供应链分析
产业链结构、上下游与瓶颈环节研究 · 共 1315 条 · 滚到底自动加载更多
-
AI数据中心供应链严重依赖亚洲，中东冲突加剧流动性风险，稀土应成美国国安优先事项
Hi Ron，这个话题分为两个不同的部分： 1. 超大规模云服务商（$ORCL、$META、$AMZN）：他们的支出超过了盈利水平（$GOOGL 是罕见的例外，而 $AAPL 相对而言支出并不多）。 因此市场担心这并没有转化为实质性的收入。 尤其是像 Oracle 这样负债累累的公司，降息是使其稍微可持续的重要因素（如果你看债务利息，数目非常庞大），而降息既能推动产能扩张，也能带来实质性的成本节省。 由于伊朗战争和原油价格上涨，原定2026年的降息预期已经落空。 雪上加霜的是，人们担心他们的 AI 数据中心建设供应链会中断。 如果你看看他们的 AI 数据中心部件产地： - 来自台湾 - 来自韩国 - 来自日本 这是一个全球化的供应链。 虽然美国看起来没问题，因为它在石油、液化天然气、氦气和其他半导体供应链所需的资源方面是绝缘的。 但我们的亚洲合作伙伴并不处于长期冲突中。 所以当你看到亚洲那些规模较小、专业化的公司时，它们正在苦苦挣扎。 这种影响会层层放大，一直传导到终端的亚马逊网络服务 AI 数据中心（成本上升、供应短缺）。 2. AI 整体而言：投资者对此非常乐观。 问题在于实现路径，而美国在表面上看起来被中国卡住了。 - 美国投入过多，而中国和其他国家正在蒸馏我们的最新模型（需要 KYC 端点管控），坐享其成。 - 随着 AI 项目的规模扩大，它正变得越来越依赖中国公司。 我认为这是因为美国缺乏稀土，这应该成为我们的头号国家安全优先事项。 这是制造新一代 AI 硬件、机器人供应链和太空探索所需的。 这些资源都由中国和俄罗斯控制。即使是领先的"西方"公司，我已发现它们仍然主要依赖中国/俄罗斯。 这是我们确保前沿项目安全最大的漏洞之一。 如果你看光子学（AI 硬件如何通过光加速），我们没有西方的供应链来满足需求。主要材料来源是中国的 $AXTI 和中国的 Vital 两家公司，然后流转到日本或其他公司，最后进入美国。 如果你看我们的人形机器人项目（我们如何将 AI 带入现实世界），$TSLA Optimus 的所有躯体都是在中国制造的。 美国供应链缺乏制造这些组件所需的稀土，因为成本太高。 3. 中东地区的冲突 近期下跌的一个重要原因是流动性问题。我们在阿联酋、中东的合作伙伴主要在为私募市场提供资金（想想 OpenAI），他们大量的支出流向 AI 市场。 他们也是从 Meta 到微软等科技七巨头最大的投资者之一。 如果他们的油田继续受到干扰，他们可能不得不从美国市场撤出流动性。 这既影响这些公司在 AI 上的支出能力，也会导致美国市场的抛售下跌。 如果有takeaway（要点）的话：请把稀土列为美国头号国家安全优先事项。特朗普总统已经投资了 $USAR、$MP 和几个单独的公司。 但还有许多更多至关重要的、尚未盈利的公司，它们对于制造 AI、自动驾驶汽车和机器人都是必需的。我们需要打破对中国/俄罗斯的依赖。
英文原文
Hi Ron, so there's two different parts to it: 1. Hyperscalers ( $ORCL, $META, $AMZN): They're spending more than they have with profit ( $GOOGL is the rare exception, and $AAPL isn't really spending much relatively). So markets are worried less worried it's translating into material revenue. Especially with companies like Oracle taking on immense amounts of debt, rate cuts were a large driving factor in why it's slightly more sustainable (if you look at debt interest, it's massive), and cutting rates drives both forward expansion and material amounts of savings. Due to the War in Iran, rising crude, former projections with cutting rates in 2026 are now gone. This is compounded that there are worries that their AI buildout supply chain gets disrupted. If you look at where their AI datacenter parts are made: - it comes from Taiwan - it comes from Korea - it comes from Japan It's global. While the US might look like it's fine, since it's insulated to Oil, LNG, helium, and others needed for the semiconductor supply chain. Our Asian partners are not in a long drawn out conflict. So when you look at companies in Asia that are more small and niche they are struggling. And this gets compounded tens of times until it reaches all the way down to the end Amazon Web Services AI datacenter (increased costs, lack of supply). 2. AI as a whole: Investors are very bullish on it. It's just how we get there, and America looks strangled by China optically. - America is spending way too much on it, with China and others distilling our latest models (need KYC endpoints), and reaping all the benefits. - As we scale our AI programs, it's becoming increasingly reliant on Chinese companies. I'd argue it's because America lacks the rare earths, which should be our #1 national security priority. This is needed to make new generations of AI hardware, robotic supply chains, and Space. They're all controlled by China and Russia. Even the leading "Western" companies, I've identified they're mainly relying on