Latest post stream

@ram_blings Not 1:1 but $LITE gross margins are ~45-48%. Having 60% gross margins for $SIVE is extreme pricing power. And that’s on top of exponentially growing revenue pipelines… at the very beginning of the CPO supercycle.

@Chaimeisenberg 💯

@KPACTAFXAMCTEP Thanks for asking that question. What an absolutely legendary response from $SIVE CEO. And this type of answer around demand just derisks Sivers a ton.

@xu159411209 哪家厉害的还没上市？

Holy crap, this is the most bullish thing I’ve heard from $SIVE so far. From earnings transcripts: “We do not look at competitors when demand far outstrips supply” (literally anything they make gets bought) Along with: “We see 60% gross margins in the future” (incredibly high) “We have two technologies that can feed into these three supercycles that are currently going on.” (Holy revenue opportunities) My high conviction CPO/photonics long is $SIVE for a reason.

@seti_park @Alisvolatprop12 👀

@nic_carter Anyone saying “quantum is next” here is a scammer.

@leemaiking @aleabitoreddit 好奇怪，问下客服？

其实大多都是旧新闻了。但是 $NOW 现在确实蛮划算的。没买很多，但是软件股反弹里，我觉得应该会有他。

https://t.co/1079I1OZBD

"Can't Build Equipment Without the Parts"… Semiconductor Test Equipment Faces "Worst-Ever" Supply Crunch The semiconductor test equipment industry is grappling with a severe component shortage—so severe that some are lamenting, "We can't build semiconductor test equipment because there are no semiconductors." According to industry sources on the 29th, the supply of components needed for semiconductor test equipment has become difficult to secure. In particular, lead times for non-memory semiconductors—such as field-programmable gate arrays (FPGAs), central processing units (CPUs), and driver integrated circuits (ICs)—have lengthened dramatically. First, the lead time for the FPGAs needed to operate the equipment has stretched from a previous 8–10 weeks to as much as 52 weeks recently. A distributor source explained, "It varies by FPGA specification, but it's typically 52 weeks," adding that "supply has become difficult to secure." FPGAs are used to analyze test data in real time, quickly identifying problems such as defects. The FPGA market is led by AMD, which acquired Xilinx. The situation is the same for driver ICs used in test equipment. Whereas distributors could previously purchase the relevant chips from inventory immediately, it now takes at least 10 weeks or more. In particular, Analog Devices' (ADI) product lineup for automatic test equipment (ATE) is causing an extreme bottleneck. ADI supplies "pin drivers"—integrated assemblies of multiple devices—for semiconductor test equipment. x86-architecture-based CPUs and graphics processing units (GPUs) are also experiencing supply shortages. One semiconductor equipment industry official noted, "For some products, supply has tightened and unit prices have risen sharply—up to threefold, from a previous 1 million won to 3 million won," adding that "Intel's server CPUs, in particular, are not easy to secure." Intel has recently been allocating its server CPU (Xeon) volumes primarily to high-margin hyperscalers—large-scale cloud service providers—and data center servers. As a result, supply to other markets has been less smooth. Mass production of "Diamond Rapids," the next-generation server CPU, has also been pushed back from the originally planned second half of this year to the middle of next year. Because of this, the development and supply of next-generation equipment that requires the product's high performance and new features have been partially delayed. In fact, one particular test equipment manufacturer recently signed an equipment supply contract worth tens of billions of won with Samsung Electronics, but due to delayed component procurement, it had no choice but to push back the delivery date by three months. An industry official said, "The current situation isn't simply a problem with specific components like FPGAs or CPUs," explaining that "a serious bottleneck is occurring across the entire non-memory semiconductor supply chain." Test equipment manufacturers are responding by placing advance orders for components. Under this approach, they negotiate equipment volumes and delivery dates with customers several months before a formal purchase order (PO) is issued, and order the necessary components ahead of time. Even so, with component delivery delays worsening, fully smooth supply remains difficult to achieve. The industry expects the supply crunch for non-memory components used in test equipment to persist for some time. This is because the boom in downstream industries—such as demand for AI and data center infrastructure—continues, driving up demand for semiconductors (components) and semiconductor test equipment simultaneously. Semiconductor manufacturers that purchase test equipment are also on high alert. As one industry official put it, "A strategy in which semiconductor manufacturers and equipment makers cooperate closely to respond proactively is increasingly becoming the new normal."

$SIVE basically took their entire revenue pipeline. In the entire company’s history. Then grew that by 77% in the first 3 months. Thats by far the clearest indication of the inflection of the CPO supercycle. It’s probably going to look exponential from here on out. https://t.co/81pYk3Y36V

@gudou_dama MSCI index inclusion inflows happen end of trading day today. Or on June 1st (which was also Nasdaq omx inflow). One of those days

@TJ_Research 这就回到了一个重要的话题，错杀的软件股里这种AI基础设施的一定会更加受益

@daniel_praguepp 单边行情磨损可以忽略

@TJ_Research 用skill / mcp 去访问数据库，而不是让它自己去搜数据。要会用

@DeJeff48985 Extremely bullish on $SIVE. I’m very brutally honest when it comes to earnings reports on the names I own. We got 77% forward growth indications from the entire company’s pipelines… over a single quarter. $JBL ramp confirmation. Then multiple volume ramp production starts from management. Multiple. Should keep exponentially growing over next two years. This was just the inflection point I called out.

@qptmvj 헉

@BigKaiWang Sequential 77% growth. Basically entire company revenue pipeline from all these years. And it increased 77% in the first few months of 2026. Absurd forward growth indications coming from $SIVE.

@dockery58 Thanks, I’m still waiting on the transcript to come out. But glad they clarified $JBL volume ramp is coming and photonics is their main growth vector. Great thesis validation so far.

Authropic最新一轮650亿美融资，最大的看点不是估值9650亿美金超越OpenAI登顶大模型厂商王座，而是美光、三星电子、SK海力士存储御三家一起参与投资，以御三家在产业链上的话语权这比几家大科技投资更具有说服力、这意味着AI大模型与算力芯片产业链进一步深度绑定。 一个月前聊到大模型厂商是各领风骚三五月，其实最近使用下来GPT 5.5 pro的体验是更好的。过段时间大家会不会发现OpenAi又成了那个被低估的了。🤔

@RyanU_1F42B This is 77% growth, from literally the start of the CPO supercycle H1 2026. I’d expect that number to just keep on compounding exponentially as we approach H2 2027, which is the true inflection of scale up CPO.

@cobra_mindset1 77% growth in 1 single quarter. From the entire company’s operating history to date. I’m not sure if markets realize how high that is and represents a structural shift in forward revenue growth indications. Extremely positive in regards to $SIVE thesis validation so far.

