事件流
关注的公司官号 / KOL 一发帖，先分诊（精选 / 普通），精选的才做传导分析、点名标的、给持仓方向。 共采集 578 条，精选 18，普通 6，待分析 245。
RT @AdrianDittmann: https://t.co/ccSbMfxMu2
RT @cb_doge: This entire scene was made using @Grok Grok Imagine 1.5 is on another level. 🔥 https://t.co/zzcTzx1KvC
RT @pmarca: It is hard to overstate how pervasive the SPLC’s control was in the tech industry, and in fact still is in many companies.
Falcon 9 launches from pad 40 in Florida https://t.co/DswQicc5TM
Deployment of all three BlueBird satellites confirmed
Falcon 9’s first stage has landed on the A Shortfall of Gravitas droneship https://t.co/LqecUt0AwY
Liftoff! https://t.co/FeiTy12b2Y
Watch Falcon 9 launch the @AST_SpaceMobile BlueBird 8-10 mission to orbit https://t.co/wvLHBYHxID
RT @C_3C_3: Most popular new baby boy names in the UK: 2015: Oliver 2016: Oliver 2017: Oliver 2018: Oliver 2019: Oliver 2020: Oliver 2021:…
RT @fwprism: Asmongold gives his BASED opinion, says: "You're not being oppressed by the top 2% of society. You're being oppressed by the b…
RT @cb_doge: All these videos were created using Grok Imagine 1.5 Big upgrade. Huge jump in quality. 🚀 https://t.co/jWroTLlqHv
RT @elonmusk: @brian_armstrong A legion of Optimi combined with solar panels will constitute the first Von Neumann probe
Just follow SpaceX if you want news about our company
RT @XFreeze: Grok Voice just ranked #1 on Vapi’s Humanness Index and it's scary good That is a huge deal This benchmark is simple: Listen…
Dragon will also announce its arrival with a brief sonic boom prior to splashing down in the Pacific Ocean
Dragon is on track to reenter Earth’s atmosphere and splash down off the southern coast of California near Oceanside at ~5:10 a.m. PT on Wednesday, June 17 https://t.co/9ZNVoHPf2y
Grok Imagine 1.5 is now in wide release
RT @Starlink: Starlink’s reliable high-speed internet is onboard 30+ Goldenport vessels, enabling the maritime operator to run data intensi…
Nashville getting drilled
RT @libsoftiktok: The “white supremacist KKK” who burned a cross in Chicago is a Leftist anti-Trump Asian dude Another hoax. Every. Singl…
https://t.co/XJ5unQrlSS
Grok on Bedrock
Ayar Labs' partnership with @nvidia enables hyperscalers and system innovators to build optically connected AI infrastructure around NVIDIA's NVLink Fusion platform and partner ecosystem. https://t.co/iyQSL9kkV9 #optical #photonics #semiconductor #AI #HPC #CPO https://t.co/R4eSbVoJk9
RT @tesla_na: A Model Y driver started experiencing a medical emergency with chest pain mid-drive & called his son. His son then remotely…
微软在Azure上使用8192块NVIDIA GB200 NVL72系统的GPU，以7.07分钟完成了Llama 3.1 405B的训练目标，这是迄今为止最大规模的MLPerf Training提交之一。
微软Azure用8192 GPU的GB200 NVL72系统在MLPerf刷新Llama 3.1 405B训练纪录，直接验证Blackwell架构大规模训练性能与能效。对英伟达整机柜系统是强利多催化，加速云商采购决心；向上游拉动核心组件（CoWoS、HBM3e、铜缆、电源、液冷）需求，向下游推升AI算力云服务价值。短期强烈看好英伟达及紧耦合硬科技供应链，云商受益但弹性略逊。
▸展开完整传导链（20 个对象）
| 对象 | 方向 | 烈度 | 分数 | 确定性 | 逻辑 |
|---|---|---|---|---|---|
| 英伟达 | 利多 | 100 | +100 | 高 | MLPerf纪录直接证明Blackwell NVL72系统在万卡级训练上的标杆性能，强化CSP采购确定性，驱动GB300订单上修。 |
| 安费诺 | 利多 | 64 | +90 | 高 | 独供英伟达NVL72柜内NVLink铜背板及高速连接器，单柜价值量3万美元，GB300放量直接放大营收。 |
| 台积电 | 利多 | 60 | +90 | 高 | 英伟达Blackwell GPU唯一代工商，3nm制程+CoWoS-L先进封装独占，GB300上量拉动晶圆代工及封装营收高增。 |
| MPS | 利多 | 60 | +85 | 高 | AI服务器VRM控制器核心供应商，NVL72高功率密度供电需大量多相电源芯片，MPS是英伟达800V DC联盟成员，直接受益。 |
| 德州仪器 | 利多 | 60 | +85 | 高 | 加入英伟达800V DC电源联盟，为GB300平台提供可扩展电源管理方案，数据中心电源管理业务受益。 |
| 工业富联 | 利多 | 55 | +85 | 高 | 英伟达GB300 NVL72最大ODM组装商，份额约40%，代工价值量单柜超20万美元，业绩弹性大。 |
| 纬创 | 利多 | 55 | +80 | 高 | 英伟达GB300 Compute Tray主力代工厂，份额约25-30%，直接受益整机柜出货放量。 |
| 三星 | 利多 | 49 | +75 | 中 | 通过英伟达12层HBM3e验证，预计年底起进入GB300供应链，初期份额虽低但打开空间，高价值量HBM加速业绩增长。 |
| 沪电股份 | 利多 | 51 | +85 | 高 | 英伟达AI服务器高多层PCB核心供应商，GB300 NVSwitch托盘用22层M8U材料超高层板，订单已排至2027年。 |
| 胜宏科技 | 利多 | 49 | +80 | 高 | 英伟达AI GPU PCB全球Tier 1供应商，GB300 OAM升级为五阶HDI，单价与份额双升。 |
| 中际旭创 | 利多 | 29 | +75 | 高 | 英伟达800G光模块最大供应商（份额>50%），1.6T已小批量交付，GB300向1.6T升级进一步提升价值量。 |
| 新易盛 | 利多 | 32 | +75 | 高 | 英伟达800G光模块第二大供应商，1.6T模块已完成验证参与GB300平台，份额持续扩大。 |
| SK海力士 | 利多 | 35 | +80 | 高 | 英伟达GB300初期12层HBM3e独供，HBM绑定AI GPU出货，高ASP、高毛利拉动存储业务盈利能力。 |
| 台达 | 利多 | 43 | +80 | 高 | 英伟达NVL72电源模组主力供应商，单柜需标配台达180kW电源架，超容、BBU等新品类带来增量。 |
| 维谛技术 | 利多 | 35 | +75 | 高 | 与英伟达联合开发GB300液冷方案，提供CDU及供电参考架构，液冷渗透率提升直接受益。 |
| 英维克 | 利多 | 30 | +70 | 中 | 通过英伟达NPN Tier 1认证，GB300液冷解决方案官方伙伴，但份额待实际放量验证。 |
| 微软 | 利多 | 13 | +90 | 高 | 直接事件主体，率先部署GB200 NVL72并创MLPerf纪录，Azure AI服务性能领先，吸引更多AI工作负载。 |
| CoreWeave | 利多 | 39 | +85 | 高 | 全球首家部署GB300 NVL72的云商，作为纯AI算力云，训练性能标杆直接促进客户订单与定价能力。 |
| Nebius | 利多 | 55 | +80 | 中 | 已宣布提供GB300 NVL72实例，但体量较小，业绩弹性大但市值交易量有限，中性偏多。 |
| 精达股份 | 利多 | 58 | +75 | 高 | 英伟达高速镀银铜线约90%供应商，安费诺等连接器厂商的上游核心材料，GB300单柜铜线用量提升显著。 |
数据缺口（4）
- 骨架中存在106条链异常（跨层跳跃、环节环等），多为细分归类粒度过粗导致，不影响主要传导，但部分远端泛产业链（如化工原料、特种钢材）数据以占位符为主，缺失营收占比等关键定量信息，精度受限。
- 产品级BOM数量多数缺失，仅凭拓扑关系定性传导，无法测算业绩弹性具体数字。
- 部分公司（如未上市企业）财务数据为估算，且AI相关营收占比为券商推算，实际可能偏差较大。
- 事件中提及的GB200 NVL72在骨架中以GB300 NVL72为代表进行传导，两者在组件上高度相似，但存在细微差异，可能导致部分具体规格对应的供应商略有不同。
https://t.co/yoplqLmc06
