Skip to content

#行业动态

7 today

Nov 2Monday

TechCrunch · AI

Omen AI adds liquid know-how in $16 million Aquatic Labs acquisition

Omen AI 以 1600 万美元现金加股票收购竞争对手 Aquatic Labs,将后者检测水体碱度的传感器与自身检测水中元素的传感器整合为单一实时监测设备,预计明年初上市。Omen 2024 年成立,6 月刚完成 3100 万美元 A 轮融资;其传感器年底将监测 250 至 300 MW 功率的芯片,目标 2027 年底覆盖相当于 1 GW 的芯片。

OpenAI News

Introducing the Grid Protectors Program

OpenAI推出全国性Grid Protectors Program,计划让未来计算园区在电网承压时降低用电需求,支持社区可靠、可负担的电力供应。拟建的佐治亚州Effingham County园区将作为试点,与Georgia Power的协议包含最高1,000 megawatts的灵活需求响应,OpenAI承诺承担项目所需的全部基础设施及供电服务成本。

Hacker News front page

Anthropic amends S-1, disclosing $17B Q3 revenue and $895B in committed spending

Anthropic's amended S-1 filing shows about $17 billion in third-quarter revenue. Purchase commitments and lease agreements together total roughly $895.4 billion in committed spending.

Why it matters: The filing pairs Anthropic's fast revenue growth with the infrastructure spending it has already committed to, giving a rare look at the cost side behind the numbers.

The Decoder

Saudi Aramco takes 49 percent stake in AI firm Humain with $2.6 billion deal

沙特阿美将收购 AI 公司 Humain 的 49 percent 股份,并在 2026 年投入约 $2.6 billion 及自有 AI 技术。Humain 由沙特主权财富基金 PIF 于 2025 年 5 月成立,建设数据中心、开发云服务并训练 ALLaM 等模型,计划到 2030 年建成 three gigawatts 的数据中心容量。

TechCrunch · AI

OpenAI might rebrand its data centers as ‘AI compute campuses’

OpenAI 正筹备中期选举后的公关攻势,其首席全球事务官 Chris Lehane 提出将数据中心重新包装为「AI 算力园区」。他还讨论了为 AI 设定「限速」、推动美国参与制定全球标准等议题,并称围绕数据中心的争论「良性且健康」。

Hacker News front page

Seoul Viosys begins production of rare-earth-free white light emitter

Seoul Viosys 已开发并量产不含稀土元素的白光 LED 芯片 WICOP 3,芯片尺寸可小至 10µm,为首款小于 100µm 的白光光半导体。首批量产中 25µm x 10µm 芯片输出 0.015lm 白光,发光效率已接近 100lm/W,目标两年内达到 200lm/W,制造成本约为传统白光 LED 的一半。

Hacker News front page

Seoul Biosys Begins Production of Rare Earth-Free White Micro LEDs

Seoul Biosys 已基于 WICOP 技术开发并量产第三代光学半导体,无需稀土荧光粉和封装即可产生白光。首个量产产品为 25µm x 10µm 的 micro LED,白光区亮度 0.015 lm,光效约 100 lm/W,公司目标两年内达到 200 lm/W,并将制造成本降至当前一半。该技术计划扩展至 AR 显示和 AI 数据传输的光通信应用。

Nov 1Sunday

The Decoder

OpenAI is steering ChatGPT users toward more paid usage

OpenAI 为 ChatGPT 增加支付卡和设备限制,并引导用户购买额外额度或用量重置以获取更多用量。限制适用于 Go、Plus、Pro 100、Pro 200 和 Pro 500,免费账户及 Business、Enterprise、Edu 不计入;达到设备上限的账户无法继续在该设备使用,退出登录也不会自动释放名额。

Hacker News front page

Anthropic takes steps to block Claude from reverse-engineering video games

Tom's Hardware 报道称,Anthropic 已悄然开始限制 Claude 参与游戏代码修改、反编译和逆向工程。Anthropic 尚未正式公布原因,材料将这些限制与 AI 辅助游戏移植、破解引发的版权及诉讼风险联系起来。材料还指出,违规使用者可能转向尚未采取同类防护措施的其他大语言模型。