China/Russia still. And this is one of the biggest vulnerabilities in securing our frontier programs. If you look at photonics (how AI hardware is sped up by light), we no Western supply chain to fulfill our needs. It's $AXTI (in China), and Vital (in China) as the two main sources of materials needed to make them, that gets passed down to Japan, or other companies before they end up in US. If you look at our humanoid program (how we move AI to the real world), all the bodies of $TSLA Optimus are made in China. American supply chains lack the rare earths needed to make the components because it gets too expensive. 3. Disruption in the Middle East A large part of the recent drop is mainly due to liquidity. Our partners in the UAE, Middle East are largely funding private markets (think OpenAI) and a lot of their spend going to AI markets. They're also one of the biggest investors in Mag7 from Meta to Microsoft. If their oil fields continue getting disrupted they may have to pull out liquidity from US markets. This impacts both the amount of money these companies can spend on AI. As well as causing a drop in American markets from selling. If there's any takeway: Please make Rare Earths Amercia's #1 national security priority. President Trump already invested in $USAR, $MP, and a few individual names. But there are many more extremely critical, yet unprofitable companies that are needed to make AI, to self driving cars, to robotics. And we need to break our reliance on China / Russia.
-
希望 SIVE 保持独立，担心被竞争对手低价买下。
@Matrix_B0SS 我其实希望 $SIVE 保持独立。 只是随手写了些洗澡时冒出来的想法，觉得另一家超大客户或竞争对手可能会用区区 3.5 亿美元就控制 $MRVL 的近端光子项目。
英文原文
@Matrix_B0SS I actually hope $SIVE remains independent. Just posting some shower thoughts about how another hyperscaler or competitor could control $MRVL near term photonics program for a cheap price of $350m.
-
认为 SIVE 可能被多方盯上，也可能因持股结构保持独立。
洗澡时冒出来的一个想法…… SIVE 现在可能已经被 $QCOM、$MRVL、$AVGO、AlChip、联发科、$AMZN、$META、$MSFT 的风投部门盯上了？ 既然现在大家都公开知道了，Broadcom 只要花 3.5 亿美元就能买下 Marvell 在光子学上的咽喉点（Amazon / Microsoft 的 ASIC 项目）…… 或者 Qualcomm 也可以用同样的方式收购 $SIVE，然后在收购 Alphawave 之后把激光 IP 做垂直整合。 又或者……超大规模客户会通过持有足够多股份，想办法让 $SIVE 继续保持独立？
英文原文
It’s funny a shower thought… Probably put $SIVE on the radar for $QCOM, $MRVL, $AVGO, AlChip, Mediatek, $AMZN, $META, $MSFT venture arms now? Since everyone publicly knows now Broadcom can just go and buy Marvell’s photonics chokepoint vulnerability (Amazon/Microsoft ASIC programs) for a rounding error of $350m… Or Qualcomm can do the same buying $SIVE then vertically integrate laser IP after Alphawave aquisition. Or maybe… hyperscalers figure out a way $SIVE remains independent by owning enough shares?
-
把 7n 高纯铟视作 InP 衬底的一个代理信号。
@ParadisLabs 7n 高纯未标准铟，在 SMM 上是 InP 衬底的一个不错代理指标。 不过对 CPO 放量来说，这件事复杂得多，我会看更多其他信号。
英文原文
@ParadisLabs 7n indium high purity nonstandard on SMM is a good proxy derivative for InP substrates. However, it’s much more multifaceted for CPO ramp and I’d look for other signals
-
只是分享想法，猜测 Marvell 的 CPO 供应链会不会被竞争对手收购卡脖子。
@Jornka329996 我只是分享一下洗澡时的想法，不知道人们或者市场会怎么反应。 也许 $MRVL 的风投方更愿意让他们的 CPO 供应链卡口被 $AVGO 或 AlChip 这样的竞争对手收购和控制？
英文原文
@Jornka329996 I’m just sharing my shower thoughts, don’t know how anyone or markets react. Maybe $MRVL ventures prefers their CPO supply chain chokepoint to be bought out and controlled by competitors like $AVGO or Alchip?