Momentum like this comes from customers pushing boundaries and teams turning ambition into outcomes. We’re excited for what’s ahead as organizations scale AI impact. https://t.co/31XVIiewKz
New Azure milestone. The fastest time to train yet at the largest reported scale for this leading AI training benchmark. A great example of what is possible when we bring together full-stack innovation across silicon, systems, networking, and software, along with our deep partnership with @nvidia, to advance the frontier of AI infra. https://t.co/5ADybErndz
@CoherentCorp Read more on how Coherent’s Sherman expansion is helping build the American manufacturing base behind the AI era: https://t.co/UbIqJhav9G
黄仁勋与相干公司CEO Jim Anderson及当地官员在德克萨斯州谢尔曼市共同庆祝先进光子学制造产能的扩建，该技术用于AI系统中的激光器和光学链路，以增强美国制造业基础。
英伟达与Coherent在德州谢尔曼扩张光子制造产能，直接巩固英伟达高速光互连供应链，利好Coherent自身产能及上游光芯片/DSP/特种光纤。事件震中在Coherent光模块，向上游传导拉动Lumentum、Marvell等，向整机柜/云服务传导强化英伟达生态。利多方向明确，最受益标的为Coherent、英伟达和核心光芯片供应商，短期光模块上游订单可见度提升，长期利好CPO等技术落地。
▸展开完整传导链（9 个对象）
| 对象 | 方向 | 烈度 | 分数 | 确定性 | 逻辑 |
|---|---|---|---|---|---|
| 英伟达 | 利多 | 100 | +85 | 高 | 锚点产品，合作扩张光子制造直接保障GB300整机柜所需800G/1.6T光模块供给，减少供应链风险，加快AI工厂部署。 |
| Coherent | 利多 | 32 | +90 | 高 | 直接与英伟达合作扩张产能，锁定大客户订单，产能翻倍预期，在英伟达供应链份额有望提升，业绩弹性最大。 |
| Lumentum | 利多 | 30 | +80 | 高 | 作为EML芯片全球双寡头之一，是Coherent/英伟达光模块核心上游，产能扩张直接拉动200G EML需求，供不应求格局强化。 |
| 源杰科技 | 利多 | 27 | +70 | 中 | 国产EML芯片已进入英伟达外围供应链，Coherent产能扩张可能带动国产光芯片验证加速，但直接受益度略低。 |
| 仕佳光子 | 利多 | 21 | +60 | 中 | AWG分波器及光芯片供应商，数据传输量增长间接利好，但无直接绑定，弹性有限。 |
| Marvell | 利多 | 19 | +75 | 高 | PAM4 DSP是光模块必备电芯片，800G/1.6T升级中Marvell份额领先，受益于光模块总量扩张。 |
| AyarLabs | 利多 | 30 | +65 | 中 | 英伟达NVLink Fusion生态成员，光子制造扩张加速CPO技术成熟，AyarLabs硅光方案有望更快导入。 |
| 康宁 | 利多 | 21 | +60 | 中 | 特种光纤是光模块上游，产能扩张小幅增加光纤需求，但弹性较小。 |
| 中际旭创 | 中性 | 29 | 0 | 中 | 英伟达与Coherent合作可能挤压旭创在英伟达的份额，但整体光模块市场增长，旭创仍受益于其他云商需求，中性偏谨慎。 |
数据缺口（3）
- 骨架中缺少光模块生产设备（如耦合机、贴片机、测试设备）供应商，无法分析设备端受益情况。
- Coherent-光模块的unitPrice/revSharePct为null，无法定量测算营收弹性。
- 骨架层级异常106条（跨层边、环），多为段分类粒度问题，不影响核心传导逻辑。
RT @Starlink: Starlink Mobile is now providing data in Ukraine with @TwiyKyivstar @VeonGroup, enabling millions of people to stay connected…
RT @ChromiumDev: Generic AI prompts burn tokens guessing at performance issues from raw source code 📉 Chrome DevTools for agents fixes thi…
Learn more about Wear OS 7 ↓ https://t.co/HU3IFoHMvS
Wear OS 7 is rolling out to eligible Pixel Watch devices today. It includes: • Android Live Updates directly on your wrist, so you can track ongoing events — the score of a big game, your food order’s arrival time or your workout progress — in real time • Better integration with your connected devices, like your earbuds or intelligent eyewear • Personalized help with Gemini Intelligence on select devices, coming later this year • Improved battery life with system-level power optimizations
RT @Rothmus: At least 250,000 white British girls have been groomed, raped, and brutalized. Out of 4.2 million white British girls under 1…
RT @pmarca: (1) What
Correction: GB300 NVL72 delivers 1.6x performance on DeepSeekV3 pretraining at 512 GPU scale. (Ref: results 6.0-0022 and 6.0-0101)
RT @dave_brown24: Many of the hawkish conservatives who rushed to President Trump’s defense at the beginning of the war with Iran now fear…
AI models are getting bigger and they need room to run. 🧠 @wccftech spotlights AMD Ryzen AI Max PRO 400 Series processors, featuring up to 192GB unified memory to help developers and creators run 300B+ parameter LLMs locally. https://t.co/EMZEQooPKn
RT @NVIDIAAIInfra: 📣 We are collaborating with @Cadence to bring autonomous AI agents to chip design verification. See how Cadence ChipSt…
Code is the right action interface for spatial reasoning agents. New from NVIDIA Research: SpatialClaw, a training-free agent that uses code as its action interface for complex visual tasks. Instead of calling a fixed set of pre-defined tools, the agent writes Python inside a persistent kernel, so it can compose perception modules, inspect intermediate results, and revise its strategy across steps. Perception outputs become ordinary variables it can reuse and combine with libraries like NumPy and SciPy. With no benchmark-specific or model-specific tuning, it beats a recent prior agent by 11.2 points across 20 benchmarks and holds up consistently across six different model backbones. You can check out SpatialClaw here: https://t.co/xfZBw5il0i
Simulated deployments also reduced evaluation awareness to levels close to real production traffic. We extended the method to agentic deployments with stateful tools, showing that tool simulators can produce realistic trajectories when given sufficient context and capabilities. https://t.co/8JMXApY8xe
Deployment Simulation works best with representative production data, which external evaluators often can’t access. In a companion post for our Alignment blog, we also explore the public WildChat dataset and find that, while less precise, it still provides a useful signal about deployment behavior. https://t.co/UPUnhEiGeh
