来源：凯恩斯 2026年注定是不平凡的一年，不仅是十五五规划的第一年，而且是众多行业的元年。A股股民其实是最 喜欢炒概念的市场之一。有一些概念2025年作为概念已经炒的很热，比如液冷，比如商业航天，比如可 控核聚变，当然，有很多一些行业还没有产品落地，但不影响概念的炒作" 1，自动驾驶元年；2，液冷元年；3，国产HBM元年；4，端侧AI元年； 5，固态电池元年；6，AI应用元年；7，量子计算元年；8，存算一体芯片商用元年；9，类脑计算元 年；10，低空经济元年；11，商业航天元年；12，人型机器人元年；13，硅光元年；14，可控核聚变元 年。 2026年会如何？根据目前机构的宏观和行业研究报告，我总结了一下2026年大概率发生的事件和十大预 测。 一，2026年美联储降息2次以上。 从11月美国的经济数据来看，基本呈现低通胀（2.6%）高增长 （4.3%）态势，考虑到新任美联储主席上任，即使从预防和储备货币政策工具的角度，美联储上半年 也会继续降息，并且大概率上半年就会降息2次，每次25个基点。下半年则要看经济数据，有不确定 性，可能只降息1次。 二，人民币升值到6.5。 美联储降息导致美元走弱，全球金资本关 ...

Core Viewpoint - Hangzhou is accelerating the development of brain-like intelligence as a key future industry, supported by substantial policy measures aimed at fostering innovation and investment in this field [1][2]. Group 1: Policy Support - The "Measures" outline financial support for brain-like intelligence projects, including a 25% subsidy on actual funding for projects not requiring local matching funds, with a maximum of 5 million yuan per project [2]. - For small and medium-sized technology enterprises with R&D increases over 2 million yuan, a 5% subsidy on the increase is available, capped at 200,000 yuan [2]. - Larger enterprises with R&D expenditures exceeding 500 million yuan can receive a direct subsidy of 300,000 yuan [2]. Group 2: Application Focus - The policy emphasizes application-oriented support, targeting key areas such as personalized medical treatment, intelligent diagnostic assistance, and rehabilitation in healthcare [2]. - In smart applications, priority scenarios include smart homes and security, intelligent transportation and autonomous driving, and robotics [2]. Group 3: Encouragement for Innovation - The policy encourages enterprises to trial new brain-like intelligence products, offering a subsidy of up to 300,000 yuan for the first product sold to non-affiliated parties, based on 10% of sales [3]. - Companies transitioning through various growth stages can receive financial rewards, including 20,000 yuan for first-time registration and additional support for maintaining status over three years [3]. - Recognition as a national or provincial "specialized and innovative" enterprise can yield rewards of up to 1 million yuan and 200,000 yuan, respectively [3]. Group 4: Spatial Layout - The policy outlines a spatial layout for the brain-like intelligence industry, supporting the establishment of a pilot area in Yuhang District and encouraging development in Binjiang and Xiaoshan Districts to create a dual-driven growth model [3].

三航脑科学与脑技术研究中心 （ SiBST ） 成立于 2023 年 9 月 ， 是西北工业大学医学研究院下设的科 研机构和人才特区，聚焦脑科学、脑机接口、类脑计算等前沿领域。" 三航 "这一名称涵盖了航空、航天和 海洋科学，该研究中心致力于将脑科学研究与 航空、航天和海洋科学 领域相结合，近期目标是将实验鼠送 入深空，长远目标是为人类在深空或深海长期生存做好准备。我们正在全球范围内招募高水平的杰出青年 人才。 编辑丨王多鱼 排版丨水成文 1、讲席教授 职位要求 ：引领国际科学前沿，在基础研究方面取得系统性、创新性成果，在国家重大科技任务中取得标 志性科学成就。候选人应为具有较高社会声誉和广泛学术影响力的战略科学家。 支持福利 ：具体问题具体协商，根据需求提供个性化支持。 2、领军人才 职位要求 ：在其学科领域开展开创性和引领性研究，获得国内外同行的高度认可。为社会发展和科学进步 做出重大贡献，符合国家重大需求。候选人应为国家级领军人才，在学科建设、科学研究、人才培养等方 面做出重大贡献。 支持福利 ：具体问题具体协商，根据需求提供个性化支持。 3、青年人才 职位要求 ：拥有博士学位，年龄在 40 岁以下。在 ...