-
把近期机构轮动理解为 CPO/ELS 上游链条的机会。
尽管指数和巨头都在跌，比如 $META / $MSFT，但市场里依然明显有一些超额收益者。 这大概说明了机构轮动，或者说某个非常强的新兴行业已经在形成。 最明显的就是 CPO / ELS 供应链。 我个人的标的包括： $TSEM / Win Semi - 光子学里的 TSM $AAOI / $SIVE / $COHR / $LITE - 光源 $SOI / $AXTI - 衬底 $IQE - 外延片 当然也有更大的名字，比如 $MRVL、Sumitomo、$AVGO。
英文原文
There’s clearly some market outperformers despite index’s and mega caps from $META / $MSFT crashing. This probably signals institutions rotation or an extremely strong up and coming sector. The obvious one is into CPO/ELS supply chains. My personal picks were: $TSEM / Win Semi- TSM of photonics $AAOI / $SIVE / $COHR / $LITE- Light Sources $SOI / $AXTI - Substrates $IQE - Epiwafers But of course there are larger names out there like $MRVL, Sumitomo, $AVGO.
-
指出美国在InP衬底、人形机器人、高端半导体等领域严重依赖外国供应商，无法自给自足。
这在事实上并不正确，而这正是我们陷入困境的完全相同的原因。 美国完全依赖日本/中国。例如：用于AI光子学的InP（磷化铟）原料/衬底。 美国依赖中国制造人形机器人的机体。 美国通过一些关键瓶颈点依赖俄罗斯/其他对手（不会透露具体名称，因为涉及国家安全风险）。 美国依赖台湾/韩国制造高端半导体。 美国依赖欧洲制造高端机器（$AXIA、$AMSL类型）或特定衬底（$SOI）。 再次强调，供应链是相互关联的。这种观点正是当前政府认为自己掌握了所有牌的精确原因。 请列举一家能够供应AI所需InP衬底足够原材料的美国公司。 请列举一家今天能够完全替代$ASML的美国公司。 你做不到，因为它们还不存在。
英文原文
This is factually not true and this is the exact same reasoning why we're in this mess. US is completely reliant on Japan/China. eg. InP feedstock/substrates for AI with photonics. US is reliant on China for manufacturing bodies of humanoids. US is reliant on Russia/other enemies (won't disclose specific names since it's national security risk) through some niche chokepoints. US is reliant on Taiwan/Korea for high end semiconductors. US is reliant on Europe for high-end machines ( $AXIA, $AMSL type) or specific substrates ( $SOI ). Again, supply chain is interconnected. This type of take is the exact reason why the current administration thought it had all the cards. Name one US company that can supply enough raw feedstock required for InP substrates for AI. Name one US company that can completely replace $ASML today. You can't because they don't exist yet.
-
全球供应链高度互联，战争将摧毁从AI到消费电子的产业链，受益者仅是以色列和俄罗斯。
我仍然看多市场（目前85%多头，15%现金/对冲）。即使在战争时期，我预期从$AAOI到$LITE的许多股票仍能跑赢大盘。然而...如今世界高度互联。现任政府可能没有意识到上游供应链的严重性和脆弱性。他们可能看到$AAPL的"美国制造"。或者$TSLA的"美国制造"，但如果你看看特斯拉人形机器人的供应链来源：-> 在中国。执行器(Actuator)材料来自俄罗斯的原料。材料来自加拿大的合作伙伴。-> 对于星链(Starlink)，大量供应链来自台湾稳懋(Win Semi)到韩国的公司。对于OpenAI、私募股权、大型科技公司和其他公司的资金来源：-> 很大一部分来自阿联酋、中东和其他地方（随着油田被炸毁，他们可能需要撤出流动性）。持续数月的美国地面入侵造成的混乱是是非常看跌的。我最初预期的是"震撼与敬畏"，美国宣称胜利，战争冲击被买回。但我不会鼓励美国走向一场旷日持久的战争。这对亚洲伙伴（韩国、日本）或我们在欧洲的盟友都没有好处。战争的主要受益者是：#1：以色列 #2：俄罗斯世界其他地区，包括美国，都被战争拖累。看起来"油价/能源价格上涨对美国有利"：但它摧毁了从$NVDA到$AAPL的超大规模数据中心(Hyperscaler)所依赖的大量供应链。
英文原文
I'm still long on markets (now 85% long, now 15% cash/hedges). Even during wartime, I expect many names from $AAOI to $LITE to outperform. However... The World is largely interconnected now. The current administration might not realize the severity/fragility of upstream supply chains. They might look at $AAPL "Made in America". or $TSLA "Made in America" but if you look at the source for Tesla Humanoids: -> It's in China. Actuator materials stem from Russia feedstock. Materials come from partners in Canada. -> For Starlink, massive amount of supply chains stem from Win Semi Taiwan to companies in Korea. For where the funding from OpenAI, private equities, megacap, and others come from: -> Large part of it comes from UAE, Middle East, and others (that might need to pull out liquidity) as their oil fields get blown up. Extended disruptions (US ground invasion) for spanning multiple months, is very bearish. I was initially expecting the "Shock and Awe" and US to claim victory and for short term War shock to be bought back. But I would not encourage US to head toward a drawn out War. This especially is not good for Asian partners (SK, Japan) or our allies in Europe either. The main beneficiaries of War are: #1: Israel #2: Russia The rest of the World, including the US, are getting dragged down from War. It looks like "Oil/Energy prices going up is good for US": But it takes down a lot of supply chains that the hyperscalers from $NVDA to even $AAPL relies on.