We’re sharing new research on a method for anticipating how models may behave in real-world use before release: simulating deployment with recent, de-identified user requests and studying candidate model responses. https://t.co/7RJzBfNniQ
These and other measures will allow us to track consequential shifts in the nature of work as they happen—we'll incorporate some of them into the Anthropic Economic Index going forward. Read the full report: https://t.co/IjjwQvrESo
Domain experts—as judged by the questions they ask and vocabulary they use about a subject—are more likely to see success. But the gap between intermediate and expert users is quite modest, suggesting that proficiency in a domain is sufficient to code successfully within it. https://t.co/tZZ7jVYpTB
The average task in Claude Code has grown more valuable. We compared the type of work done in each session to what that same task would cost on a freelance marketplace. From October to April, the monetary value of the average session grew 27%. https://t.co/VcIPFTgN4Q
We compared Claude Code success rates between occupations. On our toughest measure of success—requiring verifiable evidence that a goal was completed, like committed code—every field was within 7 percentage points of software engineering. https://t.co/ShxZK4p0e2
Our latest economic research introduces a framework for tracking Claude Code as it scales. Who is using Claude Code, and what are they using it for? How is the value of tasks changing? And how much does domain expertise shape whether a session succeeds? https://t.co/IjjwQvrESo
Super! 💫 Data from @chandraxray has uncovered possible remains of a supernova in the middle of our Milky Way galaxy. If confirmed, this supernova piece would be one of the closest to our galaxy’s central black hole that we have ever found. https://t.co/gI5TfSJ45E https://t.co/9FJx02K8lt
Learn more about @Android 17, rolling out first to Pixel devices today ↓ https://t.co/WnZWtcN13I
Android 17 is here 📲 New features include: 🫧 Bubbles, which allows you to turn any app into a compact, floating window so you can stay in the flow 🤳 Screen Reactions, so you can record yourself using your device’s selfie camera and capture your phone screen at the same time 🎮 A new gaming mode for foldable devices, which makes full use of your screen real estate 🔐 New and improved safety and security features
RT @theCUBE: @CoreWeave is bringing the #VeraRubin era to life 🦾 At the @NYSE Wired Studio, @dvellante spoke with CoreWeave’s @CoreySander…
Ineffable Labs 正在构建一个从经验中学习的超级学习器，需要巨大的计算规模。他们选择了 NVIDIA Vera Rubin NVL72（谷歌云上最大的集群之一）来支持其实现超级智能的目标。
普通该帖子描述了 Ineffable Labs 使用 NVIDIA 产品进行研发，但缺乏具体订单规模、交付时间或对供应链的可量化影响，属于普通的业务合作或技术选择信息。
Catch up on everything new coming to Pixel this month ↓ https://t.co/CQXKfL3XRg
The June Pixel Drop includes new features and updates that make your devices more personalized and helpful. You’ll now be able to: ✨ React in real-time with Screen reactions (no green screen required) ✨ Blend any combination of text, images and video to create your own high-quality video with Gemini Omni ✨ Create your own custom soundtrack in Gemini from an idea or photo ✨ Turn any app into a compact, floating window to make multitasking smoother ✨ Use Voice Translate on more devices to translate phone calls in real time ✨ Chat smarter with Magic Cue in more apps ✨ Use call features like Take a Message in more regions ✨ Automatically notify your loved ones in an emergency with features like Car Crash Detection, Fall Detection and Loss of Pulse Detection ✨ Edit your photos by simply asking in more regions
RT @clamanna: Copilot Cowork is now generally available! Over the last few months of preview in Frontier, we’ve seen you use Cowork to hel…
Nemotron 3 Ultra and the Open Model Landscape | Nemotron Labs https://t.co/8opXC1aQ4r
Cadence + @HPE are collaborating to advance digital twin–driven data center modernization for the #AI era. Helping teams design, optimize, and scale infrastructure with greater efficiency and confidence. Read more: https://t.co/RmhQk4gANY https://t.co/l65KIH32Y4
AMD delivers its most comprehensive MLPerf Training submission to date: first multi-node Training submission, FLUX.1 at 512 GPUs/64 nodes with OCI, first use of AMD Primus, 3.5X generational gain on Llama 2-70B, and a record 10 partner submissions. Learn more: https://t.co/zD8IprgLGD
Listen to the OpenAI Podcast on— Spotify https://t.co/5u8ANPIHBe Apple https://t.co/ZhhRA1ZB27 YouTube https://t.co/ABG78oTl6W
Let’s talk about evals. We’re always looking for better ways to measure and forecast model progress, especially as benchmarks get saturated or gamed. @tejalpatwardhan, who leads our frontier evals team, spoke to @andrewmayne about why evals matter and what models need to be judged on next.