TechCrunch · AI

OpenAI reportedly lowers funding target for new round

OpenAI 据报将新一轮融资目标从 $30 billion 下调至 $20 billion,首批交割金额预计更低。《金融时报》称本轮由阿联酋投资者群体领投,后续面向富裕个人的发售可能帮助达到目标;《华尔街日报》称已筹得 $11 billion,部分大型投资者反应冷淡。报道将融资困难部分归因于华尔街对高额 AI 基础设施支出的担忧,TechCrunch 已向 OpenAI 寻求回应。

Oct 31Saturday

Hacker News front page

Cathedral breaches Anthropic systems in security test, reaching staff data and internal messages

Security startup Cathedral used an AI agent to reach sensitive internal data at Anthropic in under 12 hours, as part of Anthropic's external researcher security testing program. A draft report reviewed by The Washington Post says the data included employee records, salaries, offer letters, meeting notes and Slack messages.

Why it matters: An agent found flaws in Anthropic's business software, and third-party software bugs exposed staff records and internal messages.

The Decoder

Microsoft offers voluntary buyouts to about a quarter of its staff, exempting key AI units

Microsoft is offering voluntary buyouts to about 60,000 employees, roughly a quarter of its workforce, while excluding several units important to its AI strategy. The plan covers vice president level and below in the US, Canada, India and the UK. Xbox, Windows, Copilot and finance are included; data centers, sales, cybersecurity and quantum computing are excluded.

Why it matters: About a quarter of Microsoft's staff can take a buyout, with Copilot in scope but data centers and other AI-strategy units left out.

The Decoder

OpenAI raises about $11 billion of a $30 billion target

OpenAI's latest funding round has brought in about $11 billion, less than half of its $30 billion goal, according to The Wall Street Journal. The round is valued at $1.25 trillion, below the $1.4 trillion target. The report attributes the shortfall to weaker interest from major investors and a tech selloff after soft quarterly results from Google, Amazon and Intel.

Why it matters: OpenAI missed its funding target and is turning to family offices and wealthy individuals as major investors pull back.

AI HOT · Industry

Anthropic IPO filing discloses nearly $900B in off-balance-sheet spending commitments

Anthropic's IPO filing discloses close to $900 billion in future spending commitments on servers, buildings and power. The filing lists $117.5 billion in on-book liabilities and $112 billion in cash. Off-balance-sheet obligations cover future leases and agreements to buy data center capacity, power and GPUs. Anthropic declined to comment on the size of the capital commitments and plans to cover the investments with revenue growth and fundraising.

Why it matters: Anthropic's filing puts future lease and compute-purchase commitments off the balance sheet at nearly $900 billion, far above the liabilities it reports on its books.

TechCrunch · AI

Nvidia in talks to back Nscale's West Virginia AI data center debt

Nvidia is negotiating to guarantee billions of dollars in debt financing for an Anthropic data center Nscale plans in West Virginia. The proposed arrangement is a residual-value guarantee: if Nscale defaults on construction payments, Nvidia would take over the facility temporarily and help find a buyer and new tenants.

Why it matters: Nvidia's financial support for cloud customers now reaches construction risk, beyond leases and chip purchases.

TechCrunch · AI

Ollie, a privacy-focused AI assistant, is shutting down

隐私优先的 AI 助手 Ollie 将于 11 月 6 日关停,CEO Bill Lennon 在 X 上宣布了这一消息。该产品曾获 Khosla Ventures 领投的 750 万美元种子轮,可连接用户邮箱与日历,并提供预约、备餐和账单支付功能,还通过了 SOC 2 隐私合规认证。Lennon 称随着 Muse、Instinct 等竞品入场,公司难以脱颖而出并继续融资。

Ars Technica · AI

Backdoored Anthropic VS Code extension sat in Cursor's marketplace for nearly 12 hours

A malicious version of Anthropic's official Claude VS Code extension, 2.1.313, was planted in Cursor's marketplace and reached up to 5,000 users. Researcher Kevin Beaumont says Anthropic pushed 2.1.314 hours after the bad build went live. Cursor auto-updates extensions by default, though users can change that in settings. Anthropic had not warned customers at publication, and neither side has explained how attackers took over the official account.

Why it matters: Attackers shipped a malicious extension through Anthropic's official Cursor channel, and default auto-update carried it to as many as 5,000 users.