这样一幕令人惊叹。 随着摩尔线程、沐曦相继上市，科创板短短半个月内一举诞生两个3000亿市值IPO，创投圈为之沸腾。在此之前，寒武纪、海光信息市值均一度突破6000 亿元。至此，中国"算力四贵"合计市值已超1.6万亿元。 鲜少有人注意到，梳理寒武纪、海光信息、摩尔线程、沐曦背后，意外浮现着一个共同的身影——联想创投。复盘下来，能做到独中四元的投资机构，似 乎就只有联想创投一家。 这是中国算力崛起叙事当中的一个惊喜片段。感叹于此，我们尝试去还原这段几近被忽略的投资往事。 从寒武纪到沐曦 联想创投，独中四元 回过头看，联想创投这几笔投资在当时都投在了"寂寞无声处"。 时间回到2017年，正值移动互联网的风口，芯片算力赛道几乎无人问津，鲜少有投资人会专注于此。但彼时刚成立不久的联想创投却是一个例外——当时 他们根植于联想集团"端-边-云-网-智"新IT战略，将AI算力视为智能世界的"心脏"，判断应用导向的大计算将会使得未来十年对算力需求仍然有百倍增 长。 出手寒武纪始于一个下午。投资界曾记录这段往事：2017年，联想创投总裁、创始合伙人贺志强在中国科学院第一次见到陈天石，"话不多，典型的技术 天才气质。" 经过几次 ...

先是身披 " 国产 GPU 第一股 " 光环的摩尔线程登陆科创板，上市首日股价最高涨超 500% 至 688 元 / 股，市值一度突破 3000 亿元；紧随其后的沐曦表现更是亮眼，开盘即报 700 元 / 股，市值直接站上 3500 亿元关口。 成为千亿捕手需要什么投资逻辑？ 作者丨 王满华 来源丨 投中网 若论近期一二级市场的关注焦点，AI芯片赛道无疑独占鳌头。 将投中网设为"星标⭐"，第一时间收获最新推送 如此显著的赚钱效应摆在公众面前，人们自然好奇：谁是这波红利的受益者？我也不能免俗地梳理了一番，结果发现一个有趣 的现象：尽管两家公司背后的投资机构多达数十家，但同时押中二者的却寥寥无几。 这其实不难理解。一方面，早期 VC 向来有 " 只押龙头 " 的习惯，况且在 2024 年底之前， "IPO 不确定性 " 始终是横亘 在这个领域的一道现实难题；另一方面，两家公司的估值一直不低，份额争抢也相当激烈。对多数机构而言，能投进其中一家 已属不易。 不过， 我也发现了一个例外 —— 联想创投。尤为值得一提的是，这家机构不仅同时斩获了摩尔线程和沐曦股份的 IPO 红 利，更是在寒武纪、海光信息这两家先行上市的 ...

Core Viewpoint - The article discusses the advancements and potential of embodied intelligence, particularly focusing on the development of a "brain" for robots that can adapt and learn across various forms and tasks, highlighting the contributions of companies like Qianjue Technology and Liufeng Space [2][3][4]. Group 1: Embodied Intelligence Development - Embodied intelligence has emerged as a hot investment area, with significant advancements in creating "small brains" but challenges remain in developing a comprehensive "big brain" [3][4]. - Recent scientific research indicates substantial potential for embodied intelligence, although the foundational paradigms are still evolving [4]. - Qianjue Technology aims to create a "brain in a jar" that can be utilized by various robot forms, with plans to connect 100,000 devices to its system by next year [4][5]. Group 2: Technical Approaches - Qianjue Technology employs a decoupled approach to brain modeling, allowing for independent optimization and evolution of different brain regions, which enhances efficiency [5][14]. - Liufeng Space focuses on building world models that drive embodied brains, utilizing real-time interactive space generation technology [6][11]. - The two companies represent different paths in the development of embodied intelligence, with Qianjue emphasizing brain-like structures and Liufeng leveraging world models for practical applications [8][10]. Group 3: Data and Training - Data scarcity is a significant challenge in training embodied intelligence systems, with Qianjue Technology achieving multiple generations of pre-training, which is rare in the industry [14][17]. - Liufeng Space believes that good robot data should be treated as an asset, emphasizing the importance of diverse and abundant data for effective training [12][17]. - Both companies recognize the need for extensive data to achieve effective pre-training, with estimates suggesting that a billion clips may be necessary for comprehensive training [26][27]. Group 4: Future Outlook - The timeline for achieving a mature embodied brain technology is optimistic, with both companies suggesting that significant advancements could occur within two years [26][27]. - The potential for embodied intelligence to surpass language models is highlighted, with expectations for the emergence of numerous billion-dollar companies in this sector [27].