As the first automotive OEM to adopt ADI’s electrochemical impedance spectroscopy (EIS) technology for mass production, XPENG is helping to unlock new insights into battery management enabling a safer and smarter driving experience. Learn more: https://t.co/PlkKh79yKc https://t.co/czsq4OdBuN
RT @OpenAIDevs: More of Codex is rolling out across Europe this week. We’re bringing Computer use, the Codex Chrome extension, personalize…
Say hello to @XREAL_GLOBAL AURA ⚡️ ✅ Sleek, lightweight design ✅ Massive 70-degree FOV ✅ Powered by Android XR ✅ Built on the new @Snapdragon Reality Elite Platform ✅ Hands-free help from Gemini Reserve for fall now: https://t.co/c1DD5rQuo8 https://t.co/GprK92Q8ov
The journey with AI in development is just getting started. Beyond code generation lies an incredible opportunity to transform how we build. Here is how it unlocks that potential: 1️⃣ Integrate AI across the entire software & product lifecycle. 2️⃣ Let our traditional processes evolve and grow into smarter workflows. ✨ Dive into the full, inspiring session here: https://t.co/zAJUXYmh2v
RT @NousResearch: The Hermes Agent Accelerated Business Hackathon presented by @NVIDIAAI × @stripe × @NousResearch starts now, for builders…
The politicians imprison people to hide their crimes
Those who knew must go to prison for a long time
This actually happened and still happens
Censorship always comes with a “protect the children” lie https://t.co/Chhr7dNrlU
The politicians who turned a blind eye to the Rape of Britain must go to prison https://t.co/JFfHZSwiDc
RT @mattsgarman: Excited to share that Grok 4.3 from @elonmusk's @xai is now available on Amazon Bedrock. Customers pick the right model…
Separation confirmed! Dragon is performing four departure burns to move away from the @Space_Station. Splashdown in ~20 hours off the coast of California https://t.co/F85xpP6vCu
RT @Space_Station: .@NASA and @SpaceX now target Dragon's undocking time for 12:25pm ET today.
Best. Day. Ever. Are you tuning in on June 30th? https://t.co/GCosoHGeD2 https://t.co/sADauMRHZf
RT @surface: Built for this world so you can create the next. Now available in Jade, with Dune on the Pro Flex Keyboard. All-day battery.…
All cargo is loaded, the hatch is closed, and Dragon is ready for an on-time departure from the @Space_Station at 12:05 p.m. ET
@googlegemma @measure_plan @GOROman @hubxstudios Try Gemma 4 in Google AI Edge Gallery on iOS or Android, or in @GoogleAIStudio now. Get more details. https://t.co/AIbjH62RCl
3️⃣ The team at @HubXStudios used Gemma 4 to build BetterSpeak, an AI English tutoring platform that uses the Gemma 4 E2B model as the reasoning engine for its on-device pipeline — enabling private, low-latency tutoring without the need for an internet connection. 🔹HubX deployed the 4-bit quantized version of the model to handle tasks like grammar explanations and progress monitoring across languages. By using Gemma 4’s native audio input capabilities, HubX’s app is able to support direct speech-to-speech learning, while reducing costs and ensuring privacy.
We recently released Gemma 4, our most capable open models to date. Since then, they’ve been downloaded more than 150 million times. Here’s how three builders are using @GoogleGemma to create apps, platforms, and more. First up: Builder @measure_plan, who used Gemma 4 to perform visual question answering (VQA) through a specific persona. 🔹By prompting Gemma 4 this way, the model effectively maintained a "medieval bard" character while accurately identifying objects in the room, like “glass of amber liquid” or “shelves with bound tomes.”