@chrisoc9999 77% growth of forward revenue projections… Over a single quarter is absurdly high growth. And validates volume ramp expectations with $SIVE. Just want you want to hear from qualification cycle players, along with “multiple volume ramp” production start. And that should compound a lot further as other volume ramps start

Only thing to look at with $SIVE earnings is forward growth. Nobody cares about pre-development contract earnings from 2025 or last quarter, especially for qualification cycle optical players. Having 77% growth of opportunity pipelines (revenue volume ramp projections), to $799m In a quarter is absolutely incredible growth. And, I’d expect to see that continue compounding. “The company continues to anticipate several volume production starts within AI data centers” (photonics/lasers). Is also very positive and validates the thesis about volume ramp for photonics. Now next thing to look at is earnings call transcript, once they’re indexed, which is the most important signal of what’s to come. Overall, very positive.

@_TAOLIMAO 所见略同😄

@Jeffreyvasily 主要看利润增速，今天大涨前戴尔的动态市盈率才22倍，市场倍来也没给他太高的估值。

@Megarsene @aleabitoreddit 什么意思？

戴尔财报电话需要关注的几点，说实话4月中在梳理CPU紧缺潮的时候是忽视戴尔了，没想到戴尔也是cpu紧缺的最大受益者之一。今天戴尔的电话会，赶紧来补个课。看看业绩暴涨是的核心逻辑是什么？1、戴尔COO“我们在全部三个领域夺取市场份额——如果算上AI服务器，那就是四个。所有四大主要业务：PC、服务器、存储、AI服务器，都在夺取市场份额。我们在赢。” 1）AI服务器：收入161亿美元，同比暴增757%。当季新订单244亿美元，积压订单达到创纪录的513亿美元。 2）传统服务器与网络：收入85亿美元，同比大涨92%。需求远超供应。 3）存储：收入43亿美元，同比增长8%。戴尔自研IP存储产品连续第五个季度需求增速跑赢市场。 4）PC（客户端解决方案事业部）：收入146亿美元，同比增长17%。连续第二个季度夺取市场份额，其中商用PC收入增长18%，已是连续第七个季度增长。 2、戴尔已经变成了AI基础设施的“集成商”。 电话会上，戴尔反复强调一个词：戴尔AI工厂（Dell AI Factory）。首席运营官Jeff Clarke介绍，两年前戴尔与英伟达联合推出了“AI工厂”概念。简单说，就是把AI训练和推理所需的全套硬件——GPU服务器、网络、存储——打包成一个可以快速部署的整体解决方案，交给企业直接用。 客户不是只买散件。“客户买的不是组件，他们买的是集成方案——能够快速投入生产、运行在他们自己控制的基础设施上、满足他们工作负载所需性能、安全和数据基础的方案。” 3、每个环节都在缺货 需求持续超过供应，而且这种需求是广泛的——不只是GPU，还有CPU、传统服务器和PC。” 供应链卡在哪儿？Clarke点名了四个环节：DRAM（内存）、NAND（闪存）、CPU（处理器）、硬盘。其中内存是最主要的瓶颈。他说：“每一颗内存都重要，每一颗处理器都重要，这就是我们每天在做的事。” GPU做神奇的工作，但有大量工作需要围绕它完成——IO处理、分支管理、状态维护，这些是串行的、顺序的工作，是CPU的活。CPU在Agent的每一个决策里都在循环运转。 每一次AI的调用、每一次决策、每一次状态管理，都需要一个“管理框架”来支撑。这个框架，跑在CPU上。AI正在“拉爆”传统服务器：一个没人预料到的新市场 4、大幅度上调全年指引 全年营收：1650亿至1690亿美元，中值1670亿美元，较此前上调约270亿美元，同比增长约50% AI服务器收入：全年目标600亿美元，约为上年同期的2.4倍 传统服务器：预计全年增长超60% 全年EPS（非GAAP）：17.90美元（±0.25），较此前指引上调约5美元，同比增长约75% 说实话这么大幅度的上调指引，在美股上市公司里也很少见。那只能说明业绩真的很爆 “我们的管道显示需求不是在放缓，而是在加速，并且大幅超过供应。”戴尔COO Jeff Clarke原话 当然跟川普关系密切，跟老黄互动很多，能带动情绪走高。但是只有超强，大幅度超出预期的财报才会让盘后暴涨接近40%

戴尔财报电话需要关注的几点，说实话4月中在梳理CPU紧缺潮的时候是忽视戴尔了，没想到戴尔也是cpu紧缺的最大受益者之一。今天戴尔的电话会，赶紧来补个课。看看业绩暴涨是的核心逻辑是什么？ 1、戴尔COO Jeff Clarke“我们在全部三个领域夺取市场份额——如果算上AI服务器，那就是四个。所有四大主要业务：PC、服务器、存储、AI服务器，都在夺取市场份额。我们在赢。” 1）AI服务器：收入161亿美元，同比暴增757%。当季新订单244亿美元，积压订单达到创纪录的513亿美元。 2）传统服务器与网络：收入85亿美元，同比大涨92%。需求远超供应。 3）存储：收入43亿美元，同比增长8%。戴尔自研IP存储产品连续第五个季度需求增速跑赢市场。 4）PC（客户端解决方案事业部）：收入146亿美元，同比增长17%。连续第二个季度夺取市场份额，其中商用PC收入增长18%，已是连续第七个季度增长。 2、戴尔已经变成了AI基础设施的“集成商”。 电话会上，戴尔反复强调一个词：戴尔AI工厂（Dell AI Factory）。首席运营官Jeff Clarke介绍，两年前戴尔与英伟达联合推出了“AI工厂”概念。简单说，就是把AI训练和推理所需的全套硬件——GPU服务器、网络、存储——打包成一个可以快速部署的整体解决方案，交给企业直接用。 客户不是只买散件。“客户买的不是组件，他们买的是集成方案——能够快速投入生产、运行在他们自己控制的基础设施上、满足他们工作负载所需性能、安全和数据基础的方案。” 3、每个环节都在缺货 需求持续超过供应，而且这种需求是广泛的——不只是GPU，还有CPU、传统服务器和PC。” 供应链卡在哪儿？Clarke点名了四个环节：DRAM（内存）、NAND（闪存）、CPU（处理器）、硬盘。其中内存是最主要的瓶颈。他说：“每一颗内存都重要，每一颗处理器都重要，这就是我们每天在做的事。” GPU做神奇的工作，但有大量工作需要围绕它完成——IO处理、分支管理、状态维护，这些是串行的、顺序的工作，是CPU的活。CPU在Agent的每一个决策里都在循环运转。 每一次AI的调用、每一次决策、每一次状态管理，都需要一个“管理框架”来支撑。这个框架，跑在CPU上。AI正在“拉爆”传统服务器：一个没人预料到的新市场 4、大幅度上调全年指引 全年营收：1650亿至1690亿美元，中值1670亿美元，较此前上调约270亿美元，同比增长约50% AI服务器收入：全年目标600亿美元，约为上年同期的2.4倍 传统服务器：预计全年增长超60% 全年EPS（非GAAP）：17.90美元（±0.25），较此前指引上调约5美元，同比增长约75% 说实话这么大幅度的上调指引，在美股上市公司里也很少见。那只能说明业绩真的很爆 “我们的管道显示需求不是在放缓，而是在加速，并且大幅超过供应。”戴尔COO Jeff Clarke原话 当然跟川普关系密切，跟老黄互动很多，能带动情绪走高。但是只有超强，大幅度超出预期的财报才会让盘后暴涨接近40%