-
对比光通信供应链各玩家对CPO光源业务的直接敞口差异
当然，这更像是直接敞口与市值的对比。 光源环境如下： 头部玩家：Lumentum (550亿美元)、Broadcom (1万亿美元)、Coherent (550亿美元)、Sumitomo (规模庞大) 大型玩家：$AAOI (约80亿美元)、Furukawa (约70亿美元)、$MTSI (约190亿美元) 新兴玩家：$INTC (约2300亿美元+)、$SIVE (2.3亿美元)、Luxnet (17亿美元) 你可以买$AVGO来获得光子学敞口，但因为它规模太大……他们的其他业务部门可能会稀释敞口。 另一方面，如果你看$SIVE，由于规模很小，它几乎是CPO光源升级的纯概念股。 所以我最终选择了最直接的敞口。 很多人对市值(MC)感到不适，所以Furukawa、$MTSI、Sumitomo这类公司可能更受机构分析师青睐。
英文原文
Sure, it's more like direct exposure vs. marketcap. The light source environment looks like this: Leading players: Lumentum ($55B), Broadcom ($1T), Coherent ($55B), Sumitomo (massive) Large Players: $AAOI (~$8B), Furukawa (~$7B), $MTSI (~$19B) Emerging players: $INTC (~$230B+), $SIVE ($230M), Luxnet ($1.7B) You can buy $AVGO for photonics exposure, but because it's so large... their other segments will probably be diluting exposure. On the other hand, if you look at $SIVE, it's almost pure-play to the light source ramp for CPO because it's so small. So I ended up picking the most direct exposure. Lot of people are uncomfortable with MC, so companies like Furukawa, $MTSI, Sumitomo might be more popular with institutional analysts.
-
把 AEHR、TSEM、AAOI 分别对应测试、代工和激光供应链。
@mi20483980476 所以这三者其实分属三个不同类别（当然有重叠），我都喜欢。 $AEHR 做测试。 $TSEM 更像是晶圆代工。 $AAOI 则更像是激光 fab -> 设计 -> 组装。 这就是一条供应链。
英文原文
@mi20483980476 So they all sit in three separate categories (with some overlap) and I like all three. $AEHR does testing. $TSEM is more like a foundry then $AAOI is more like laser fab -> design -> assembly It's a supply chain.
-
博主分享其对AI供应链瓶颈轮动的精准判断，从存储到光子学再到CPO/ELS，并引用研报分析AI投资超级周期的瓶颈解决序列和CPO商业化拐点
我认为我精准把握了机构层面的瓶颈轮动。 -> 在存储股的上涨尾声捕捉到了 $SNDK、三星、SK海力士、$MU -> 在光子学领域领先机构布局了 $AAOI、$AXTI、$LITE、$COHR -> 现在再次行动，大量加仓SiPh、ELS和CPO：$SIVE、$TSEM、$SOI、$AEHR、Win Semi以及其他 当然如果你想稳健操作： $MRVL（产能自用）、$AVGO（产能自用）、$TSM和$NOK也都有涉及，但他们是更大的玩家。直接曝险下一个超级周期才是理想的。 我仍然认为从存储到EML光收发器的增长空间巨大，但最大的繁荣出现在新架构周期起点/拐点处。 除了宏观因素干扰了一些交易外，预计资本将很快轮动到CPO/ELS供应链。 券商研报的摘要：AI基础设施投资超级周期遵循严格的"瓶颈解决"序列： 算力/GPU（2023）-> 存储/HBM（2024）-> 互联/网络（2025+） 翻译： "我们判断AI价值链的第三轮投资周期已正式启动。继GPU（2023）和HBM（2024）之后，2025年后，光互联将成为增长最快的核心环节。 2027-2028年将是关键拐点，CPO商业化、1.6T标准化和Scale-Up光互联将形成共振，结构性扩大相关行业TAM。" 一些要点： 1. CPO（共封装光学）正从实验室走向商业化量产 2. 2027-2028年是重大结构性拐点（现在在2026年提前布局是个好主意，从测试端$AEHR到ELS端$SIVE或封装端$POET） 3. 光学部件、材料和测试设备的市场总规模（TAM）预计将结构性扩大3到5倍（我认为这还是保守估计）。
英文原文
I think I nailed the institutional bottleneck rotation. -> Caught the tail end of memory name rise with $SNDK, Samsung, SK Hynix, $MU -> Frontran institutions with photonics with names like $AAOI, $AXTI $LITE, $COHR. -> Doing it again now by adding heavily toward SiPh, ELS, and CPO: $SIVE, $TSEM, $SOI, $AEHR, Win Semi, and others. Of course if you want to play it safe: $MRVL (captive), $AVGO (captive), $TSM, and $NOK all