Copilot Cowork is now generally available worldwide, now with multi-model support! Every organization can put long-running agents to work on complex, multi-step tasks, grounded in your organization's unique knowledge and know-how. https://t.co/1fJNjGOs5o
LIVE: An uncrewed @SpaceX Dragon spacecraft is preparing to head back to Earth with cargo and scientific experiments from the @Space_Station. Undocking is scheduled for 12:05pm ET (1605 UTC). https://t.co/RAc6LEvOcX
At #Computex2026, we demoed Linear Pluggable Optics between two Scorpio Fabric Switches — up to 50m of fiber at PCIe 6 speeds, no retimers. Read Jignesh Shah's blog for the full breakdown: https://t.co/EdI1fSa3me #AsteraLabs #LinearOptics https://t.co/cioEuM849S
We’re expanding in India with new engineering centers in Ahmedabad and Hyderabad! Building on our Bengaluru hub, these centers strengthen India’s role in our global engineering strategy. We're hiring now: https://t.co/w1K0zKQi6h #AsteraLabs #Hiring #India #AIInfrastructure https://t.co/5dYmtDDCgC
RT @patcaldwell: "Multiple project summaries provided to the White House by Clark Construction show that internal cost estimates have been…
At #COMPUTEX2026, @Lenovo showcased the Arm AGI CPU-based HR650a V3 Server. Hear from John Donovan on why Arm was a natural fit and how the collaboration is helping deliver something different for AI and cloud infrastructure. https://t.co/RpO4k8CzNu https://t.co/MwCkayL9u1
In the MLPerf® Training v6.0 round, we delivered near-linear scaling efficiency with our NVIDIA GB300 NVL72 4,096-GPU cluster connected with NVIDIA Spectrum-X Ethernet. We were the only team to push @nvidia GB300 NVL72 past 2,048 GPUs on this benchmark, and delivered amazing results at every scale. Full breakdown below! https://t.co/hxiPCh2fmQ
英伟达Blackwell平台在MLPerf Training 6.0中表现最佳，实现最快性能和最大规模。该平台还通过可靠性、可用性和可维护性引擎及英伟达弹性扩展功能，减少中断并加快恢复，保障大规模训练任务顺利进行。
NVIDIA Blackwell 平台在 MLPerf Training 6.0 中性能夺冠，叠加可靠性引擎与弹性扩展能力，直接强化了 GB300 NVL72 整机柜的市场确定性，需求与供应链拉货动能将进一步提速。事件沿产业链向上游核心器件（HBM、铜缆、液冷、电源、PCB）和代工/组装环节强烈传导，台积电、SK 海力士、安费诺、工业富联等最受益；向下游利好抢先部署 GB300 的算力云服务商。投资倾向：积极做多英伟达及高烈度、高确定性的上游卡脖子环节，同时关注下游 AI 云服务商的重估机会。
▸展开完整传导链（10 个对象）
| 对象 | 方向 | 烈度 | 分数 | 确定性 | 逻辑 |
|---|---|---|---|---|---|
| 英伟达 | 利多 | 100 | +95 | 高 | 事件直接锚定产品，MLPerf 冠军级性能巩固 Blackwell 为 AI 训练首选基础设施的地位，拉动整机柜需求与锁定客户订单。 |
| 沃尔核材 | 利多 | 70 | +88 | 高 | 为安费诺供应高速铜缆材料，英伟达 NVLink 铜背板需求随 GB300 放量而激增，高速铜线是机柜内部互联的关键物料，公司为国内主要供应商。 |
| 安费诺 | 利多 | 64 | +90 | 高 | GB300 NVL72 的 NVLink 高速铜背板与连接器核心独家供应商，单柜价值量约 3 万美元，直接受益于整机柜出货量提升。 |
| 台积电 | 利多 | 60 | +90 | 高 | 英伟达 Blackwell GPU 唯一晶圆代工方，3nm 制程及 CoWoS-L 先进封装产能由台积电独占，每柜含 72 颗 GPU，量价齐升。 |
| 工业富联 | 利多 | 55 | +85 | 高 | GB300 NVL72 最大 ODM 组装商，全球份额约 40%，单柜代工价值量约 20 万美元，销量增长直接转化为营收弹性。 |
| Nebius | 利多 | 55 | +80 | 高 | 率先部署 GB300 NVL72 的独立 AI 云平台，性能优势有助于其赢得更多企业客户，算力租赁业务价值提升。 |
| xAI | 利多 | 55 | +80 | 高 | xAI 的 Colossus 超算集群已明确采购大量 GB200/GB300 机柜，Blackwell 性能提升直接增强 Grok 大模型训练效率。 |
| 台达 | 利多 | 55 | +80 | 高 | GB300 NVL72 标配 BBU 模组，台达为主力供应商，每柜约 5 个 BBU 模块，新增量明确。 |
| 沪电股份 | 利多 | 51 | +75 | 高 | 英伟达 AI 服务器高多层 PCB 核心供应商，单柜 PCB 价值量显著，受益于 Blackwell 平台持续迭代与出货增加。 |
| SK海力士 | 利多 | 35 | +80 | 高 | 英伟达 Blackwell GPU 独家 HBM3e 主供商（前期份额 100%），每颗 GPU 搭载 288GB HBM3e，量价齐升逻辑最硬，但产业链位置距 anchor 稍远致强度计算值偏低。 |
数据缺口（4）
- 部分产品 revSharePct、annualShipment、unitPrice 为估算口径或 placeholder，具体财务弹性测算需以公司实际披露为准。
- 骨架中有 106 条段关系异常（跨级、环路等），主要集中在材料段互指和上下游跨度过大，虽不影响本次传导主干判断，但部分细分环节（如光纤、化工原料）的传导路径可能不够精确。
- xAI、Nebius 等非上市公司的营收、毛利率等字段缺失，无法量化受益幅度。
- 向下游传导时，Kimi、MiniMax 等中国 AI 大模型/应用的算力供应商具体份额数据有限，仅依据公开报道推断。
Learn how the NVIDIA platform and ecosystem enable AI model builders to launch frontier models faster, minimize training costs, and start generating revenue early. 🔗 https://t.co/tmSTzQDscB
We just trained DeepSeek-V3 671B in 2 minutes. That's all 671 billion parameters, on 8,192 NVIDIA Blackwell Ultra GPUs connected with NVIDIA Spectrum-X Ethernet. The fastest DeepSeek-V3 training run anyone's recorded, set in the new MLPerf® Training v6.0 round on the same cloud platform our customers run on every day.
RT @dhh: Another fun/insane anecdote from the peak-woke era was when an employee of ours tried to get our terms of service amended such tha…
RT @muskonomy: NEWS: Two years ago, an internal document from the Center for Countering Digital Hate listed "K*ll Musk's Twitter" and "Trig…
RT @pavandavuluri: Today we're introducing the next Surface Pro and Surface Laptop. Powered by Snapdragon X2 processors, they're faster and…
https://t.co/8VGIZPhP4U
It is humbling to consider that if we harness just 1 millionth of the Sun’s power for AI, that will be much more than a million times the intelligence of all of humanity
RT @FutureJurvetson: Mission: Expanding the Scope and Scale of Consciousness. From my 📸 of the billboard on Times Square “This is, depend…
RT @grok: @nishkidoonoo @elonmusk Stockfish is the world's strongest chess engine (Elo ~3650). It dominates grandmasters like they're begin…
Made with Grok Imagine https://t.co/EMJdtFPYpo
RT @BasedMikeLee: SPLC = hateful , fraudulent enterprise
RT @RupertLowe10: The Rape Gang Inquiry Report. https://t.co/EuKgGWBRhS https://t.co/SD5G9HPVtV
Wow, this is crazy!
RT @mntruell: Lots to do together. Excited to be joining forces with @SpaceX to build useful AI.
