刘泽龙 长江证券通信分析师: 各位投资人大家晚上好，我是长江通信的分析师刘德龙，给大家汇报一下我们对整个 26 年升腾行业的一个观点和展望。首先我们在整个 26 年是非常看好整个升腾及产业链后续 的这个发展和上涨的。那首先的话，我们看好它主要基于这个，一方面我们非常看好今年 国产算力整体的贝塔。在春节期间大家也看到了一个非常明确的产业变化，叫 ToG 出海。 其实这个出海，背后代表了两件事情。一个是我们目前，其实就两个事，一个是性能够， 一个是中国的模型出海的话，单位 Tokens 的价格比较便宜。 性能够的话，我们更多会觉得，这体现在这个当你模型的性能迭代有了一定放缓之后，中 国的性能差距相当于海外没有那么大，中国厂商觉得是能够海外客户是觉得是能够满足需 求，同时的话，那我们会看到中国的 token 的单位价格是便宜的。可能大 家会有疑惑， 说我们国产算力为什么你如果按理说单位的算力下的价格，其实不如海外的英伟达呀，然 后包括这些厂商有那么便宜的价格但是我们会有更多的运营成本的优势。包括一些配套的 零部件，包括一些供电，包括一些人员维护，我们觉得把这些成本综合起来看的话，其实 在一定程度上可以抹平国产算力和 ...

关于我们 联系方式 网站声明 用户反馈 网站地图 ID: 100104 帮助 CUILIT 首页 Faris 旗 FM 156 H T 狙击龙虎榜】情绪与指数有所背离市场或未走出调整 深 t 受益于Al等新兴产业长期需求访动的品种 黄河旋风：2月23日Akash Systems 宣布, 已向印度主权云服务商 NxGen Al Pvt Ltd 交付全球首批搭载 Diamond Cooling® 技术的英伟达GPU服务器。这是全球首次将"金刚石导热技术"正式部署于商用AI服务器体系。早在2024年12月，Akash Syste ms就已与印度云服务商NxtGen签署了一份价值2700万美元的合同，计划部署金刚石冷却服务器，所以这次交付并非一次性的 产品测试,而是有长期合同支撑的商业化落地,具备重要意义。黄河旋风已成功研制出国内可量产的最大8英寸金刚石热沉 片,计划于2026年2月投入量产。其多晶热沉片已通过华为验证,并进入"6-8英寸晶圆级"量产阶段。未来最大的预期是与中 国"超赢钻石科技"一起切入英伟达散热供应链。 网宿科技：Token的成本结构中，电力和算力占比超过70%。当一个美国开发者调用中国大模型时，数 ...

我身边有位朋友，前阵子看到AI板块大涨就急忙跟进，结果刚买入就遇上调整，亏了不少。事后复盘才发现，当 时看似强势的上涨，其实早已被「获利回吐」行为主导，只是他只看了股价涨跌，没看懂背后的交易行为。很多 时候，机构资金会利用上涨的表象"演戏"，即便在高位派发筹码，也会维持股价涨势，让普通投资者误以为行情 还能延续，傻乎乎地接盘。 看图2： 近期A股市场热点轮动快得像坐过山车，AI硬件板块突然爆发，CPO、液冷服务器等细分领域掀起涨停潮，电网 设备板块也跟着火了一把。不少人看着盘面涨跌心跳加速，涨了就追，跌了就跑，结果往往是追在高位、割在低 点，反复被市场"教育"。其实很多人都忽略了，市场的本质不是简单的买卖博弈，而是由无数复杂交易行为构成 的多维系统。利好未必涨、利空未必跌的背后，藏着资金真实的交易意图，而量化大数据，正是帮我们跳出涨跌 迷局、看清市场真相的关键工具。 一、跳出涨跌，看见真实交易行为 大多数人对交易的理解还停留在"买的人多涨、卖的人多跌"的表层，但通过量化大数据对交易行为的长期跟踪， 能发现市场里的交易远不止这两种，核心可归纳为四类：「做多主导」代表资金积极参与行情；「获利回吐」是 资金少量兑现 ...