目前全球规模最大的杠杆单票ETF，是一只提供SK海力士（港股代码：7709 HK）两倍多头敞口的产品。 与此同时，韩国本周新上市的SK海力士单票ETF和三星电子单票ETF在上市仅两天内，分别吸引了约12亿美元和8.16亿美元的资金涌入，认购热度罕见。 恭喜先行者南方东英！

OpenAI is reportedly preparing a pocket-sized device… interesting. https://t.co/zRWdHmJO8s

@mingchikuo Is this true? "For the new iPhone 18 Pro / Pro Max in 2H26, the variable aperture lens reportedly carries a 50% higher ASP than the high-end 7P lens, with Sunny’s supply share expected to reach 40–50%." Has Largan been displaced?

https://t.co/IKLk4NVYX7

Samsung Electronics: "Taylor Fab Ready for Operation... Mass Production Next Year" Samsung Electronics has reaffirmed that it will begin ramp production at its Taylor fab in Texas, USA, starting next year. Margaret Han, Vice President of Samsung Electronics' Foundry U.S. division, stated at the SAFE (Samsung Advanced Foundry Ecosystem) Forum held at Samsung's U.S. headquarters on the 28th (local time), "Customers will begin production at the Taylor fab starting next year," adding, "We are ready." Previously, during the first-quarter earnings conference call this year, Samsung had described the Taylor fab as "under construction," with a goal of timely operation within the year. Construction of the Taylor fab began in 2022. The total expected investment is $17 billion (approximately ₩25.5 trillion), and it will produce Tesla's autonomous-driving chips AI5 and AI6, among others. The fab will also adopt the Advanced 2nm process. Han said, "We will strengthen our long-term production capacity at the Taylor fab," and "We will establish the most advanced 2nm production capacity at Taylor Fab 1 this year." Samsung is analyzed to be building out the SF2P+ (second-generation enhancement) process at the Taylor fab. Samsung had previously announced that SF2P+ would enter operation in 2026–2027. The 2nm process is divided into SF2 (first generation), SF2P (second generation), SF2P+, and SF2X (third generation for AI/HPC), among others. With each successive generation, power, performance, and area (PPA) improve. SF2P+ is a process that improves PPA while maintaining the same design IP as SF2P. It has been finely tuned for AI workloads. Performance is reportedly improved by up to 30%. Detailed information on the leading-edge processes is scheduled to be shared with partner companies this coming July. Samsung will hold the SAFE Forum again at its Seocho headquarters in July. The SAFE Forum is an occasion where Samsung and its foundry partners share technology trends across fields such as IP, design, and packaging, and strengthen collaboration. Later that same month, Samsung is reported to host the Samsung Foundry Forum (SFF). The Foundry Forum is an event where Samsung unveils its foundry technology roadmap to partner companies.

@jake_ayes https://t.co/BeeKbgPMXR

This is a really remarkable report from UBS. I was genuinely stunned by a lot of it. In particular, the point that Kioxia's NAND wafer cost is $2,000 — only half of Samsung's — and that Kioxia's lateral (horizontal) scaling architecture is what gives the company its industry-leading cost advantage was something I really had no idea about. There's so much else in there too — HBF, the outlook for the NAND flash market itself, and more.

@SleepNotPayTax 怎么还是腾讯

@syusaizhuqi Q3开始出货，Q4放量爬坡，2027年上半年继续加速出货。老黄前几天讲的，Vera Rubin的整个生命周期内都将处于供不应求状态。

@chrisw09010901 👍

@musukataisakabu reason?

Is there any reason why wafer companies are surging today? https://t.co/SjhvajgNKM

AI需要数据库（MDB），需要实时可扩展、AI-native的数据云底座（SNOW），应该也需要AI需要更强、更智能的防护墙”。AI时代的安全需求不是“替代”，也会爆发式新增，逻辑在于： 1）AI Agent、云迁移、多模态数据让攻击面大幅扩大（数据泄露、模型投毒、提示注入等新风险）； 2）企业部署AI必须同步加强“零信任”、云安全、端点检测（EDR）、AI安全平台； 之前大摩测AI催生的增量安全机会高达2200亿美元，远超AI可能颠覆现有厂商的负面影响。 2月中的时候，也明确聊过安全类软件是被误杀的。Agent发展很多原来的软件应用场景都会被Agent替代掉，安全是硬需求，而且agent越普及安全需求会更高。 现在回看2月份这条聊安全股的推文，几乎是Call在安全从去年十月调整跌下来的最底部。当时聊到的cloudfare(net)、CrowdStrike（CRWD)）、Palo Alto Networks（PANW）、Zscaler（ZS）、Fortinet（FTNT）在过去三个月走势都很不错，CRWD接近翻倍。 最近市场对于SaaS的看法不再是整体的一遍倒的负面评价，大家接受AI越发展并不是所有的软件公司都会被冲击，开始区分和筛选哪些是AI受益的软件板块，哪些是AI受损的板块。个人角度，安全应该就是AI受益而非受损的板块。

听人说我在中文圈都快”闪光宝可梦”了, 哈哈哈。 拜托我也就是个普通人好吗！ 平时也就是随手写点自己的想法而已。感觉最近关注度高了，也莫名其妙招来了不少黑子和瞎编的谣言 跟最近那些胡说八道的传闻相反，其实我在推特上从来没删过任何一篇分析帖。 大家估计是把我跟哪个喊单 $IREN 的大board member给搞混了吧。 我就是喜欢找点不一样的选股角度，先让行情走一段，最后再看看到底打没打脸 我觉得最核心的无非就是踩中了下个大叙事，比如去年的 Neoclouds 和光子板块，或者今年的 800 vdc 和 CPO。 不过… 我还是非常感谢大家温暖的支持！

