企查查APP显示，近日，南宁润云智算科技有限公司成立，注册资本1000万元，经营范围包含：人工智 能应用软件开发；互联网数据服务；人工智能基础软件开发；信息系统集成服务；人工智能公共数据平 台等。企查查股权穿透显示，该公司由润建股份等共同持股。 ...

RT Matteo Franceschetti (@m_franceschetti)BREAKING: We just raised $100M at @eightsleep to build the AI that finally fixes sleep.For decades, sleep has been passive. You lie down, close your eyes, and hope for the best. We’re changing that.Over 1B hours of biometric sleep data → distilled into fully autonomous AI that knows your body better than you do.We will do this in two ways:Step 1: Your personal Sleep Agent, running infinite nightly scenarios before they happen to deliver the most restorative sleep yo ...

中商产业研究院近日发布报告称，人工智能发展已进入关键转折期，AI智能体正以前所未有的速度从 实验室走向产业核心。2024年全球AI智能体的市场规模约为51亿美元，并预测2025年全球AI智能体市 场规模将达113亿美元，2030年将接近500亿美元。 另据央视新闻报道，一批重大科技产业项目在上海自贸区临港新片区集中签约落地，项目涵盖集成电 路、高端装备、汽车软件、人工智能等方向，总投资额超400亿元，包括三维芯片集成项目、重型燃气 轮机项目、脑机接口验证平台项目、水下机器人项目等六个项目集中签约落地。 【环球网财经综合报道】日前，上海市经济和信息化委员会 上海市发展和改革委员会 上海市国有资产 监督管理委员会关于印发《上海市加快推动"AI+制造"发展的实施方案》（沪经信软〔2025〕556 号），宣布实施"模塑申城·AI+制造"行动，推动人工智能技术与制造业深度融合，加快赋能新型工业 化，形成新质生产力。 实施方案提出，通过三年时间努力，本市制造业智能化发展水平进一步提升，在语料、模型、平台、场 景等领域形成一批创新成果。推动3000家制造业企业实现智能化应用；打造10个行业标杆模型，形成 100个标杆智能产 ...

Group 1 - Fujian Zhican Ark Technology Co., Ltd. has undergone a business change, with SenseTime Group's wholly-owned subsidiary, Shanghai SenseTime Intelligent Technology Co., Ltd., added as a shareholder [1] - The company was established in 2024 with a registered capital of 100 million yuan [1] - Its business scope includes artificial intelligence industry application system integration services, big data services, data processing services, and artificial intelligence application software development [1]

公众号记得加星标⭐️，第一时间看推送不会错过。 多个标准组织正在积极规划 448G 电气物理层 (PHY) 的发展路径。光互联网络论坛 (OIF) 于 2024 年 7 月启动了 CEI-448G 框架项目，为定义信道特性、调制目标和可达性目标奠定了基础。IEEE P802.3dj 工作组正在将以太网标准扩展至 1.6T 和每通道 200G，并以 448G PHY 为关键构建模块。 超级以太网联盟 (UEC) 和 UALink 正在使电气接口规范与人工智能规模架构要求保持一致，而存储 网络行业协会 (SNIA) 正在举办研讨会，汇聚人工智能、存储和网络领域的观点。开放计算项目 (OCP) 持续推动以部署为导向的规范，解决超大规模采用的外形尺寸、集成模型和运营考量。 这种协作环境确保最终规范在技术上是稳健的并且易于部署，即使每个组织都有独特的侧重点——无 论是以太网协议合规性、电气互操作性、光电链路、人工智能优化还是系统集成。 来源 ：内容 编译自semiwiki 。 迈向更高速网络的步伐始终秉持着相同的核心目标：提高数据速率、降低延迟、提升可靠性、降低功 耗，并在控制成本的同时保持或扩展覆盖范围。对于下一代高 ...