do it as well, but they're larger players. And getting direct exposure to the next supercycle is ideal. I still think there's tons of room to grow for memory to EML optical transceivers, but the largest boom is at the start/inflection point of a new architectural cycle. Macro messing up some trades aside, expecting capital rotation soon toward CPO / ELS supply chains. TLDR from analyst note: The AI infrastructure investment supercycle follows a strict "bottleneck resolution" sequence: Compute/GPUs (2023) -> Memory/HBM (2024) -> Interconnect/Networking (2025+). Translation: "We judge that the third investment cycle of the AI value chain has officially begun. Following GPUs (2023) and HBM (2024), post-2025, optical interconnects will become the fastest-growing core segment. 2027–2028 will be a critical inflection point where CPO commercialization. 1.6T standardization, and Scale-Up optical transitions align, structurally expanding the TAM of related industries" Some takeaways: 1. CPO (Co-Packaged Optics) is moving from the lab to commercial mass production 2. 2027-2028 is the major structural inflection point (good idea to frontrun this now in 2026 from testing with $AEHR to ELS with $SIVE or packaging with $POET). 3. The Total Addressable Market (TAM) for optical components, materials, and testing equipment is expected to structurally expand 3x to 5x (I think this is sandbagging a bit).
-
Mirae分析师梳理CPO产业格局，作者认为遗漏上游公司但框架有参考价值
CPO产业格局 Mirae分析师笔记： 横向扩展： CPO ASIC：$AVGO、$MRVL 光收发器：$COHR、$LITE、Innolight DSP/PAM4：$AVGO、$MRVL 相干DCI：$CIEN、$NOK OCS设备：iPronics、Polatis 光缆/光纤：$GLW、Prysmian、Furukawa HCF：$LITE、OFS DCI相干：Ciena、Nokia、Huawei 光放大器：$LITE、$COHR OCS网关：KDDI 纵向扩展： SiPh代工：imec、$GFS、$TSM SiPh调制器：$NVDA（内部MRM）、$INTC ELS：NTT、Furukawa、$LITE、$COHR 太赫兹互联：研发阶段？ CPO测试：「新玩家不断进入」 微透镜/光学系统：「新玩家不断进入」 总结：纵向扩展的CPO即将到来。认为分析师笔记遗漏了一些上游公司名字，并将ELS与光源混为一谈。 但作为高层视角，能看到他们认为的领先玩家还是有参考价值的。
英文原文
CPO Landscape Mirae Analyst Note: Scale-Across: CPO ASIC: $AVGO, $MRVL Optical Transceiver: $COHR, $LITE, Innolight DSP/PAM4: $AVGO, $MRVL Coherent DCI: $CIEN, $NOK OCS Equipment: iPronics, Polatis Optical Cable / Fiber: $GLW, Prysmian, Furukawa HCF: $LITE, OFS DCI Coherent: Ciena, Nokia, Huawei Optical Amplifier: $LITE, $COHR OCS Gateway: KDDI Scale-Up: SiPh Foundry: imec, $GFS, $TSM SiPh Modulator: $NVDA (in-house MRM), $INTC ELS: NTT, Furukawa, $LITE, $COHR THz Interconnect: R&D Stage? CPO Test: "Expanding entry of new players" Micro Lens / Optical Systems: "Expanding entry of new players" TLDR: Scale Up CPO is coming next. Think the analyst note missed a bunch of upstream names and conflated ELS with light source. But it's helpful to see who they think the leading players are as a very high-level view.