"人脑以仅约20瓦的功耗支撑起千亿级神经元的复杂运作，这为AI提供了极致的能效范例。借鉴人脑机制将成为突破 低功耗、长序列与通用性三大核心难题的关键。" 在"从脑机接口到脑机共生"主题论坛期间，在接受《科创板日报》等媒体采访时，对于AI与脑机接口融合的下一步发 展方向，中国科学院自动化所研究员李国齐如是阐述。 在这场会议上，天桥脑科学研究院研究院创始人雒芊芊宣布天桥脑科学研究院成立尖峰智能实验室，由李国齐团队研 发的国内首款类脑脉冲大模型"瞬悉1.0"也同步亮相。该模型在国产GPU算力平台上完成训练与推理，并与国产GPU企 业沐曦科技协同，打通了从类脑基础模型、国产算力平台到类脑芯片的全栈式研究链条。 打通国产GPU算力平台 "瞬悉1.0"被视为我国类脑计算与大模型融合的一次关键突破。该模型由李国齐团队研发，是国内首个类脑脉冲大模 型。与当前主流基于Transformer架构的大模型不同，类脑模型借鉴人脑以脉冲形式进行信息传递和处理的机制，试图 从根本上破解高能耗、长序列建模和泛化能力受限等问题。 李国齐向《科创板日报》等媒体介绍，发现式智能的一个关键能力是神经动力学。人脑以仅约20瓦的功耗支撑起千亿 级神经 ...

Core Insights - The research team at Jiangnan University has made significant advancements in multifunctional optoelectronic devices, proposing a device scheme that integrates neuromorphic computing and hardware encryption imaging through bias control [1][2] - This innovation addresses the challenges of achieving high sensitivity UV light detection, brain-like information processing, and robust security encryption within a single device, overcoming the limitations of traditional multi-component systems [1] Summary by Sections Technological Advancements - The team developed a device based on AlScN/GaN heterojunctions, enabling reconfigurable functionality through precise bias control [1] - At low bias, the device functions as a high-speed, highly sensitive UV detector, while at high bias, it exhibits significant persistent photoconductivity, simulating biological synaptic learning and memory [1] Security and Integration - The research introduces a dynamic bias protocol that acts as a physical "key" for hardware-level encryption of images, enhancing information security [1] - This technology is expected to pave the way for integrated detection, computation, storage, and encryption within a single device, providing a new physical layer solution for information security [2] Future Applications - The technology is anticipated to be applied in fields requiring high security and integration, promoting the development of safer and more efficient intelligent hardware systems by deeply integrating neuromorphic computing with sensor encryption [2]

Group 1 - The 2025 World Computing Conference was held in Changsha on November 20-21, showcasing the rapid increase in global computing power and the widespread application of large models in consumer products such as smartphones, cars, and smart home devices, leading to significant changes in production and lifestyle [1][2] - The conference highlighted ten global computing innovations, including the exponential growth of global computing capabilities, the acceleration of generative AI applications, and the commercial deployment of neuromorphic processors, marking a new era in brain-like computing [2][3] - Advanced computing-driven intelligent factories reported an average reduction of product development cycles by 28.4%, an increase in production efficiency by 22.3%, a decrease in defect rates by 50.2%, and an average reduction in carbon emissions by 20.4% [4] Group 2 - The strategic importance of advanced computing in reshaping the global industrial landscape is increasingly recognized, with the computing technology evolution driving significant changes in the information technology industry and the digital economy [4] - The World Computing Conference has evolved into an international and professional platform for communication in the computing field since its establishment in Changsha in 2019, now hosting its seventh edition [4]

在过去20多年时间里，金耀初一直在思考从自然演化和生物发育角度来研究、理解人的智能的可能性。 在他看来，人的神经系统和形态是协同发育，神经系统的发育过程跟环境、形态是密不可分的，这与近 期比较热的具身智能在原理上是相似的。 例如，当演化一个有不同模块、不同结构深度的脉冲大模型时，技术人员往往会基于Transformer架构加 入一些类脑的机制，实现不同的通道，有自上而下的通道，也有自下而上的通道，也有注意力机制的调 整，使得结构更灵活。这与人脑内有不同通道，也有很多神经调控机制，有着很大的关联。 大会上，金耀初还发布了西湖大学人工智能领域最近几年的研发成果。据他介绍，在具身智能领域，西 湖大学在大模型的辅助下，目前已经能够做到更加自主地感知、自主决策和自主执行。（文猛） 责任编辑：常福强 炒股就看金麒麟分析师研报，权威，专业，及时，全面，助您挖掘潜力主题机会！ 新浪科技讯 11月9日下午消息，2025年世界互联网大会乌镇峰会期间，欧洲科学院院士、西湖大学人工 智能讲席教授金耀初指出，"传统大模型有很多局限性，包括能耗大，没有自主学习的瓶颈等，基于类 脑神经网络，将类脑计算与具身智能结合，才能让类脑大模型真正具 ...