听人说我在中文圈都快”闪光宝可梦”了, 哈哈哈。 拜托我也就是个普通人好吗！ 平时也就是随手写点自己的想法而已。感觉最近关注度高了，也莫名其妙招来了不少黑子和瞎编的谣言 跟最近那些胡说八道的传闻相反，其实我在推特上从来没删过任何一篇分析帖。 大家估计是把我跟哪个喊单 $IREN 的大V给搞混了吧。 我就是喜欢找点不一样的选股角度，先让行情走一段，最后再看看到底打没打脸 我觉得最核心的无非就是踩中了下个大叙事，比如去年的 Neoclouds 和光子板块，或者今年的 800 vdc 和 CPO。 不过… 我还是非常感谢大家温暖的支持！

@curtuner 自动驾驶还没到ChatGPT时刻，地平线扩张太厉害，以为自己血槽厚。地平线的J6架构很一般，然后负责芯片设计的陈鹏走了。

美伊现在舆论战玩的很溜： 前天晚上伊朗放出一版协议备忘录，美国否认； 昨晚美国放出一版协议备忘录，伊朗否认。 不过市场都买单。 本质上都是在通过舆论施压和影响对方，双方想达成的意愿很高，但都不想关键问题上让步或者说即使让步也要考虑国内国际上的影响。通过舆论战，一方面表明态度，另一方面也是向国内其他立场派别喊话，打预防针，美伊皆如此。 还是之前这里https://t.co/zL4AWawz9t聊到的：美伊双方还是越来越有可能达成昨天聊到的去年中美贸易战的休战模式。 本条由@bitget_zh赞助，「Bitget 买美股：秒级入场，丝滑交易 」

@ShuaiShao12 关闭港股交易权限，防止手贱。勾选“日本”交易许可。

地平线机器人已经跌到5港币了，光头余什么感受？

港股就是屎坑，即使某家公司看起来很好，动手前也要冷静下。一个诚恳的建议。

Blue Origin 的失败 变相利好 SpaceX 和Rocket Lab。坐火箭是一件门槛非常高的事。而目前稳定安全的供应商就只有这么几个

事实告诉我们如何最快成为股神——先当选美国总统

买了一堆韩股日股的问题就是，看着日元/韩元价格有点茫然。每天都疯狂做汇率兑换成美金来看。

川普的国家资本主义正在四处开花，这次是无人机。看华尔街日报昨晚报道川普政府正在谈判通过五角大楼战略资本办公室资助美国无人机制造商。相同的11亿美元无人机主导计划，目标到2027年生产30万架攻击无人机。 简单梳理下美国的无人机生态： 1）直接受益者： $UMAC：与特朗普Jr.有关联，已赢得陆军电机合同。 $RCAT：Teal Drones，蓝UAS名单，位于小型无人机采购领域。 $AVAV：Switchblade和Puma，美国上市的最大攻击无人机纯正玩家。 $KTOS：XQ-58 Valkyrie，更广泛的无人系统推进。 $ONDS：军事和国土安全无人机系统。 2）间接供应链： $KULR：来自无人机主导计划客户的2026年4月FPV电池订单。2026年最高500万美元。 $KOPN：320万美元Sentinel FPV光学模块订单，到2028年潜在4万副护目镜。首次正式进入军事FPV领域。 看起来小型无人机生态系统也得美国制造，川普概念股正在外溢

澜起科技投资方的减持已被接了一堆买方接了。冲个400不过分吧？

Opus 4.8 来了，有 5 档选项，一时间不知道怎么选，看了头疼 🤦 让它自己来介绍了一下，大家可以参考 为啥就不能有一个 Auto 的选项呢，根据任务的难度来自动选择 https://t.co/VisO5rLBBa

之前定投的 $IGV 终于开始涨了 AI 时代，的确大部分软件都会被替代掉，但通用大模型之外，至少有 2 类机会： - 补齐大模型能力的 Infra: 比如支付、安全、搜索… - 极其垂直的场景，需要额外的努力来获取 context，形成完整解决方案：金融、医疗、教育、军工… 本质上是围绕 token 的生意: token 生产、加工和转卖，附加值最高的还是在转卖环节 按照Cloud云计算的商业规律，软件应用层还有很大的增长空间，春天布局，秋天收获

@qcl92429686 Taiyo Yuden也很好。Musashi这个月也翻倍了。本周自从我周末喊它，已经接近60%。

Another 10 pct today. Great week for Musashi. https://t.co/ySbdBjQuMq

Jensen, please bless Ibiden too.🙏

https://t.co/rVUzI49t6v

Samsung Electronics Eyed for Anthropic Chip Order… Foundry Revival Begins Samsung Electronics has made an investment in Anthropic, the U.S. AI company behind the AI model "Claude," raising the possibility of a foundry partnership in which Samsung would manufacture Anthropic's chips on a contract basis. Samsung's foundry business, which has run at a loss for several years, has recently landed a string of major global customers, fueling growing expectations for a return to profitability next year. On the 28th (local time), Anthropic announced that it had raised $65 billion in its recent Series H funding round, with its post-money valuation assessed at $965 billion (about 1,440 trillion won). The three major memory semiconductor manufacturers—Samsung Electronics, SK Hynix, and Micron—participated in this funding round as "strategic infrastructure partners." Anthropic emphasized that "the technologies of these companies play a critical role in the global supply of memory, storage, and logic chips," adding that "partnerships with them will be a great help in reliably scaling computing capacity to meet customer needs." Anthropic's mention of "logic chips" in its announcement has prompted speculation about a foundry partnership with Samsung Electronics. The process of manufacturing logic chips is foundry work, and aside from Samsung Electronics, neither Micron nor SK Hynix has a foundry division. For this reason, the industry analyzes that Samsung Electronics and Anthropic will newly collaborate not just at the memory level but also in foundry. This implies a high likelihood that Samsung Electronics will produce the AI chips used in Anthropic's flagship model "Claude." Anthropic, alongside OpenAI, is regarded as one of the world's leading AI model companies and a core AI player. The prevailing view is that Samsung's foundry has secured yet another major customer, following Tesla, NVIDIA, and others. Earlier, Samsung Electronics won orders for Tesla's next-generation AI chips, "AI5" and "AI6." Samsung's foundry is also handling production of an upgraded version of "AI4." NVIDIA's inference-dedicated language processing unit (LPU) chip "Grok3" is likewise being produced through Samsung's foundry. Samsung Electronics also plans to supply the image sensor to be mounted in Apple's new iPhone next year. Accordingly, attention is focused on whether Samsung Electronics' foundry business—which has run at a loss in recent years—can regain its vitality. As of last year, Samsung Electronics held second place (7.2%) in the global foundry market, but the gap with first-place TSMC (69.9%) reaches 62.7 percentage points. Regarding Samsung's investment in Anthropic, the industry views it as strategic cooperation spanning "AI semiconductors, memory, foundry, and AI infrastructure." Samsung Electronics is positioning a "one-stop solution"—encompassing high-bandwidth memory (HBM), advanced foundry processes, and advanced packaging—as its differentiation point. As the AI market evolves into an ecosystem centered not only on HBM supply but also AI chip design, production, packaging, and system optimization, the move is interpreted as Samsung strengthening partnerships by leveraging its competitiveness across the semiconductor spectrum. A semiconductor industry official said, "Samsung's investment is not a simple equity stake in an AI company, but is interpreted as a signal that Samsung is earnestly expanding strategic relationships with the key players of the AI era," adding, "Expectations are growing as to whether Samsung's foundry business can seize another opportunity amid the expansion of the AI market."

