人虽不强有弓一长

2K posts

人虽不强有弓一长

人虽不强有弓一长

@guoqiang

产业分析师和技术分析师,璞玉智库和阅码场创始人。正在成为一个准博物学家的路上。

Beijing Katılım Ocak 2009
1K Takip Edilen668 Takipçiler
Sabitlenmiş Tweet
人虽不强有弓一长
人虽不强有弓一长@guoqiang·
范式转移来自两个轴,一个是发展轴,这个是从西方科学革命开始,到工业革命,再到信息革命以来一直的主线,系统的总结来自马克思。另外一个是安全轴,这个来自于自然演化的力量,由天变,地变和人变组成。这两个轴向正在创造一个奇点,将会把人类社会推入一个前所未有的发展轨道,但也催生了灾难轨道。
人虽不强有弓一长 tweet media
中文
1
2
5
0
人虽不强有弓一长
期待
Eden@EdenKollcinaku

Things Google will speak about at Google I/O 2026: - Android (including Android 17 details) - Wear OS - Google AI / Gemini (major model updates, multimodal capabilities, media generation, robotics, intelligent agents, agentic AI) - Agentic coding / vibe-coding tools (new AI-powered development workflows and automation) - Chrome (new web capabilities, browser updates for 2026) - Cloud (integrations and infrastructure for AI/app scaling) - Google Play (latest tools and updates for developers and business growth) - Firebase (evolving into an agent-native platform, AI Studio & Antigravity integrations, vibe-coding full-stack apps) - Gemma open model family (new additions, tools for discovery/deployment across cloud/desktop/mobile) - Android development tools / Android Studio (new features, Gemini-powered app development, workflow acceleration) - Jetpack Compose (UI breakthroughs, productivity improvements, support for new form factors) - Flutter (performance updates, Flutter GenUI for adaptive/AI-generated user experiences) - Web UI / web development (new interactive capabilities, CSS/HTML primitives, scroll-triggered animations, view transitions) - Adaptive development for the expanding Android ecosystem ("Adaptive Everywhere" – seamless experiences across phones, foldables, desktops, cars, TVs, living rooms, and immersive/XR environments) - Aluminium OS (Android for PCs/laptops, desktop mode, large-screen enhancements) - Android XR (XR platform updates, immersive environments, smart glasses/headsets) - Performance, media & camera improvements (in Android 17 and related apps) - Agentic automation for users (empowering faster everyday tasks)

日本語
0
0
0
29
人虽不强有弓一长 retweetledi
Shishir
Shishir@ShishirShelke1·
We now have more details about the alleged OpenAI smartphone 🚨 - Mass production could begin in early 2027 - MediaTek expected as the sole chip supplier - Custom chip based on Dimensity 9600 - Built on TSMC’s advanced N2P node - LPDDR6 + UFS 5.0 - Major focus on camera/visual AI (ISP) with improved HDR - Dual-NPU setup for stronger AI performance - pKVM + inline hashing for enhanced security Estimated shipments: ~30M units (2027–2028) Via: .@mingchikuo (Image for reference)
Shishir tweet media
English
145
90
1.2K
195.8K
人虽不强有弓一长
eswin也受益
Macro_Lin | 市场观察员@LinQingV

SemiAnalysis最近发了一组硅晶圆行业的分析,逻辑很清晰,简单转述一下。 需求端两条线。一条是先进制程逻辑芯片(7nm以下),晶圆需求预计明年开始加速,到CY28接近100万片/月,占300mm等效总需求约10%。用的是外延片(epi wafer),供需平衡收紧速度比市场预期快。另一条是HBM,8到16层DRAM die加一层逻辑base die,每GB的晶圆消耗远高于普通DRAM。HBM用的是抛光片,虽然不直接收紧epi的供需,但能帮整个晶圆市场消化前几年积累的深度过剩。 供给端几乎被锁死。四大晶圆材料商(信越、SUMCO、环球晶圆、Siltronic)的资本开支从CY23高点大幅收缩,CY26E只剩峰值的约30%。原因很简单,没有客户长约承诺就不扩产。关键设备epi反应炉交期超过6个月,就算客户现在签约,新产能最早也要2H27才能上线。 价格拐点正在酝酿。过去几年ASP持续承压,现有LTA锁定了到2027年的大部分定价。但SemiAnalysis认为市场已经在转向,TXN和STM等模拟芯片公司的数据也能印证。他们预计2028年及以后的新长约谈判将在明年初启动,epi wafer新合同ASP可能比现价高出20%左右。整体节奏类似2016到2017年那轮由存储驱动的晶圆上行周期。 核心受益标的就是四大硅晶圆厂商:GlobalWafers、SUMCO、信越化学、Siltronic。