TechCrunch · AI

Intel’s stock crashes, knocking $12B from government’s stake

Intel第四季度业绩展望低于预期,周五股价暴跌约28%,美国政府所持股份价值缩水$12 billion。公司预计第四季度收入为$15 billion至$16 billion,低于分析师预期的约$17 billion,毛利率将从43%降至37%,且因大规模投资预计无法产生正现金流。

Hacker News front page

Embraer's Atech aims to double in size by 2030 with drone traffic control

Embraer 控股的 Atech 计划到 2030 年将公司规模翻倍,依托无人机交通管理(UTM)等核心技术拓展海外市场。其空中交通管理(ATM)方案已在南非落地交付,此前还出口至印度、巴拉圭等市场。Atech 正关注亚洲、非洲和拉丁美洲的空中交通管理系统现代化及防务安全需求。

The Verge · AI

Google’s satellite AI server transmits its first response to Earth

Google 的 Project Suncatcher 卫星成功启动 TPU,由搭载的 Gemma 4 生成关于身处太空的俳句,并向地球传回首条响应。Google 还测试了太空主题打油诗和遥测提示词,该卫星于本月早些时候进入轨道。后续测试将评估 TPU 在轨道热应力和辐射暴露下的表现,为 2027 年两颗互联卫星的下一阶段目标做准备。

OpenAI News

Making it easier for websites to express AI training preferences

OpenAI 宣布将遵循 robots.txt 中一项通用信号,让网站运营方一次性向多家 AI 提供商表达内容用于基础生成式 AI 训练的选择。该信号源自 IETF AI Preferences 工作组,OpenAI 计划在未来几个月内实施,现有爬虫控制规则继续有效。

Hacker News front page

Proton's CEO Cancels AI 'Re-Interviews'

Proton CEO Andy Yen 周五取消了近乎覆盖全公司的 AI 使用情况重新面试,此前员工担忧使用不足可能导致解雇。404 Media 的调查披露了公司推动 AI 编码时的安全争议,Yen 承认沟通失误,同时表示将继续推进 AI,下月重组不会因 AI 导致岗位流失。员工对取消安排反应不一,其中一人称原定面试的衡量项将成为绩效考核的一部分;Proton 未回应置评请求。

TechCrunch · AI

Google’s first data center satellite is online and writing poetry

Google Suncatcher 太空数据中心原型卫星已上线并与地面通信,搭载的 Google DeepMind Gemma 4 本周按提示写了一首太空主题俳句。卫星配备 4 个 TPU,发电功率最高为 1 kW,但供电限制使其每次只能运行 1 个 TPU,持续约 20 分钟。后续测试将检查芯片在热应力与太空辐射环境下的表现,评估辐射引发的位翻转对推理工作负载的影响。

Hacker News front page

Tracks along the longest drive on another world

NASA 毅力号在火星 Sol 2006(2026 年 10 月 11 日)里程表达到 45.22 公里(28.10 英里),创下地外天体最远驾驶纪录,其地球工程双胞胎获吉尼斯世界纪录牌匾。

Hacker News front page

US tech stocks slide on AI buildout cost fears; Nasdaq opens down 2.5%

US tech stocks kept falling as investors worried about rising AI buildout costs and the returns on huge spending. Nasdaq opened down 2.5% on Friday. Intel fell as much as 30% after saying cash income won't cover spending this year. Amazon's results missed expectations, with spending up sixfold in three months. Alphabet lost over $500bn in market value Thursday as its AI summaries hit core search ad revenue, and Reddit warned Google's AI search is hurting referral traffic.

Why it matters: Intel and Amazon keep raising AI spending under cash pressure, while Google's AI search is already hitting ad revenue and traffic to outside sites.

OpenAI News

Credit Saison achieves 954% ROI in a ChatGPT Enterprise pilot

Credit Saison 在 315 名员工参与的 ChatGPT Enterprise 试点中实现 954% ROI,人均每月节省 14 小时,远超 500% ROI 和每月 7.5 小时的初始目标。试点结束后仅 1 个月,公司便决定向约 3700 名日本员工全面推广,工程团队使用 Codex 平均每月节省 850 小时。

Hacker News front page

ICE wants its mask back, and hired a company to ask

FlokiNET拒绝了ZeroFox代表ICE提出的Icepick网站下架请求,认为其公开政府雇员工作身份具有公共监督价值。ZeroFox通过AI检测寻找目标材料,DHS于2026年7月与其签订$15 million合同,授权其代表部门提交请求至2028年。