Damn, I couldn’t buy Dell because I didn’t have enough cash. https://t.co/HxlUjoFUUI

@blackchoic12035 @aleabitoreddit 之前有海外地址证明就可以开

@Wutarded I kinda gave up after blocking 90 of them under 1 post. Hope X does something

$HPS.A is close is to a 2M timeframe. Transformers in the Sky are up 83.3% since. Bumblebee go brrr? But in all reality, transformers demand visibility is very high, solid backlog, high market share (of dry). Not exactly parabolic, but looks like a compelling compounder. https://t.co/bNGwjqfZ4M

I’m actually even more bullish on $AAOI at $13B MC given all the recent laser bottlenecks… Than I was back at $2B or $6B. I also think markets missed the analyst note around potential long term supply agreements with $NVDA or $AMD. If they’re projecting $471M in H1 2027… that’s absurd ramp. But of course the $600M ATM is a short term overhang. Just a matter of waiting time? Given it’s more about keeping up with demand… So more of a matter of how much they can make. Probably my favorite US-based photonics long stock now that I own.

Holy crap… My $EWY 2028 leaps are now up 428%+. So 5.2x ROI trade, in 3 months. Thats what happens when IV increases and underlying memory assets in SK Hynix / Samsung also appreciate. https://t.co/XdE21DOyoz

@SaintAgnesBets Lot of my for you are just a bunch of EU locals shorting their own frontier companies, while US institutions buy ownership. Very interesting to see https://t.co/CUehLRtPkU

@MatteoDeDycker $600m ATM, still long term bullish on $AAOI

@Frenchie_ Basically, so much negative coverage on $SOI just for it to go up 340%+ afterward. It’s weird Europeans keep shorting their own frontier companies, rather than just letting it grow.

Always a good time interacting with European outlets. Back at $44 EUR, European media said $SOI was an: - -“overvalued stock” - “buying now would undoubtedly be a big timing errors” - “purely speculative bases” And traditions analysts had no clue why it had risen. $SOI is now up 342% since by the way.

CLAUDE OPUS 4.8 JUST DROPPED https://t.co/BtDrCAtCMT

“我们的数据中心业务燃爆了”，“订单异常强劲”，这是MRVL的CEO在财报电话会上的原话，记录下电话会上值得重视的要点： 1、连续多个季度上调指引 1）Q2指引方面，公司预计营收约27亿美元（上下浮动5%），同比增长约35%，高于分析师此前预期的26亿美元 2）Marvell将2027财年营收预期上调至约115亿美元，同比增长约40%。三个月前，公司给出的指引是“接近110亿美元” 2、数据中心业务增速在加速 1）Q1数据中心营收18.3亿美元，同比增长27%，环比增长11%，占总营收76%。 Marvell对这块业务的增速预测是： FY2026：+46%（已实现） FY2027：约+50% FY2028：约+55% 2）互联业务 AI数据中心互联（Interconnect）业务是Marvell数据中心业务中体量最大的部分，覆盖光学互联、DCI模块、相干光等产品线。 这个业务的年度增速预期，在过去几个季度里被连续上调：去年9月前后约为30%，之后上调至50%，如今再度上调至超过70%。 CEO的原话是“Murphy在被分析师追问时直接说： 我认为这里存在大量向上偏的可能性。我们的传统DSP业务明年将有大幅跃升，1.6T产品线是更高价值的内容，DCI在加速，还有retimer、AEC这些新业务，以及规模化上行光学（scale-up optics）……这是我们一个重大增长周期的开始。” 3、定制芯片 FY2027定制芯片营收：同比增长超20% FY2028定制芯片营收：预计同比翻倍，高于上季度预期 FY2029目标：超过100亿美元（此前目标为约80亿美元） 这意味着FY2028定制芯片营收在40亿美元以上，然后到FY2029跳升至100亿美元，也就是单年增量超50至60亿美元。 4、与英伟达扩大合作，三大方向落地Marvell宣布与英伟达扩大战略合作，电话会上详细介绍了三个核心方向： 1）光学互联合作：Marvell长期为英伟达提供DSP、TIA和驱动器，双方现在进一步合作开发硅光子技术，这被认为是纵向扩展（scale-up）网络的关键使能技术。 2）NVLink Fusion集成：允许Marvell构建能与英伟达基础设施无缝对接的自定义芯片和网络半导体。Murphy表示，这为超大规模云厂商提供了更大的灵活性，可以在自定义芯片和英伟达芯片之间自由混搭，"Marvell独特地提供了这两种架构之间的桥梁"，并将为双方创造新的市场机会。 3）AI-RAN：Marvell将增强其Octeon基站处理器，使其能与英伟达 GPU直接协同工作，在同一硬件平台上同时运行5G/6G无线工作负载和AI应用。 除了之前市场公认的联接/光互连、定制芯片两大业务板块之外，现在又多了一个AI-ran的边缘计算业务。

🤡 https://t.co/2naCT63igG

@xenoxidal1234 No, ByteDance is building this in-house.