Core Viewpoint - The demand for printed circuit boards (PCBs) is surging due to the accelerated construction of AI computing power infrastructure, with significant growth expected in high-layer boards, high-density interconnect (HDI) boards, and IC substrates. Domestic manufacturers are actively expanding high-end capacity, and it is anticipated that domestic PCB equipment manufacturers will seize the AI PCB boom, validating emerging technologies and capturing incremental demand, leading to an increase in domestic market share and value enhancement [1]. Supply and Demand Trends - Structural demand for high-layer boards, HDI boards, and IC substrates is robust due to AI servers. The overall demand for AI PCBs is outpacing supply, with a projected compound annual growth rate (CAGR) of 11.6% for global market server/storage PCBs from 2024 to 2029, driven by AI demand. The planned output value for high-layer boards, HDI boards, and IC substrates is growing rapidly, with an estimated investment of 41.9 billion yuan from leading domestic AI PCB companies by 2025-2026 [1][2]. AI PCB Upgrades - The usage and specifications of PCBs are evolving due to AI servers. For instance, the configuration of Nvidia's NVL36 includes 36 GPUs, 18 CPUs, and 9 NVSwitches, with even higher usage in the NV72 model. The demand for high-layer and HDI PCBs is increasing, leading to higher requirements for wiring, copper plating, and blind/buried holes [2]. Beneficiaries in AI PCB Equipment - The increase in layers and precision in PCBs is benefiting exposure, drilling, electroplating, and inspection equipment. According to Prismark, the global market value shares for these equipment types are 17%, 21%, 7%, and 15%, respectively. The demand for more precise wiring is raising requirements for LDI equipment, while higher layer counts and complex HDI structures are increasing the need for micro-drilling and laser drilling technologies. The uniformity requirements for micro-hole filling are also rising, with expectations for VCP penetration rates to improve from 55% [3][4].

Core Insights - Meta is set to launch new smart glasses named Hypernova in September, starting at a price of $800, down from an initial plan of $1,000 [2][3] - The Hypernova glasses will feature a display for apps and alerts, and will be controlled via a wrist accessory, indicating a shift towards glasses potentially replacing smartphones in the future [3][4] Pricing and Market Context - Current Meta Ray-Ban glasses are priced between $200 to $400, while Oakley smart glasses can cost up to $500, positioning Hypernova at a premium price point [2] - The price reduction for Hypernova suggests a strategic move to capture a larger market share in the smart glasses segment [2] Technological Features - Hypernova is rumored to include a smartphone-quality camera and a voice-activated AI tool, reflecting advancements in wearable technology [4] - The glasses are designed to work in conjunction with mobile phones, hinting at a future where they may serve as standalone devices [3][4] Industry Trends - EssilorLuxottica reported a more than 200% increase in sales of Ray-Ban Meta glasses in the first half of the year, indicating strong market demand for smart eyewear [5] - The convergence of AI, sensory technology, and healthcare data is seen as a transformative trend in the eyewear industry, as stated by EssilorLuxottica's leadership [5] Competitive Landscape - Google is also entering the smart glasses market with its Android XR operating system, partnering with eyewear brands to create AI-equipped glasses, which may intensify competition in the sector [6][7] - The vision of AI assistants integrated into eyewear is a common theme among tech companies, suggesting a growing trend towards augmented reality and smart devices [7]

| 机器之心报道 | | --- | | 编辑：Panda | | 昨晚，深度求索在用户群里宣布「DeepSeek 线上模型版本已升级至 V3.1，上下文长度拓展至 128k」并更新了 UI （去掉了 DeepThink 旁的 R1 标示）之后，在 Hugging Face 发布了一款新模型 DeepSeek-V3.1-Base 。 | | 模型地址：https://huggingface.co/deepseek-ai/DeepSeek-V3.1-Base | | 从名字也能看出来，该模型是 DeepSeek-V3 系列最新的基础模型。至于为什么命名为 V3.1，而不是像之前以前命名为 V3 带四位日期数字的形式（如 V3-0324）， | | 尽管社区有诸多猜测，但深度求索官方尚未给出明确说明 —— 和该公司之前的操作一样，这一次同样是模型先行，说明和宣传还在后面。 | | 该模型一发布就吸引了 AI 社区的广泛关注，短短几个小时就已经冲上了 Hugging Face 热门模型榜第 4 位！ | 具体技术指标上，DeepSeek-V3.1-Base 与 DeepSeek-V3 差别不大，同样的参数量，采用了 ...

Fake Job Seekers Are Exploiting AI To Scam Job Hunters And Businesses https://t.co/Nxp0Aik874 https://t.co/Nxp0Aik874 ...

.@Meta confirmed that it is breaking up its AI Superintelligence Labs into four divisions focused on research, infrastructure, and product development, part of a broader effort to accelerate progress toward so-called “superintelligence." https://t.co/gcGdOFK3cP ...