-
看好HIMS与TEM结合布局个性化AI医疗，认为市场潜力巨大
个性化AI医疗很可能是一个巨大的市场。没有人想早死。将 $HIMS 与 $TEM 结合具有巨大潜力。这里有一个GitLab CEO的例子：2023年，Sijbrandij被诊断出骨肉瘤；2024年7月，他宣布癌症复发了，尽管进行了多种治疗；到2024年底，他已经用尽了标准医疗方案的全部选择。没有选择接受死亡，Sijbrandij决定像构建GitLab一样将他的癌症作为一个工程问题来处理。他开始向ChatGPT输入大量的个人医疗数据，包括扫描结果、血液检测结果和组织样本数据。ChatGPT和AI在识别突变后提出了新的靶向治疗方案，然后确定如何攻击这些突变。在此基础上，他申请了未获批准、处于试验阶段的药物并获得了用药。这个方法奏效了。Sijbrandij用AI阻止并扭转了晚期预判。我们已经看到了这个成功案例（ $TEM 明确列为他用于肿瘤批量RNA测序的合作伙伴）。最近也有一个人用AI治愈了自己的狗。如果你能够将 $HIMS（直销医疗分销网络）和 $TEM（个性化测序和AI治疗）结合成一个可扩展的产品，潜力是巨大的（尽管极具挑战性）。全球人口都是潜在市场。这类似于 $AXTI InP（磷化铟）在超大规模数据中心的情况：人们会不惜代价避免死亡（这就是瓶颈）。这只是未来医疗发展方向的一个缩影。
英文原文
Personalized AI medicine will likely be huge. Nobody wants to die early. Combining $HIMS to $TEM has immense potential. Here's just one example with Gitlab's CEO: > In 2023, Sijbrandij was diagnosed with osteosarcoma > In July 2024, he announced the cancer had returned, despite multiple treatments > By late 2024, he had reached the absolute end of standard of care medical options. > Rather than accepting death, Sijbrandij decided to treat his cancer the same way he built GitLab as an engineering problem > He began feeding ChatGPT massive amounts of his personal medical data from scans, blood test results, and tissue sample data. > ChatGPT and AI made novel targeted treatments after identifying mutations, then how to attack them > From there, he applied for unapproved, experimental drug and received it > The approach worked. Sijbrandij managed to stall and turn around a terminal prognosis with AI. We've seen this success story with Gitlab's CEO ( $TEM was explicitly listed the partner he used for the bulk RNA sequencing of his tumor ). As well as someone curing their own dog with AI recently. The potential if you can combine $HIMS (DTC distribution network) and $TEM (Personalized sequencing and AI treatment) into a scalable product is massive (though extremely challenging). The entire human population is the TAM. And similar to the $AXTI InP situation with hyperscalers: People will pay anything to not die (bottleneck). This is just a hint into where the future of healthcare is heading.
-
说很多业内高管在关注 SIVE 这样的瓶颈点，接下来几个月值得看。
@TheValueist 是啊，自从我发了最初那篇帖子后，不少行业高管都开始对 $SIVE 感兴趣。 接下来几个月看看这些瓶颈点会怎么演化，会很有意思。
英文原文
@TheValueist Yep, lot of industry executives got interested in $SIVE after I made the original post. Should be interesting to see how these chokepoints plan out over the next few months
-
认为上游光源被控住后，MRVL 的光子路线图会被低成本卡住。
大致就是 $SIVE -> Win -> $POET -> Celestial -> $MRVL 这条链。 你只要把上游光源买下来，就能用很低的成本控制 Marvell 的供应链。 他们当然可以绕开，但认证时间和延迟会拖慢路线图。 或者 $MRVL 也可以入股，确保 $SIVE 保持独立。
英文原文
It goes $SIVE -> Win -> $POET -> Celestial -> $MRVL basically. You just buy out the upstream light source and Marvell’s supply chain gets controlled for dirt cheap. They can engineer around it but qualification time and delays roadmap. Or $MRVL can secure a stake to make sure $SIVE remains independent
-
承认现金少和历史包袱，但认为独立 cw 激光供应商的价值足以抵消下行。
当然，现金余额低、来自母公司的历史包袱、后续还会有进一步稀释。 Ayar 在做多来源采购，而且更偏向 $LITE。可能还会遇到 Win 的认证问题，以及 CPO 路线图延迟。 但我认为，作为全球少数几家独立 cw 激光供应商之一，它的价值足以抵消这些下行风险。这个身份本身就非常有价值。 如果跌得足够便宜，对冲基金或者像 Broadcom、Marvell 这样的公司，完全可以用很低的价钱把它收购，然后向上游一体化，或者把竞争对手的关键路径卡住。
英文原文
Sure, low cash balance, legacy drag from parent company, further dilution along the way. Ayar multi sourcing and leaning more heavily toward $LITE. Maybe qualification issues with Win. CPO roadmap delays. I do think any downside risk is negated by how valuable being one of the only independent cw laser suppliers in the world. So that itself is really valuable. If it ever drops enough a hedge fund or company like Broadcom or Marvell can just buy it out for pennies and either vertically integrate upstream or chokepoint out competitors.