国内头部PCB厂商的资本开支在加速。9家头部企业2025年合计capex 267亿，同比+111%。2026年Q1单季度125亿，同比+182%，斜率还在变陡。鹏鼎2026年全年规划168亿，2025年是66亿。胜宏25Q4单季度接近30亿，环比翻倍。深南25Q4也有14亿，环比+76%。而且目前主要是胜宏、沪电、鹏鼎三家在拉，深南、景旺、方正、广合还没真正放量，后面有接力。 产能扩张去向高度一致。高阶HDI、HLC、mSAP、类载板，全部对准AI服务器和高速交换机。传统消费电子PCB没有新增投入。东南亚建厂同步在落地，鹏鼎泰国43亿，深南泰国12.74亿，胜宏在泰国、越南、马来多地同时动工，北美CSP对供应链分散的要求已经变成了资本开支的真金白银。 这些钱流到上游之后，谁拿到了超额利润。 钻孔、曝光、电镀三个工序合计占PCB设备投资的44%左右，每个环节恰好有一家国产公司占着位置。 钻孔是这轮变化最大的工序。AI服务器PCB层数从十几层往30层以上走，18层以上多层板2025年增长超50%。层数翻了，钻孔量也跟着翻，背钻精度和超高厚径比成了硬门槛。大族数控的CCD六轴背钻机搭载3D背钻，已经在多家高多层板龙头量产。超快激光在1.6T光模块上拿到了订单。2025年PCB设备收入57.73亿，+72.68%。26Q1末存货较年末涨了28%，合同负债也在涨，接单和备货同时在加速。 曝光环节和mSAP的渗透速度直接挂钩。线宽线距压到20μm以下，传统菲林曝光就做不了了，必须换LDI。芯碁微装2025年营收14.08亿(+47.61%)，MAS 6P的载板级解析能力做到6μm，效率比国际主流同类设备高50%以上。26Q1累计订单突破8亿，营收预增91%-123%。另外WLP直写光刻已经拿到头部客户量产订单，等于同时切入了半导体先进封装。 电镀方面就是东威科技。高层数PCB通孔长径比大幅提高，垂直连续电镀变成标配。东威在这个细分市场的份额超过50%。2025年营收10.98亿(+46.45%)，26Q1合同负债8.72亿，同比翻倍。下半年新厂房投产，水平三合一和VCP合计新增产值空间超过10亿。 PCB扩产计划高度同质化，几乎所有人同时在扩高阶HDI和HLC，全在抢英伟达和北美CSP的单。如果AI服务器需求增速有任何放缓，这些集中投产的高端产能利用率会很难看，而且产线降维做低端板几乎不可能。上游设备的处境完全反过来。CCD背钻成套方案只有大族能做，载板级LDI只有芯碁微装做到6μm以下，VCP电镀东威份额稳定在50%以上。下游越同质化地扩，上游这几家的议价权反而越强。 需求端的节奏。英伟达已经确认Vera Rubin全面量产，2026年下半年开始向全球CSP发货。摩根士丹利的数据显示，Rubin平台上GPU在整机柜成本中的占比从Blackwell时代的65%降到了51%，PCB、MLCC、ABF载板这些配套环节的价值量反而在爆发。沪电和英伟达已经在测试下一代M10材料，目标是Rubin Ultra和Feynman平台的正交背板，78到104层。供给端同样有明确的投产节奏，沪电43亿的新产能2026年Q3试产，爬坡半年后2027年满产。这意味着2026年下半年到2027年，Rubin发货放量和新产能投产几乎同步发生。如果Rubin Ultra和Feynman的节奏不断档，设备采购就不是一个脉冲而是一个持续的台阶。

@CertainBasket 主动国产化。目前BESI的设备还没被荷兰政府明确禁运。

https://t.co/bIaS7VaTL3

Micron EVP and Chief Business Officer Sumit Sadana: Even with all the new investments, meaningful memory chip supply will only start coming online in late 2027 and ramp up in 2028. $MU

@Yuchenj_UW Someone told me to pay them money and they’d stop spamming me…

@Yuchenj_UW 👏

"New demand is emerging for in-chip thermal conductive materials. SK hynix’s HBM4 reportedly faces relatively significant heat dissipation issues, and a Japanese supplier has quoted RMB 6 million per ton for a new type of filler material. By comparison, the value of conventional filler materials is typically around RMB 100,000 to 1 million per ton." What exactly is this material? I don’t think it’s iHBM.

@TJ_Research 最近的感受是gpt 5.5 Pro非常好，准确率也挺高

@fanli819 Yes, Germany’s leading ficonTEC is a 100% wholly owned Chinese subsidiary now. Because RoboTechnik over in Suzhou bought them for $260m… because the company lacked funding and Germany/EU wouldn’t support them with enough capital. So China is a large beneficiary of the photonics upside, and will likely do well now over in Shenzhen’s stock exchange.

混合键合设备那部分估计以后都是拓荆科技的

Arm又创新高了，媒体报道字节也要自研CPU了，其实也不用考了，选Arm就得了。毕竟英伟达的Vera CPU，谷歌、亚马逊、微软、meta的自研CPU都是用arm架构。 https://t.co/9EKzqGteII

Just a shower thought to European media: Maybe instead of complaining about the influx of capital. To keep your critical companies alive and subsidize growth. Maybe welcome it… So you don’t keep selling off your monopolies or critical chokepoints like ficonTEC from a lack of funding to foreign countries such as China. Especially when they’re necessary to US hyperscaler supply chains.

@Gxba4aUBU7d8TvO 芯原是有做5nm/7nm后端设计的能力，但是Server CPU芯原好像之前没做过。不知道字节会不会找他们做后端/物理设计。

@Sigurd_Drivenes This type of narrative-add BS reporting is partly why trust in institutions/media has been crashing in the US and around the world. https://t.co/jPnwJfKfCn

Reuters今天报道字节正在自研CPU，同时评估Arm和RISC-V两条路线。 Agentic AI让server CPU的角色变了。每个agent session里CPU跑的是prompt组装、JSON解析、tool call调度、状态管理这些串行逻辑，单线程不重，但一台推理服务器上千个session并发，host CPU直接卡在关键路径上。大厂自研推理芯片之后，配套的CPU自己做是顺理成章的事。Google拿Axion配TPU，AWS拿Graviton配Trainium，字节走到这一步不意外。 如果走RISC-V，香山昆明湖是目前起步门槛最低的server级开源RISC-V核。开芯院的联合开发模式让字节不用从零画CPU微架构，直接在一颗验证覆盖率接近100%、已有下游企业流片验过的核上面做SoC集成。从CPU核到一颗能量产的server芯片，中间还隔着DDR5控制器、片上互连、PCIe/CXL、RAS、整颗SoC级验证，百人团队加两三年的投入跑不掉。但跟从零起步或者受制于商业授权比，成本结构完全不一样。 昆明湖在国产CPU里什么水平？基准IP核大约15分/GHz（SPECint2006），进迭时空基于昆明湖定制的X200核已经做到16分以上。仅次于龙芯3A6000的17-18分，高于海光Zen1的12-13分，甩开鲲鹏920约40%。龙芯IPC最强但LoongArch生态做server不现实，海光Zen1授权冻结了没有后续。昆明湖是国产阵营里唯一IPC够用、同时还能持续迭代的server级CPU核。 Tenstorrent的Ascalon-X更强，22+分/GHz，比昆明湖高47%。但那是对标Neoverse V2/V3的性能核，8-wide decode加双256-bit向量单元，为HPC设计的。Agent推理的host CPU大量时间在等I/O，瓶颈在并发session数和内存带宽，单核IPC到15分/GHz就够了。性能核跑agent orchestration，相当于拿跑车当公交车开，😇。 昆明湖15分/GHz配64核甚至128核的吞吐方案，对agent场景性价比更高，和Graviton选N系列效率核的逻辑一样。开源让字节可以自己动微架构、加定制指令、调NoC和内存子系统，商业授权IP做不到这一点。今年3月阿里达摩院已经宣布玄铁参与下一代香山昆明湖v3的联合研发，生态在加速聚拢。进迭时空两周前刚宣布X200核完成研发，第四代X300瞄准2027年超级智算数据中心场景。字节现在入局联合开发，时间窗口刚好。