日本語
0
0
1
64
CHUIP LEUNG
CHUIP LEUNG@WayWLeung·
Intel 18A 14A要来了。 当Apple都在台积电产能受困时,AMD更加会无法拿到足够的产能生产足够的供应。这也是我为何看空AMD的原因。 比起产能无法控制的AMD,具有foundry的Intel的竞争力强多了。Intel的重资产模式正在因为地缘政治和逆全球化而变得越来越重要。 控制和具备自身完整的制造体系是非常必要的,这点马斯克做得非常出色。而国家层面上,中国是这方面的霸主。美国在川普主导下也在回归这条必须的路线。
*Walter Bloomberg@DeItaone

$AAPL - APPLE CEO TIM COOK SAYS IPHONE FACED SUPPLY CONSTRAINTS OF ADVANCED NODE CHIPS: REUTERS INTERVIEW

中文
11
9
65
9.1K
人虽不强有弓一长
cpu卖爆,ifs绑定头部,布局量子。典型吃着碗里的,看着锅里的,盯着案板上。
PsiQuantum@PsiQuantum

We are thrilled to announce that @LipBuTan1, Intel CEO and longtime veteran of the semiconductor industry, has joined PsiQuantum's Board of Directors. His appointment reflects PsiQuantum's ongoing work to leverage existing semiconductor manufacturing as the company rapidly scales its silicon photonics platform and drives toward deploying utility-scale quantum computers. Learn more: psiquantum.com/news-import/li…

中文
0
0
0
84
人虽不强有弓一长
如果AI的叙事崩了,支撑美股的下一个叙事是什么?
Aakash Gupta@aakashgupta

A 5-year backlog on grid transformers just killed half of America's 2026 AI data centers. Sightline Climate tracked 12 GW of 2026 US data center capacity announced across 140 projects. Only 5 GW is actually under construction. 11 GW sits in the "announced" stage with no physical progress despite typical build times of 12-18 months. 25% of those projects haven't disclosed a power strategy at all. That last number is the tell. A quarter of "planned 2026 AI capacity" has no sourced answer to where the electrons come from. Call those projects what they are: vapor capex with a press release attached. Nvidia is shipping. The gating constraint is high-voltage transformers, switchgear, and grid-tie batteries. Pre-2020 lead time on a high-power transformer was 24-30 months. Today it stretches to 5 years. Electrical equipment is under 10% of total data center cost and 100% of the bottleneck. This breaks the standard analyst model. When a hyperscaler announces $50B of capex, the Street treats it as compute coming online in 18 months. If the transformer order wasn't placed in 2022, that money sits as commitment without capacity. You cannot pay for a transformer that doesn't exist yet. The winners under this regime are whoever locked in power purchase agreements and electrical equipment orders 3-4 years ago, before anyone was modeling hundreds of megawatts of inference load. Everyone else is waiting in line behind them. Second order is uglier. Hyperscalers buying $50B of GPUs that sit unpowered depreciate against Nvidia's annual cadence while paying carrying costs on empty data center shells. Every quarter dark is a quarter of compounding waste. The "we're 6 months from running out of compute" panic just became "we're 5 years from running out of transformers." Capital fixes one. Capital cannot manufacture a transformer.

中文
0
0
1
110
人虽不强有弓一长
好像中国一重在好多地方都超过JSW了吧。
World Data Analysis@World_Data_A

The steel heart of nuclear power: No Japan no nuclear reactor core Nuclear reactors are often defined by design, capacity, or location. But their most critical component is rarely discussed: the reactor pressure vessel (RPV). At the core of this system are ultra-large steel forgings, massive, single-piece components that must withstand: Extreme heat (up to ~350°C) High pressure (15+ MPa) Decades of radiation exposure If this component fails, the entire plant fails. There is no redundancy. The product These forgings form the “steel heart” of nuclear reactors. They are produced from 600–670 ton steel ingots, shaped under presses up to 14,000 tons, and can reach 10–12 meters in size. Who makes them A very small number of companies globally have this capability. Among them, Japan Steel Works (JSW) stands out. Founded in 1907 Muroran (Hokkaido) is its main facility Decades of experience in nuclear-grade forgings Supplies major reactor projects in Europe, the U.S., China, and Japan Market position JSW has historically claimed around 80% of the global market for large forged components used in nuclear reactors. This does not mean Japan dominates all nuclear construction, but it highlights a critical bottleneck in the supply chain. Where these components are used JSW’s forgings have been used in major reactor designs worldwide: EPR (France) AP1000 (USA) VVER (Finland/China) In many cases, the internal structural core of these reactors originates from Japan. Why this matters This is not a typical manufacturing advantage. Only 5–6 companies globally can produce these components Requires decades of metallurgical expertise Extremely high capital investment Long production cycles (often 2–3 years per unit) This creates a structural chokepoint in the global nuclear supply chain. Regulations and barriers to entry Entry is not just about technology: Must comply with strict nuclear standards (e.g., ASME Section III) Requires approval from multiple regulators (NRC, ASN, etc.) Hundreds of inspections and material tests Once qualified, suppliers are rarely replaced Requalification is so difficult that existing suppliers maintain long-term dominance So these make JSW dominant Source: @WorldNuclear, @AAF, Coral Capital