-
认为 Broadcom 可以通过收购光子子公司来切入军工和 CHIPS 相关供应链。
@Ud197601 他们的母公司因为是 CHIPS 法案受益者，而且还和军工承包商合作。 光子学子公司也许会成为可争夺的目标。 而且 $AVGO 也不是什么陌生公司，所以他们完全做得到。
英文原文
@Ud197601 Their parent company bc it’s CHIPS act recipient and partners with military contractors. The photonics subsidiary might be fair game. And $AVGO isn’t an unknown company so they can pull it off
-
调侃如果竞争对手控制了 MRVL 的 CPO 路线图会更好笑。
@StocksAREnuts 如果有竞争对手以 2.8 亿美元拿下并控制 $MRVL 的 CPO 近端路线图，那会更好笑。
英文原文
@StocksAREnuts Would be even funnier if a competitor controlled or delayed $MRVL CPO near term roadmap for $280m.
-
说更好的做法是便宜地拿下竞争对手整套技术栈并纵向整合。
@Hiraweb3 或者更好……用很便宜的价格直接拿下竞争对手的整套技术栈。 这样还能更好地做纵向整合。
英文原文
@Hiraweb3 Or better yet… own their competitor’s stack for dirt cheap prices. And they can vertically integrate better too.
-
认为 SIVE 太上游，不太会卷进反垄断，Broadcom 可以轻松控制 MRVL 路线图。
@InFoTheLongTerm $SIVE 作为激光组件供应商，位置太上游了，所以不会被反垄断问题缠住，这样 Broadcom 就能把这件事做成。 有意思的是，这里其实就是一个瓶颈点，你等于可以在中短期内掌控 $MRVL 的光子学路线图，因为新供应商的认证需要时间。
英文原文
@InFoTheLongTerm $SIVE is too upstream as a laser component vendor to be caught up in antitrust so @Broadcom can pull it off. But the funny thing is it’s a chokepoint and you’d effectively own $MRVL photonics roadmap near-medium term since qualification of new suppliers takes time.
-
设想 Broadcom 收购上游光源后控制 Marvell 短中期光子路线图。
只是一个洗澡时突然冒出来的想法： 像 $AVGO 这样的竞争对手，如果用 2.8 亿美元收购 $MRVL 的上游光源 $SIVE，那会很聪明。 这样你就能控制他们短中期的光子学路线图，连 Celestial 也在你手里。 有点像 $QLCM 当年把 Alphawave 和 OpenLight 生态圈拿下。 已经有不少对冲基金在谈潜在并购机会了。 问了 Gemini，回答还挺好笑。
英文原文
Just a shower thought: Would be smart for a competitor like $AVGO to buy $MRVL upstream light source for $280M with $SIVE. Then you control their short-mid term photonics roadmap with Celestial. Kind of like $QLCM did with Alphawave and OpenLight ecosystem. Quite a few hedge funds already talked about potential M&A opportunities. Asked Gemini, response was funny
-
分析光学组件为AI集群关键瓶颈，Sivers垂直整合价值凸显
因为我预期一些公司，尤其是光子学个股会显著跑赢大盘。Sivers还拥有关键的激光技术，$MRVL和其他公司会想要收购来实现垂直整合……如果股价进一步下跌，任何人都可以用零钱收购整家公司。当我说伊朗扰乱全球流动性和能源时，这并不意味着在最坏情况下$NVDA会跌到零。也许一些超级计算用户会保守地削减资本支出，一些数据中心建设会放缓。但如果光学组件是扩展AI集群的必要部件，那么它们已经售罄，而且在BOM中占比很小因为它们很便宜。削减支出根本无法改变这个瓶颈。
英文原文
Bc I expect some companies, especially individual photonics names strongly outperform. Sivers also hold the critical laser tech that $MRVL and others would want to buy for vertical integration… and if it dropped any further, someone can just acquire the entire company for spare change. When I say Iran disrupting global liquidity/energy, that doesn’t mean $NVDA drops to 0 in a worse case scenario. Maybe some hyperscalers cut back on capex spend to be conservative, some DC buildout stalls. If optical components are necessary to scale AI clusters. Then they’re sold out and such a little part of BOM since they’re cheap. Cutting back on spend won’t change that bottleneck at all
-
认为用有限 TAM 思路来理解 SIVE 这种瓶颈点是错的，应该看扩张/博弈。
很高兴这有帮助。 我说 $SIVE 让我想起 $LITE，大概就是我脑子里在想的这些东西。 我也看到有人试图把像 $AXTI 这种东西建模成有限或稳定复利型 TAM，而不是把它看成扩张或瓶颈博弈——但那是看错了这些细分供应链卡点的方式。
英文原文
Glad it’s helpful. When I say $SIVE reminds me of $LITE, this is kinda what I’m thinking about in the back of my mind. I also saw people try to model finite or steady compounding TAM (like $AXTI) rather than expansion or bottleneck game theory, which is the wrong approach for these niche supply chain chokepoints
-
认为 LITE 已经用过类似打法，管理层路线大体正确，只是时点早了。
我们刚刚看到 $LITE 用同样的打法跑出了 3600% 以上的涨幅…… 你不需要重新发明轮子。 我其实认为管理层是有能力的，他们已经在手头资源条件下走对了路（比如非稀释性补助、Win 的半无晶圆厂模式等等）。 之后他们在 2024 年 2 月到 2025 年尝试做光子业务分拆，但没成功，因为那时候还没到光子学超级周期。 但我认为最近这波关注，可能已经帮助他们摆脱了瑞典小盘股的死亡谷，而且现在有不少对冲基金在研究他们。 所以我觉得，这和一开始投资 $LITE 的思路是相通的。
英文原文
We’ve just seen $LITE pull off this exact same playbook and go up 3600%+... You don’t need to reinvent the wheel. I actually think management is competent and they’ve gone down the right path with the resources they had (eg. Nondilutive grants, Win semi fabless, etc.) Then they attempted photonics spinoff in 2024-Feb 2025, but didn’t work because that was before photonics had a super cycle. But I think the recent attention may have helped them escape the valley of doom in small cap Sweden, lot of hedge funds are researching them now. So I think it’s with the shot of investing in $LITE at the start.