Few months ago, a European publication called my $RPI idea: - “mass stupidity” And said: - $RPI shares would: “come crashing back to reality” Then called it a: - “Meme stock”. Earnings report came out? Blew away revenue expectations. It’s very interesting that media can just write all this slander… Then pretend it never happened when it ages so badly. Amount of media slander from $SOI to $SIVE has been pretty incredible.

@Sofigoodboy @TechIsInsane Yeah, already discovered the other day some were launching bot farms on X to spread disinformation. Lot of money is at stake I guess, especially before Nasdaq/MSCI structural inflow next week.

@duedilligently I haven’t seen anyone credible doubt Win Semi’s ability to scale capacity. They’re extremely dominant, and are a critical part of current supply chains from SpaceX to $AVGO.

美股最近频频老树开新花，从诺基亚到Dell又到IBM到黑莓...

$SIVE is the most compelling CPO/photonics exposure to me. Addressing the disinformation: I haven’t sold and don’t plan to sell a single share. I do think this ends up the next $80B+ $LITE one day from ~$2.1B. And I personally have plans to acquire more ownership + support their M&A prospects. I believe earnings transcripts will be strongly positive. As in the part few months we’ve discovered: > AlChip/Amazon private placements, which is positive for Ayar -> $SIVE implying Trainium 4 design in > Wiwynn + Ayar CPO scale up > $JBL 1.6T optical transceiver ramp with Sivers incoming faster than markets expected (with relatively dramatic moat + demand as much as they can produce) > O-Net scaling up ELS efforts with $SIVE > $YSS acquisition of $SIVE allspace lead partner, designing Sivers into Space defense primes > New CHIPS ACT funding for $SIVE > $POET H2 volume ramp and their new $50m -> $500m order (with $SIVE as light source) > information discovery around $AAPL using $SIVE lasers for next gen consumer devices > information discovery around links to Lightelligence (went public $10B+ MC) + Lightmatter as likely customers. > Celestial volume ramp with $MRVL indicators. > new customers working on TFLN with $SIVE like Lightium > $AMD going with $GFS for CPO, and GFS listing sivers as one of two laser suppliers > Ayar removing $MTSI / $LITE from their website and signaling $SIVE as primary source/sole source > Ayar raising $500m for volume ramp (intel, Mediatek, Nvidia, amd etc) > pluggable TAM expansion signaled from 2025 annual report > Nasdaq listing expected soon > MSCI small cap index / Nasdaq omx inclusion, making Blackrock, Vanguard and others passive buyers > M&A signaled from 2025 annual report + 2 new board members that have experience in that area > $NOK as likely customer from 2025 annual report. > $LITE getting cw bottlenecked from EML contracts, $SIVE signaling capacity agreements in place with Win, making the a likely bottleneck owner + chokepoint in CPO sector. All of this market research was done before earnings. Any results is just confirmation of supply chain mapping done. I don’t think anyone cares about former quarter revenue since $SIVE is an exceptionally compelling 2027 long, especially H2 onward. Only thing I’m looking at are: > TAM expansion of the overall photonics supercycle (eg. optical engine, ELS, pluggables) either from M&A or developments > volume ramp expectations from existing companies > Nasdaq listing timelines for more liquidity to support their M&A efforts > any new customers signaled for CPO/Pluggables

$MRVL 电话会要点： 1️⃣收入预期上调主要由数据中心业务驱动，预计该业务本财年将增长约50%。其中，互连业务预计同比增长超过70%，显著高于此前50%的预期。 2️⃣互连业务的增长将继续快于云CapEx，主要受益于scale-out网络对1.6T的强劲需求，以及scale-up和scale-across网络的更大贡献。 3️⃣和Nvdia的合作方向： -光学合作：从DSP、TIA和驱动扩展到硅光子技术 -NVLink Fusion整合：支持Marvell开发与NVIDIA无缝互联的定制芯片 -AI RAN：将AI与无线基础设施融合 4️⃣关键战略： 加大AI投资 锁定产能（预付款约10亿美元） 持续回购股票 5️⃣定制业务是下一个增长点： 2027年增长 >20% 2028年增长 >100% 预期2029财年收入 >100亿美元 $MRVL 的财报进一步确认网络需求的增长以及对于光学需求的确定性，互连 + 光学是 Data Center 里增长最快、最频繁强调的板块。 这和昨天的 $SMTC 财报是相互印证的。 比较特殊的是XPU的增长性，目前只是Data Center 里的一部分；但如果按照管理层的预期加速（>100% y/y），再到 FY29 走向 100 亿美元规模，那会成为下一个增长引擎。 补充一些研报数据： -在 1.6T 光模块市场中，Marvell 在 DSP、TIA 和驱动芯片上的整体市占率约为 60%。 -在 800G 市场中，Marvell 在 DSP、TIA 和驱动芯片上的市占率约为 70%（主要原因是 Broadcom 会优先把资源投向 1.6T DSP 市场，导致其在 800G 上相对保守）。

The process looks highly irregular. https://t.co/AeBPPlGrAB

Bruh, according to Geekerwan’s TEM analysis, Samsung’s 2026 SF2 still lags slightly behind Intel’s 2025 18A in terms of geometry control and process consistency. $INTC https://t.co/fsL07lJHhJ

Bruh, according to Geekerwan’s TEM analysis, Samsung’s 2026 SF2 still lags slightly behind Intel’s 2025 18A in terms of geometry control and process consistency. https://t.co/dBS1uoneSo

@Justin_Clanky Yep! Definitely getting a lot of random real life threats and personal attacks due to recent popularity. Which is why I prefer to stay anonymous for now. Who knew publishing free information discovery to the public would invite this type of response? Anyway just here to help out individuals, I just threat that as noise along the way.