Core Viewpoint - AGM Group Holdings Inc. is strategically positioning itself to transition from an AI hardware supplier to a comprehensive AI + Web3 ecosystem platform through initiatives in Real World Asset (RWA) tokenization, mass production of Kraken All-Flash AI Servers, and a partnership with MusicDog.ai [1][2][5] Group 1: RWA Financing - The company signed a Memorandum of Understanding (MOU) with Amber Premium to explore RWA tokenization of its real-world assets, including high-performance servers and data center assets [3] - This partnership aims to leverage Amber Premium's blockchain technology to activate new asset tokenization pathways [3] Group 2: Kraken Servers - AGM launched the ValleyVerse Kraken All-Flash Cluster Storage Server, designed for generative AI and high-performance computing, featuring ultra-high IOPS and superior scalability [4] - The company plans to achieve mass shipments of Kraken servers in Q2 2026, generating revenue through direct sales and partnerships [6] Group 3: MusicDog.ai Partnership - A Letter of Intent (LOI) was signed with MusicDog.ai to enhance collaboration in three areas: cash flow improvement, hardware revenue realization, and computing power for AI music applications [5] - The partnership aims to develop audio-specific ASICs and explore blockchain-based copyright mechanisms for AI-generated music [6] Group 4: Strategic Vision - The CEO emphasized the integration of RWA, Kraken servers, and MusicDog.ai as part of a broader strategy to drive AI innovation and build a Web3 ecosystem [6] - The company is focused on creating a sustainable strategic value for shareholders through its "Cash Flow + Hardware Infrastructure + AI Smart Applications" model [2][7]

Core Viewpoint - The A-share market showed a collective increase in the morning session, indicating positive investor sentiment and market activity [1] Market Performance - The Shanghai Composite Index rose by 0.12% to 4137.06 points [1] - The Shenzhen Component Index increased by 0.81% [1] - The ChiNext Index saw a rise of 1.18% [1] - The North China 50 Index gained 0.49% [1] - The total trading volume in the Shanghai and Shenzhen markets reached 1.34 trillion yuan, an increase of 30.5 billion yuan compared to the previous day [1] - Over 2700 stocks in the market experienced an increase [1] Sector Performance - Strong performance was noted in sectors such as power grid equipment, small metals, and AI hardware [1] - Adjustments were observed in sectors including film and television, tourism, and banking [1]

AI 陪伴赛道复盘：2026 年了，为什么还没有一款千万级 DAU 的产品跑出来？ 2025 年到现在，各种 AI 硬件大爆发，录音卡片、智能眼镜、AI 毛绒玩具、智能指环，各种新硬件产 品层出不穷。 但现实是：绝大多数用户对当下的 AI 硬件普遍不满意。在 Founder Park 的调研中，「不尽如人 意」、「做得不够好」几乎是被频繁提到的。 可以说，整个 AI 硬件市场，仍然还在一个极早期的阶段。2% 的技术尝鲜者和 10% 的早期采用者，目 前几乎所有的 AI 硬件产品，都还在这两类早期用户中打转。 所以，在第三期 AI 产品市集 Meetup 上，我们决定回到原点，和 AI 硬件创业者，以及深度使用过 AI 硬件的专业用户们聊聊： AI 硬件的不可替代性究竟在哪？在手机这个无比强大的计算中心面前，独立的 AI 硬件凭什么能让用户 额外付费且持续使用？ 以下，是我们整理的 Meetup 现场一些核心观点与讨论。 过往 Meetup 集锦： ⬆️ 关注 Founder Park，最及时最干货的创业分享 Meetup 是「AI 产品市集」的线下延伸活动，每期聚焦一个垂直赛道，邀请一线创业者、资深从业者与核 ...

Core Insights - Rabbit, once hailed as a leading AI hardware product, faced a downfall due to high return rates and cash flow issues, highlighting the challenges in meeting user expectations in the AI hardware market [1][2] - The current AI hardware landscape is characterized by transitional products that are not yet ready for mass adoption, indicating a need for further innovation and development [2] Group 1: Market Challenges - Rabbit's initial success was overshadowed by negative user feedback and high return rates, leading to financial difficulties for the small team behind it [1] - Many AI hardware products on the market fail to deliver on their promises, often resembling toys rather than functional devices [1][3] - The high cost of computational power and limitations in edge computing technology hinder the development of effective AI hardware solutions [3][4] Group 2: Investment Opportunities - Despite the challenges, there is a growing interest in AI hardware, with significant investment flowing into the sector, particularly towards established entrepreneurs and innovative startups [6][7] - The Chinese market presents unique opportunities due to its mature supply chain and engineering talent, making it an attractive landscape for AI hardware development [6] - Successful AI hardware products often find niche markets and differentiate themselves through unique functionalities, as seen with products like Plaud and Oura [11][13] Group 3: Future Directions - The AI hardware sector is evolving, with a shift towards products that can effectively integrate AI capabilities into everyday use, indicating a potential for significant market transformation [5][6] - Entrepreneurs are encouraged to focus on solving real user problems rather than merely incorporating AI concepts into their products [4][9] - The lack of a clear standard for successful AI hardware products suggests that companies must remain agile and responsive to market changes to thrive [10][12]

