projected to rise to over 426 terawatt-hours by 2030, which could account for more than 10% of total U.S. electricity consumption [10] - The report highlights nuclear energy as a potential alternative, noting that most new nuclear capacity is being built in China, while the U.S. has seen a stagnation in new approvals [10] Industrial Restructuring - The report outlines a significant industrial restructuring, with China surpassing the U.S. as the largest manufacturing nation and having more robots installed than the rest of the world combined [11] - The U.S. defense industrial base is described as being in crisis, with significant shortages in key military supplies and a stark contrast in manufacturing capabilities compared to China [12] Social Trends and AI Companionship - The report discusses the rise of loneliness as a social trend, with increased solitary time among Americans, particularly among younger demographics [13] - AI companions are emerging as a response to this social void, with a significant percentage of Gen Z users expressing a belief that AI can replace human companionship [14]

Core Viewpoint - The report from the U.S. Department of Energy emphasizes that the lack of plasma measurement capabilities is a critical bottleneck in the commercialization of fusion energy, which is often underestimated [3][5][14] Group 1: Report Overview - The report titled "Basic Research Needs for Measurement Innovation" was led by top scientists from various institutions and highlights the urgent need for advancements in plasma measurement technology to achieve fusion energy commercialization by 2035-2040 [3][4][13] - The report identifies a significant data crisis, with the National Ignition Facility (NIF) generating over 100GB of data per experiment, and potential fusion power plants producing over 5PB of data per hour, which current processing methods cannot handle [6][10] Group 2: Measurement Challenges - The report outlines seven independent yet interconnected research areas, each with unique measurement challenges, emphasizing the need for real-time measurement and control in magnetic confinement fusion [8][9] - In the inertial confinement fusion (ICF) area, the report stresses the necessity for automated data processing and AI-assisted analysis to keep pace with high-frequency experimental runs [9][10] Group 3: Role of AI and Data Infrastructure - The report discusses the potential of AI and machine learning in improving plasma measurement, while also cautioning against overestimating their capabilities without proper validation [11][12] - It highlights the need for standardized data infrastructure to facilitate data sharing and AI model training across different institutions, proposing the establishment of national networks for diagnostic and calibration purposes [12][13] Group 4: Talent and Time Constraints - The report indicates a severe shortage of qualified scientists and engineers in the field of plasma diagnostics, particularly in tritium handling and accounting, which are critical for the success of fusion pilot plants [13][14] - It emphasizes the limited time window for developing measurement technologies, as many diagnostic systems require years for design and validation before the pilot plants can operate [13][14]

Core Insights - The report "State of AI 2026" by AI Technologies outlines significant advancements in AI technology over the past year and provides nine predictions for 2026, emphasizing the report's influence among global decision-makers and investors [3] Investment Trends - In 2025, global venture capital investment in AI is projected to reach approximately $200 billion, accounting for 50% of total VC investments, with 58% directed towards "super rounds" exceeding $500 million [4] - Major players like OpenAI, Scale AI, and Anthropic have secured substantial funding, indicating a "winner-takes-all" market dynamic [4] Infrastructure and Energy - The competition for computing power is intensifying, with tech giants expected to invest over $300 billion in data center expansion by 2025 [4] - Data centers are projected to consume about 10% of global energy by 2030, highlighting a mismatch in infrastructure capabilities [5] Hardware Landscape - NVIDIA maintains a dominant position in the hardware market, with projected revenues of around $130 billion in 2025, but faces increasing competition from AMD and local Chinese manufacturers [6] Model Development and Adoption - The AI model landscape is characterized by a dual dominance of the US and emerging players in China, with top closed-source models led by US firms and open-source models increasingly controlled by Chinese companies [7] - Despite high adoption rates of generative AI in enterprises, the success rate of AI projects in production remains low, with a failure rate of 88% to 95% for pilot projects [8] Consumer Market Dynamics - OpenAI leads the consumer market with 5.5 billion monthly visits, while the AI companion market has reached a size of $32 billion, growing at a CAGR of 30% [9][10] Security and Governance Challenges - Cybersecurity incidents have escalated, with DDoS attacks increasing by 30% and ransomware attacks by 32% in 2025 [11] - The global regulatory landscape is becoming increasingly fragmented, with the EU establishing comprehensive AI regulations while the US adopts a more relaxed approach [12] Geopolitical Context - The report highlights the ongoing technological rivalry between the US and China, particularly in areas like chip export controls and data sovereignty [13] Robotics and Scientific Advancements - The robotics market is expected to exceed $200 billion, with significant advancements in AI applications in scientific research and drug approval processes [14] Future Predictions - Key predictions for 2027 include a decline in NVIDIA's market share, increased investment in AI applications, and a significant rise in the robotics sector [15][16]