中文
1
0
1
83
人虽不强有弓一长
原料、燃料、肥料正在影像全世界。对于中国来说,这三样东西都可以从煤化工中获得。
中文
0
0
0
45
人虽不强有弓一长
真是个天才。
Bill The Investor@billtheinvestor

一名19岁的学生通过在课间阅读过期专利,发现了一款年利润达4.2万美元的产品。第一个月他什么也没做。他搭建了一个小型 Mac mini 系统,用于扫描旧的美国专利商标局(USPTO)备案文件,并将480万份过期专利筛选至28万个消费品创意。其中大部分都是垃圾:医疗器械、过时的电子产品、没人会买的工业零件,以及在首次销售前就需要8万美元模具费的产品。接着,一份备案脱颖而出。那是一家多年前倒闭的初创公司设计的折叠式宠物水碗。该专利包含了完整的蓝图:图纸、材料、锁定机制和组装顺序。 他将图纸发给了5家工厂,其中两家回复了,其中一家报价单价0.92美元。而类似的碗在亚马逊上的售价为12-18美元。那些 Mac mini 在后台运行着枯燥的工作:抓取备案文件、清洗文档、为产品创意评分、检查评论投诉,并将旧专利转化为供应商可用的简报。没有办公室,没有团队,没有研究实验室。只有一堆在他在上课时运行的二手 Mac mini。他订购了300个样品,并通过 TikTok Shop 和亚马逊售出了第一批产品。扣除费用和运费后,每件产品的利润为4.70美元。 以每月750件的销量计算,月利润为3,525美元。没有供应商秘籍,没有付费产品调研工具,也没有所谓的“爆款产品”表格。有的只是廉价的硬件,以及一种认知:一些倒闭的公司会留下可以使用的产品。

中文
0
0
0
66
人虽不强有弓一长 retweetledi
Winston
Winston@ChurchillWw·
The US nuclear pipeline is resurging after nearly three decades of near-standstill. The country added ~95 GW between 1969 and 1990, then almost nothing until Watts Bar 2 in 2016 and Vogtle 3-4 in 2023-24. The pipeline announced through 2035 reaches 16 GW: 2.6 GW under construction or active restart (Palisades 1, Duane Arnold, TMI 1, Kemmerer Natrium), while the rest are hyperscaler deals from Meta-TerraPower, Oracle, Fermi America, and Amazon-X-energy, still without NRC permits. EIA tracker: eia.gov/todayinenergy/…
Winston tweet media
English
3
20
87
5.1K
人虽不强有弓一长 retweetledi
Rosy Prosperity
Rosy Prosperity@rosyprosperity·
If those predictions materialise, I have some of mine: 1. HBM demand will probably also accelerate due to the EMIB support for more interfaces than CoWoS. 2. HBM demand will accelerate even further due to the higher number of GPUs being shipped, as currently packaging was a bottleneck.
English
2
2
15
5.5K
人虽不强有弓一长
难道今天涨是因为它?
SemiVision👁️👁️@semivision_tw

According to an exclusive report from Taiwan’s DigiTimes, NVIDIA is reportedly planning to shift part of the foundry orders for its next-generation Feynman GPU to Intel. While the critical GPU core is still expected to be manufactured by TSMC, the I/O chiplet responsible for communication is likely to adopt Intel’s 18A or 14A process. Around 25% of the GPUs may also use Intel’s EMIB advanced packaging technology. According to DigiTimes, following NVIDIA’s investment in Intel, the company has preliminarily decided to adopt a dual-foundry model for its “Feynman” architecture platform, scheduled for launch in 2028. This move is seen as a strategic retreat aimed at aligning with “Made in America” policy pressure amid tariff threats from the Trump administration. Current rumors suggest that the Feynman GPU core will remain with TSMC, while the communication I/O die may shift to Intel’s 18A or 14A process. In addition, 25% of the GPUs are expected to use Intel’s EMIB advanced packaging technology. DigiTimes noted that Intel taking part of NVIDIA’s orders could be “more beneficial than harmful” for TSMC. First, modestly giving up some market share could ease regulatory concerns over TSMC’s perceived monopoly. Second, it could help relieve political pressure from the U.S. government. Finally, because the orders flowing to Intel are likely to be smaller-volume and lower-end in nature, TSMC would still retain control over the most advanced and highest-margin core manufacturing processes. This kind of multi-sourcing model is likely to become standard practice for U.S. chipmakers over the next several years. For Intel, the key question will be whether it can deliver competitive 14A yield on schedule by 2028. $INTC open.substack.com/pub/tspasemico…

中文
0
0
0
42
人虽不强有弓一长 retweetledi
SemiVision👁️👁️
SemiVision👁️👁️@semivision_tw·
#NVIDIA key industry partners for adoption of 800V DC architecture
SemiVision👁️👁️ tweet media
English
0
13
109
8.8K