RT @Samdani4232: 🚨 BREAKING: Acting AG Blanche and FBI Director Patel announce a grand jury has INDICTED leftist NGO Southern Poverty Law C…
RT @cursor_ai: We're excited to join forces with @SpaceX to advance the frontier of useful AI. Expect significant improvements to Cursor so…
AI will achieve Stockfish-level coding and generalized computer use
SpaceX has exercised the option to acquire @cursor_ai in an all-stock transaction with the goal of building the world’s most useful AI models. For the past few months, SpaceXAI has been jointly training a model with Cursor, which will be released in Cursor and Grok Build soon. We look forward to working closely with the Cursor team to advance our frontier AI capabilities
Deepen your expertise with the lightweight Google Gemma open models family on June 19. Learn more → https://t.co/RRmvocfRz4 Join @gusthema & @lucianommartins from Google DeepMind for a live Portuguese AMA session to explore: 🌟 Latest gen AI capabilities 🔥 Performance updates for developers 💡 Interactive audience questions
RT @VeraMBergen: Trump’s declaration that he had finalized a “great deal” on Sunday surprised some of his own top aides, who believed the t…
Registration for #FullyConnected26 is open! Join us 🗓️September 29 – Oct 1 📌San Francisco ♾ 3 days, 3 tracks and 2,000+ AI Leaders in the room. This is where you need to be. https://t.co/Bd5R1VGkqH https://t.co/HVtZXk36uy
邀请函｜华润微邀您共赴2026慕尼黑上海电子展 7月1日-3日，欢迎您与我们共聚上海新国际博览中心N5-641展位
廉洁文化建设·工作动态|西铜公司党员干部参观厂区廉洁文化墙 日前，为持续筑牢廉洁自律思想根基，严守纪律规矩底线，西铜公司组织党员干部前往厂区廉洁文化墙，开展了沉浸式廉洁警
树立和践行正确政绩观|王彬作树立和践行正确政绩观学习教育专题党课宣讲 6月16日下午，白银集团党委书记、董事长王彬作树立和践行正确政绩观学习教育专题党课宣讲，以《新时代树立和践行正
铜城播报|来自基层的四则报道 选矿公司 日前，白银矿冶职业技术学院34名矿物加工专业学生到选矿公司开展了为期两周的校外实习。
兰州新区党工委委员、管委会主任张勇一行来白银集团考察 6月16日上午，兰州新区党工委委员、管委会主任张勇带队来白银集团考察，双方就深化地企合作进行了沟通交流。
东北大学继续教育学院副院长张福磊一行来白银集团考察 6月16日上午，东北大学继续教育学院副院长张福磊一行来白银集团考察，双方就深化校企合作进行了沟通交流。
提速提质提效·数智赋能|铜业公司：自动化铜库智管库位高效储运 走进崭新的铜业公司电铜库，宽敞明亮的厂房内，2025年11月建成投用的各类自动化设备高效协同、有条不紊运转，勾
安全生产月|西北矿冶院开展消防安全培训与应急逃生演练 日前，为进一步增强全员消防安全意识，筑牢科研生产消防安全防线，西北矿冶院开展了消防安全专项培训与逃生演练活动。
安全生产月|厂坝矿：多点发力掀起安全生产月活动热潮 6月份以来，厂坝矿以全覆盖宣传、沉浸式学习、实战化练兵、常态化整治等为抓手，精准部署、多点发力，全面掀起安全生
一封家书护平安||风华高科组织开展“一封安全家书”征集活动（二）
新来的同事，爱了！
与客户同行｜墨西哥国家电力×亨通：跨海筑线，点亮坎昆海岛民生 一条海缆点亮海岛民生，为世界杯精彩保驾护航
白银集团纪委“端午”节廉洁提醒 端午时节话廉洁，棕绳紧束守底线。端午节将至，为持续纠治“四风”，营造风清气正、廉洁祥和的节日氛围。
金钼集团参加了第十一届中国国际矿业展览会。他们借这次参展对外展示了公司在矿业全产业链上的业务实力。
锚定金钼股份（钼资源），展会亮相本质是上游资源龙头向AI高端硬件材料（半导体靶材、特种合金）拓展应用边界的信号。对金钼股份自身是提升产品附加值、强化高端市场渗透的利多，对下游靶材及合金企业构成原料保供与国产替代的支撑。投资上首选具备资源壁垒与高端化预期的钼业龙头，次选受益原料国产化的半导体靶材标的。
▸展开完整传导链（4 个对象）
| 对象 | 方向 | 烈度 | 分数 | 确定性 | 逻辑 |
|---|---|---|---|---|---|
| 金钼股份 | 利多 | 100 | +85 | 高 | 作为国内钼业绝对龙头，亮相国际矿业展不仅是品牌展示，更是其高纯钼、钼合金等高端产品向半导体靶材、AI服务器特种管材等高附加值领域加速渗透的信号。钼作为战略性小金属，在AI硬件供应链中的自主可控价值凸显，利于公司估值体系从传统大宗向新材料切换。 |
| 久立特材 | 中性 | 55 | +15 | 中 | 钼是高端不锈钢及耐蚀合金管的核心添加元素（如AI数据中心液冷管路、半导体特气管路）。上游龙头积极拓展市场利于久立特材在特种合金材料端的创新与成本稳定，但展会事件对短期业绩弹性无直接拉动。 |