IBKR 邀请链接 https://t.co/lAoO3SB9QT ，注册最多可获得价值 1000 美元的 IBKR 股票

. @aleabitoreddit 推荐的不少【供应链瓶颈】都是欧洲股票，刚刚他/她也写了一个复盘，涨幅在 2-20 倍不等 分享一个信息差，IBKR 上可以交易全球各个市场的股票，包括欧洲市场，具体设置：交易设置→交易许可→股票→选中你想交易的市场→申请开通 申请之后，一般需要 1-2 个工作日 需要注意的是, 一些小票流动性不好，最好使用限价单，免得下单滑点过大 也欢迎走我的 IBKR 邀请链接，最多可获得价值 1000 美元的 IBKR 股票，链接放在评论区

@cchronobreakk I have something planned for $HOOD retail users and international equities like $SIVE, don’t worry.

@TBU12345678 NVIDIA has placed additional HBM4 orders with SK Hynix. Does Samsung really have the best-performing HBM4?

This growth is just astronomical, I’m at a loss for words. 450,000 followers now, and 40,000 subscribers. Only 7K more subscribers and I become the #1 on the entire X platform? Thank you everyone. I think this growth is just testimony to how people see free and actually compelling research democratization. Compared to old models where any alpha gets sold to institutions for tens of thousands… then retail only find out hundreds of percent later. Or where “influencers” have expensive paywalls to sell snake oil. Hope to set a long lasting example by publishing research / ideas / thought processes for free. And encouraging positive sum growth by focusing on substance.

@MikeFritzell You got it exactly right.

Just in case people are wondering about my track record with European equities: $RPI: $280 -> $800 (agentic AI hardware demand thesis). $LPK: ~$6, thesis at $13 -> $24.2 (glass cores substrates close monopoly) $SOI: $44 -> $181 (silicon photonics, monopoly over substrates) $SIVE: $4 -> $71 (CPO, critical chokepoints over lasers). $IQE: $12 -> $47 (latent epiwafer capacity, information discovery around downstream photonics companies). $ALRIB: $5 -> $15 (duopoly, synthesis around quantum buyers with photonics growth verticals). And now $XFAB at $9. I’m not always right. But every single one of my European longs thesis have been validated so far by either earnings, investments (eg. $MTSI in IQE) or market returns.

@SleepNotPayTax 董小姐一定不买了🙈我是想给家里加个小冰箱放放饮料和冰淇淋🍦

从 $DDOG 到 $SNOW ，最近的软件股交出漂亮财报的已经迎来了相当的反弹。 $NOW 会不会是下一个呢？ 总统概念加扎实业绩🧐🤔

@SleepNotPayTax 我也看了下能买啥🥲感觉可能农作物大宗和冰箱家电吧

@reedy1991 I mean all my European longs from $SOI, $IQE, $RPI, $SIVE, $ALRIB followed this trend if we just look at chart. Long term they’ve all ended up directionally right, and now they’re all up hundreds of percent. I don’t control or trade volatility with $XFAB, but interested in being a part of their long term growth vectors with power semis/photonics.

@Orcawellian There was a wave of negative news when I first went long on $SOI that caused unnecessary volatility. It’s now up 300%, turned out institutions bought up the float, after my thesis. Regardless of unintended volatility, my goal is to secure critical nodes critical to the US supply chains like $XFAB.

I genuinely think $XFAB is very compelling at $1.5B MC, despite recent volatility. Per NIST filings stated they were the only high volume SiC foundry in America. Making them extremely “critical infrastructure”… Verbatim from the US Gov. And they’re the first comprehensive pure play foundry for SiC/GaN. So it’s very compelling exposure as $NVDA pushes 800 VDC and as other power semi players from $NVTS and $WOLF are all re-rating hard. For long term photonics exposure: They were literally listed in CHIPS Act 2 blueprints… with $NOK / $NVDA evaluations right now. And production ramp up in 2027/volume production H1 2028 expected. So you have a company at $1.5B MC: Critical both on the power semis front to the US government. And critical to both photonics front to the EU government. Coming off of a legacy drag cycle with auto and others (similar to Soitec). I think I’d follow EU/US government signals for what’s critical infrastructure given expected dual continent subsidies… Over random media analysts trying to cause unnecessary volatility saying it’s a memestock with no fundamentals.

@mazo7846 This is a pretty stupid way to dismiss the core idea of every long I hold.

Why Are Silicon Capacitors Becoming Important? And Why Is SEMCO's Stock Surging? The reason silicon capacitors have failed to grow into a large market until now is clear. TSMC, which holds an overwhelming share in substrates for AI packaging, has historically sourced silicon capacitors in-house. As a result, it has been effectively impossible for silicon capacitor suppliers outside of TSMC to break into the AI packaging value chain. But with the rise of Intel's EMIB, the situation is changing. Silicon capacitor suppliers like Murata and SEMCO are now supplying silicon capacitors for Intel's EMIB, and have begun to benefit directly from AI demand. The existing silicon capacitor market was effectively an oligopoly dominated by Murata and TSMC. SEMCO was closer to a latecomer. Yet even SEMCO—the latecomer—has managed to land a supply contract worth more than $1 billion. For context, SEMCO's silicon capacitor revenue last year was a mere few million dollars. So how did SEMCO ramp its revenue this quickly? It comes down to the fact that SEMCO is the only company in the world that operates both a silicon capacitor business and a substrate business. The reason silicon capacitors have become important for Intel's EMIB is that Intel cannot make its own silicon capacitors and must source them externally. On top of that, within Intel's EMIB BoM, the highest value-added component is the substrate. So if a company can deliver both the substrate and the silicon capacitor to Intel's EMIB on a turnkey basis, that maximizes SEMCO's bargaining power and value. This is precisely why SEMCO's stock has been surging recently. Of course, this isn't to say Murata will benefit less. Murata still maintains a very close relationship with Ibiden—Intel's first-vendor for substrates—and is clearly positioned to benefit from EMIB. That said, my view is that the emergence of a new player like SEMCO should by no means be underestimated.

@x4tsr Yeah, well deserved victory lap that it’s now up hundreds of percent since European media kept shtting on my $SOI idea.

Oh look, $SOI is now up 4X+ since I went long. And another 15%+ after earnings. Despite media“analysts”claiming there was nothing new with Soitec. The fact they’re trying to reorganize the company to answer demand for SOI substrates for photonics. It’s a pretty positive read-through on what’s to come for the company in the next two years. Again with companies with forward growth opportunities, it’s about looking at the future… not past earnings, unlike what some Europeans might only care about.

超赞的文章! 看看大家怎么看待我的过往经历真的挺好玩的 最近我一直在努力多跟对供应链和AI感兴趣的小伙伴们互动 所以只要是想了解我思考逻辑的朋友，我都会尽量去关注

@159Adir52550 CPO is largely h2 2027. $JBL already announced timelines for mass production on their 1.6T LRO using $SIVE so that’s material revenue coming in within 3-10m.