Core Viewpoint - The report redefines the concept of "smart economy" as "capability economy," marking a significant shift from previous narratives that viewed it merely as an advanced stage of the digital economy or as AI empowering traditional industries [2][3]. Policy Narrative and National Strategy - The report outlines a clear trajectory of policy evolution in China's AI strategy, highlighting key milestones from 2017 to 2026, culminating in the recognition of "building a new form of smart economy" as a national development coordinate [4][6]. Industry Classification - The report introduces the concept of a "fourth industry," asserting that AI-driven smart technology has the potential to become an independent sector, distinct from traditional primary, secondary, and tertiary industries [7]. AI Industry Landscape - The report describes the competitive landscape of China's AI industry, characterized by a "four big + six small tigers" model, indicating a shift towards international capability output [8]. Characteristics and Mechanisms of Smart Economy - The report identifies 20 fundamental characteristics of the smart economy, including the potential for exponential productivity liberation and the risk of a "fifth decoupling" between humans and labor [9][10]. Governance Framework - The report proposes a governance framework consisting of 21 principles aimed at ensuring that AI development is human-centered and addresses global public affairs, social responsibility, and ethical considerations [11]. Conclusion - The report emphasizes that the future of the smart economy will be shaped by market forces and collective choices, highlighting the importance of institutional design and social consensus alongside technological breakthroughs [12].

Core Viewpoint - The article emphasizes the evolution of Physical AI, which signifies a transition from digital space to the physical world, requiring robust systems capable of executing real-world tasks safely and effectively [2][3][4]. Group 1: Transition from Digital to Physical - Physical AI represents a significant shift in technology, moving from generating information to executing actions in the real world, which necessitates a system with strict safety mechanisms due to the low tolerance for errors in physical environments [2][3]. - The integration of large models with physical devices like robotic arms or autonomous vehicles poses engineering challenges due to unpredictable physical conditions, requiring high robustness in systems [3][4]. Group 2: Five-Dimensional Core Capabilities - The white paper outlines a five-dimensional framework for Physical AI, which includes perception, decision-making, verification, execution, and system feedback, forming a tightly coupled system for reliable operation in complex environments [3][4]. - Perception in Physical AI goes beyond simple object recognition to actively output structured features for physical operations, marking the starting point for machines to understand three-dimensional environments [3]. Group 3: Decision-Making and Safety - The decision-making layer translates high-level tasks into executable instructions, with large language models serving as tools for intent understanding, while strict physical constraints govern machine control [4]. - The verification process is crucial, as the costs of trial and error in irreversible real-world scenarios are high; thus, systems must filter dangers in virtual simulations before real-world execution [4]. Group 4: Execution and Feedback Mechanisms - The execution phase involves converting abstract strategies into precise mechanical movements, overcoming mechanical errors, and adapting to dynamic load variations [4]. - A feedback module transforms physical execution results into usable data, enabling continuous learning and evolution of the system, distinguishing Physical AI from traditional automation [4]. Group 5: Paradigm Shift in Core Technologies - The performance of Physical AI relies on breakthroughs in several intelligent cores, including strategy models that map high-level planning to specific action control [5]. - The world model is key for cognitive leaps, allowing systems to predict physical consequences of actions in a multi-dimensional digital space, reducing reliance on extensive real-world interaction data [5]. Group 6: Data Generation and Simulation - Developers can now automate the construction of physical work scenarios, generating synthetic training datasets with precise physical parameters in a short time [6]. - Digital twin platforms facilitate real-time synchronization between high-fidelity virtual testing environments and actual device operations, requiring significant upgrades in computational infrastructure [6]. Group 7: Safety and Control Mechanisms - Real-time local inference and closed-loop control must be integrated into end devices to handle unexpected physical situations effectively [7]. - End devices are equipped with independent safety monitoring programs that can trigger emergency stops if any parameters exceed physical limits, ensuring safety even in extreme conditions [7]. Group 8: Industry Ecosystem and Future Directions - Physical AI is creating a vast new industry chain, from foundational computational infrastructure to specialized commercial solutions, with China having advantages in implementation scenarios and hardware supply chains [8]. - In heavy industrial manufacturing, Physical AI is driving a shift from rigid automation to adaptive flexible production, enabling real-time understanding and adjustment of complex processing intentions [8]. Group 9: Intelligent Environments and User Experience - Static environments are transforming into intelligent spaces with holistic physical perception, allowing proactive management of facilities based on real-time data [9]. - The integration of various systems into a cohesive intelligent network enhances operational efficiency and user experience, marking a significant leap in physical technology [9]. Group 10: Challenges and Market Viability - Successful commercialization of breakthrough technologies requires clear economic calculations and a balance between cost control and technological innovation [10]. - The digital transformation driven by Physical AI is just beginning, demanding respect for the constraints of the physical world while pushing for rapid advancements in production capabilities [10].