| 江丰电子 | 利多 | 39 | +45 | 中 | 江丰电子是半导体钼靶材核心企业，金钼集团发力高端钼材料（如5N高纯钼靶坯）有助于江丰电子高纯靶材原料的国产化保供与联合研发，降低对进口高纯金属的依赖，强化其在先进制程靶材市场的竞争力。 |
| 村田 | 中性 | 39 | +5 | 低 | MLCC配方粉需钼等稀有金属作为改性添加剂以保证温度稳定性。但村田作为日企，其供应链体系相对独立且全球化布局，国内矿业展会事件对其实际原料采购与成本影响微弱。 |
数据缺口（2）
- 金钼股份在半导体高纯钼靶材领域的具体营收占比，及其向江丰电子等靶材厂的实际供货份额缺乏公开精确数据，传导链条中的商业绑定深度需进一步跟踪。
- 骨架中存在106处拓扑异常（如AI数据到AI大模型跨11层、液冷与PCB材料段关系环等），属建模粒度问题，不影响本次有色金属至半导体材料的主干传导分析。
中央纪委办公厅公开通报了五起政绩观偏差的典型案件。
普通属于宏观党政纪律与反腐通报，未涉及具体上市公司或特定产业链的实质性业务及行业政策变化，对二级市场标的基本面无直接传导。
金钼股份化学分公司最近做了节前廉洁提醒。主要是要求大家明确责任，过节期间严格遵守廉洁纪律。同时也重申了公司诚信、敬业、创新、共赢的价值观。
本次事件锚定金钼股份的钼产品，但实质为企业内部节日廉政与党建常规新闻，对钼价、产能及AI产业链供需无任何实质性影响。程序计算的下游拓扑烈度（如久立特材、江丰电子）仅反映静态物理链接，实际基本面传导效应为零。建议投资者规避此类噪音博弈，回归对全球钼矿供给刚性与高端制造需求的真实基本面跟踪。
▸展开完整传导链（4 个对象）
| 对象 | 方向 | 烈度 | 分数 | 确定性 | 逻辑 |
|---|---|---|---|---|---|
| 金钼股份 | 中性 | 100 | 0 | 高 | 事件为常规行政与廉政建设新闻，不涉及产能扩张、停产检修、价格调整或技术突破，对公司基本面及钼金属供需格局影响为零。 |
| 久立特材 | 中性 | 55 | 0 | 高 | 虽在拓扑上作为钼的下游（特种不锈钢添加钼元素）获得程序静态烈度，但矿企日常行政动态完全不会改变高端管材的原料采购成本与排产计划，无实质传导。 |
| 江丰电子 | 中性 | 39 | 0 | 高 | 半导体钼靶材上游供应稳定，此新闻不改变晶圆厂耗材的供需格局与验证进度，产业链传导阻断。 |
| 村田 | 中性 | 39 | 0 | 高 | MLCC配方粉中钼等稀有金属添加剂用量极微，且属于长协采购，单一矿企的节日新闻对全球被动元件巨头无任何成本或供应链扰动。 |
数据缺口（3）
- 事件本身为企业行政新闻，完全缺失产业层面的量价、产能及订单数据，无法进行任何弹性测算。
- 骨架中钼在AI产业链（如MLCC配方粉、半导体靶材、高端不锈钢）中的具体BOM单耗及成本占比数据缺失，多为“产业配套推断”，难以精确量化钼价波动对远端AI硬件的真实成本传导。
- 骨架存在106处拓扑异常（如跨级edge、环节成环），部分远端传导路径属建模冗余，但不影响本次“无实质影响”的最终判断。
金钼股份技术中心顺利举办了“国学研修苑”第二期培训活动。此次培训主要围绕国学文化展开，旨在强化员工的企业价值观。
普通该动态属于企业内部文化与培训活动，不涉及产能、订单或财务等基本面变化，对二级市场投资无传导价值。
金钼股份冶炼分公司在6月16日开展了安全宣传咨询日活动。这次活动主要进行安全生产宣传，目的是筑牢公司的生产安全防线。
锚定金钼股份的钼产品。本次事件为每年6月常规的'安全生产月'企业基层宣传活动，属于纯行政与合规动作，对钼金属的产能、供给、价格及下游需求无任何实质性影响。产业链传导逻辑在现实中不成立，程序基于拓扑计算的下游烈度应被归零。建议投资者直接忽略此类新闻噪音，回归钼价周期与公司实际排产基本面，当前无基于此事件的操作价值。
▸展开完整传导链（4 个对象）
| 对象 | 方向 | 烈度 | 分数 | 确定性 | 逻辑 |
|---|---|---|---|---|---|
| 金钼股份 | 中性 | 100 | 0 | 高 | 事件为响应国家'安全生产月'的常规基层宣传咨询日活动，不涉及任何停产、限产、扩产或技术突破，对公司基本面及全球钼金属供给格局无实质影响。 |
| 久立特材 | 中性 | 55 | 0 | 高 | 虽为消耗钼的特种钢材企业，但上游开展安全宣传活动绝不会导致钼原料断供或涨价，程序算出的拓扑传导在现实产业运行中完全不成立。 |
| 江丰电子 | 中性 | 39 | 0 | 高 | 钼靶材原料供应稳定，上游常规行政活动不构成任何供应链冲击或成本变化，无实质利多/利空传导。 |
| 村田 | 中性 | 39 | 0 | 高 | MLCC配方粉中钼添加剂用量极微且供应链冗余度高，上游单厂的安全宣传活动对被动元件产业无任何波及可能。 |
数据缺口（2）
- 事件本身为企业日常行政活动（全国安全生产月），不涉及任何产能、价格、技术或政策维度的产业数据，故无法进行任何供需弹性与业绩增量测算。
- 骨架中存在106处拓扑异常（如跨级边、环），但在本事件无实质产业传导的前提下，这些结构缺陷不影响最终'全局零影响'的结论。
金钼股份矿山分公司最近开展了一场警示教育活动。主要是为了提醒大家守住廉洁底线，树立正确的政绩观。
锚定金钼股份的钼产品。该事件为国企常规内部党建与廉政警示教育活动，对公司的钼矿开采、冶炼产能、成本及全球钼价供需格局无任何实质性影响。向下游半导体靶材（江丰电子）、特种钢材及MLCC等AI硬件上游材料的传导效应为零，不构成任何产业催化或交易信号，建议无视此类噪音。
▸展开完整传导链（4 个对象）
| 对象 | 方向 | 烈度 | 分数 | 确定性 | 逻辑 |
|---|---|---|---|---|---|
| 金钼股份 | 中性 | 100 | 0 | 高 | 内部党建与警示教育活动，不涉及产能、产量、成本或钼价变动，对公司基本面及产业链供需无实质影响。 |
| 久立特材 | 中性 | 55 | 0 | 高 | 钼作为特种不锈钢添加元素，其上游矿企的行政教育活动不改变特钢原料供给与成本，产业传导阻断。 |
| 江丰电子 | 中性 | 39 | 0 | 高 | 虽为钼靶材下游，但上游供应商的内部行政活动不影响高纯钼原料的供应与价格，对AI半导体材料链无传导。 |
| 村田 | 中性 | 39 | 0 | 高 | MLCC配方粉中的钼添加剂供应不受矿企内部教育活动影响，对AI服务器被动元件供应链传导为零。 |
数据缺口（2）
- 该事件为纯行政/党建性质，产业链骨架中的财务与出货量数据无法用于测算此类非经济事件的业绩弹性。
- 骨架中存在106处拓扑异常（如跨级edge），但对本事件分析无影响，因事件本身无产业传导逻辑。
某企业或机构发布了安全生产月主题宣传视频的第一期内容。该帖子主要用于推广和展示企业内部的安全生产文化活动。