Group 1 - China successfully launched a reusable experimental spacecraft on February 7, using the Long March 2F rocket, aimed at verifying reusable spacecraft technology for peaceful space utilization [2] - SpaceX has postponed its Mars mission originally scheduled for this year, shifting focus to NASA's lunar program, with plans for an unmanned lunar landing in March 2027 [2] - OpenAI's first hardware, named Dime, is expected to be released this year, designed as an AI audio headset, with a strategic shift from a more complex device due to high hardware costs and supply chain issues [2]

Core Viewpoint - Vista Equity Partners is leading a $350 million funding round in AI chip startup SambaNova Systems, marking a shift from its traditional focus on enterprise software [1][4]. Group 1: Investment Details - Vista, in partnership with Cambium Capital, is participating in the Series E funding round for SambaNova, with Intel Corp also investing approximately $100 million, potentially increasing to $150 million [2][3]. - The funding aims to help SambaNova compete with Nvidia Corp in the growing demand for inference chips used in AI applications [3]. Group 2: Market Context - The investment comes amid a selloff in global software stocks, which have lost nearly $1 trillion in value as AI transitions from a supportive factor to a potential disruptor [5]. - Interest in AI hardware has surged, driven by increased dealmaking among companies seeking efficient chips for AI applications [5]. Group 3: Competitive Landscape - Other AI chipmakers, such as Cerebras Systems, have also raised significant funding, with Cerebras securing $1 billion at a valuation of $23 billion [6]. - OpenAI is exploring partnerships with companies like Groq and Cerebras to find alternatives to Nvidia GPUs for their computing needs [7].

Core Insights - The article emphasizes a significant shift in the AI hardware industry, indicating that only products that solve specific problems and have user willingness to pay will survive post-capital frenzy [9][12]. Market Opportunities - Focusing on specific, high-frequency, and payable pain points is more viable than creating "all-powerful" products, leading to a division between opportunity zones and concept zones [11]. - Products like AI recording pens, smart translation headphones, and pet companionship robots represent "AI enhancement of traditional categories," providing immediate and perceivable value [12]. - The AI recording and audio efficiency hardware are identified as core opportunity points for 2026, with examples like the iFlytek headphones addressing business and academic needs [12]. - The pet companionship robot market is characterized by a young female consumer base, but faces challenges due to severe functional homogeneity [13]. - A more complex product design combining robotics and emotional feedback is suggested as a way to differentiate in the pet market [13][15]. Concept Zone - General-purpose humanoid robots and all-functional AI office assistants remain in the conceptual exploration phase, often failing to meet real needs due to high costs and unclear applications [16]. - The necessity of independent AI office devices is questioned, with a potential future shift towards an "AI hub + multiple smart terminals" model [16]. Cost Control - Effective cost control is essential for profitability, with a focus on managing hardware costs, particularly motors and sensors [18]. - The current limitation is not computational power but rather the cost of hardware components, with a recommendation to utilize cost-effective domestic models [18]. - Companies must either incorporate AI functionality costs into hardware pricing or adopt a subscription model to align price and cost [19]. - The supply chain landscape is evolving, with smaller teams facing challenges in assembly and needing to innovate in branding and market positioning [20]. - The growth of domestic AI chips is driven by government and market demands, influencing the technology choices of hardware manufacturers [21]. Commercial Validation - 2026 is highlighted as a pivotal year for commercial validation in AI hardware, where success will be measured by tangible business results rather than flashy concepts [23]. - Sales data, including monthly sales and shipment volumes, are becoming essential metrics for industry players [24]. - A balanced approach between technology and sales is recommended, with a focus on practical applications and market feedback to refine products [25]. - The prevailing business model for AI hardware is expected to be a combination of one-time hardware purchases and subscription-based AI services [26]. Conclusion - The article concludes that the future winners in the AI hardware space will be those who can clearly articulate the specific problems they solve, why users will pay for solutions, and their competitive advantages [28].