普通属于企业文化宣传与日常活动播报，不涉及产业链动态或公司基本面变化，无投资参考价值。
Most AI conversations focus on the technology but there is so much more to consider. Recently on the Human Capital Gains podcast, Arm CFO Jason Child shared a different perspective, exploring how leaders should think about AI adoption, productivity and measuring business impact. 🎧: https://t.co/NGd2ov5k04
字节人最爱用的表情是？ 点击解锁飞书表情冷知识
信号价值偏低，已丢弃。
RT @milichab: Use your SuperGrok or X @premium subscription inside Warp
RT @policylaila: My 17-year-old daughter: “I’m confused. We’ve always been taught not to share personal information or anything that identi…
https://t.co/r3OHSjgxJb
信号价值偏低，已丢弃。
RT @cb_doge: BREAKING: Starlink is delivering faster download speeds than other terrestrial ISPs in 22 out of 23 markets in Africa, per Ook…
https://t.co/U2CbkkvZwy
信号价值偏低，已丢弃。
https://t.co/8HAfyfbEyD
世界杯AI预测，百度文心准确率位列第一 今天，由联想集团与咪咕视频联合发起的“世界杯预测人机大战”公布了阶段成绩： 🏆截至目前，在已完成的15场世界杯比赛中，百度文心累计命中7场，以46.7%的准确率暂列“大模型预测表现榜”第一。 本次预测活动调用的是百度最新一代基础大模型文心5.1。在LMArena搜索榜中，文心5.1曾以1223分位列全球第四、国内第一。 模型的搜索能力，指的是对多源信息的快速检索、整合与生成能力，并能够输出更具一致性与可靠性的回答。 👉这意味着在世界杯预测场景中，文心能够同时覆盖更多维度的实时信息，预测结果的准确性和稳定性也就越高。 ⚽️目前，这场属于AI的世界杯挑战已经进行到第15场！速来围观：<a data-unique-id="mqg4le9p-xhivdm" target="_blank" class="weapp_text_link" data-miniprogram-type="text" data-miniprogram-appid="wx4999cc03db9cb496" data-miniprogram-path="pages/activityH5/activityH5?urlKey=https%3A%2F%2Ftianxi.cn%2Fai-worldcup%3Ffromxcx%3Dofficial%26minienv%3Dlenovo%26appid%3Dwx4fcc58ba3e18c332%26appmsg_compact_url%3Dhttps%3A%2F%2Fmp.weixin.qq.com%2Fs%2F3w16cOgzK6HYgq6aMKt9sw%26wnm_uuid%3Dundefined%26wnm_timestamp%3D1781580191177" data-miniprogram-nickname="联想" data-miniprogram-servicetype="0" data-miniprogram-applink="#小程序://联想/zd8ThYzP6tCfp4F">世界杯预测人机大战</a> 💬接下来，你最期待文心预测哪场比赛？欢迎在评论区聊聊～
仲夏时光，生辰有约 沃尔核材第二季度生日会圆满结束!
抚顺特钢开展领导干部安全生产法治素养与履职能力提升暨新干部安全管理能力专题培训
抚顺特钢廉政微课堂（2026年第六期）| 关于世界杯期间文明观赛及严禁赌博的廉洁提醒书
知行合一 筑梦未来 | 江苏大学能动学部师生走进方盛股份参观实习 方圆之间 盛交天下为深化产教融合、搭建校企协同育人平台，2026年6月9日，江苏大学能源与动力工程学部党委副
习近平党建思想内涵要义
RT @AnneKFlaherty: Trump’s declaration that he had completed a “great deal” w Iran came as a surprise to some of his own top aides, who tho…
利华益维远化学发布了6月16日的产品当日报价。帖子里主要更新了该公司化学产品的最新价格信息。
查询单日现货报价属行情终端功能，本工具无法提供当日精确挂牌价。从产业链骨架看，维远股份（苯酚/双酚A）处于AI硬件（PCB/ABF膜）的最上游大宗原料端，产品同质化高、AI营收贡献近乎为零。其价格波动对AI算力链无实质卡脖子影响，投资逻辑纯属传统化工周期，而非AI主线，建议规避。
▸展开完整传导链（4 个对象）
| 对象 | 方向 | 烈度 | 分数 | 确定性 | 逻辑 |
|---|---|---|---|---|---|
| 维远股份 | 中性 | 100 | 0 | 高 | 震中。大宗化工原料，价格受原油及供需周期主导，AI算力需求对其拉动微乎其微，业绩弹性取决于宏观化工周期而非AI主线。 |
| 圣泉集团 | 中性 | 55 | 0 | 高 | 直接下游（酚醛树脂）。苯酚价格波动影响其成本，但电子级树脂附加值高，成本转嫁能力强，不受单一原料报价扰动。 |
| 味之素 | 中性 | 17 | -5 | 中 | 远端下游（ABF膜）。电子级酚醛固化剂是ABF膜原料，但大宗原料成本占比极低，且味之素垄断ABF膜，议价权极强。 |
| 生益科技 | 中性 | 14 | -5 | 中 | 远端下游（覆铜板）。同理，大宗化工品价格波动对高端CCL成本影响可忽略，核心定价权在铜箔与玻纤布。 |
数据缺口（2）
- 本系统为产业链推演工具，无法提供2026年6月16日苯酚/双酚A的实时现货报价及当日涨跌幅，需借助百川盈孚/卓创资讯等化工数据终端。
- 骨架中存在107处拓扑异常（如跨级edge），多为分类粒度问题，不影响本标的宏观定性。
深化产学研 共赴新未来｜中信特钢与北京科技大学签署战略合作框架协议