-
提出让 SIVE 沿光子链条扩张 TAM、借资本注入和拆分推进 ELS 模块规模化。
难道 $SIVE 就不能学 $LITE 那样，把 TAM 往 ELS 整个堆栈一路往下扩吗？ 我已经开始想起早期的 $LITE 了（Oclaro、NeoPhotonics、Cloud Light）…… 下面是我的论点/设想： Sivers 如何把大约 300-400 美元的 ELS（来自 50-100 美元的阵列）吃下来，并成长为一家 100 亿美元以上的公司。 路径大概是： -> Sivers 需要融资。 -> 拿 NRE 预付款或者 PIPE（比如 $AXTI/Northland） -> 直接告诉市场：“$AVGO 和 $LITE 会把你们定制 AI 集群的利润率卡住。” ----> 让 $META / $AMZN / $MSFT 投资注资，来支持未来 ASIC 集群的多来源供应，因为他们会害怕。 尤其是 $NVDA 已经出资 40 亿美元来保障从 $LITE 和 $COHR 的供应……再加上当前 EML 产能瓶颈带来的 PTSD。 其实比大家想的容易。 -> 把光子学业务分拆到美国（市场不喜欢给老牌公司投资）。 - - - - - > 这件事大概率会在 2026 年奏效，因为 CPO / photonics 现在是热潮，而不再是 2024 年到 2025 年初那样。 -> 筹到足够的钱后，自动化晶圆探针机，并给模块 NPI 提供资金。 -> 采用轻资本 fab-lite 测试模型：自建探针卡并内部测试，但把 capex 交给 OSAT。 -> 收购硅光封装 IP（或者继续和 O-Net、$POET 做 JDM）。 -> 保留 Glasgow 的 InP 晶圆厂，用于内部测试、迭代和原始外延。 -> 把光刻、刻蚀等环节外包给 Win Semi 和其他厂商（别自己做重资本环节）。 -> 然后 $SIVE 只要把经过测试的激光器和封装 IP 交给 $FN。 然后就成了。 你拥有了一个独立、可大规模生产、高毛利的 ELS 模块，而 ELSFP 也会成为标准。 之后他们还能做别的事，比如购买 DSP 设计并开始整合技术栈。 不再是每个 10 美元的激光芯片乘 6，做成 50 到 100 美元的激光阵列，而是变成 400 美元的可热插拔模块。 基本上，他们的起点是掌握光子学里激光层面的一个瓶颈护城河…… 然后他们就能像 $LITE 当年那样，往下游做 EML 和光模块。 TLDR： Sivers（$SIVE）在世界正向 CPO 转型的这个时点，持有一个稀缺、独立的 InP 激光瓶颈点。 全球只有少数几家独立公司能做这件事，比如 $SIVE（例如住友、$LITE、$COHR）。 但当然，他们需要资金、需要正确的愿景，才能实现 TAM 扩张。
英文原文
Can't $SIVE just pull a $LITE and TAM expansion down the ELS stack? I'm getting early stage $LITE flashbacks (Oclaro, NeoPhotonics, Cloud Light)... Here's my thesis/idea: on how Sivers can capture the ~$300-$400 ELS (from $50-100 arrays) and become a $10B+ company. How this plays out: -> Sivers is needs funding. -> Get NRE prepay or PIPE (eg. $AXTI/Northland) -> Say "Hey, $AVGO and $LITE are going to choke your custom AI cluster margins". ----> get $META / $AMZN / $MSFT capital injection for multi-source supply for next ASIC clusters cause they're scared. Esp. with $NVDA funding $4B to securing supply from $LITE and $COHR... then PTSD from the current EML capacity bottleneck. It's easier than people think. -> US spinoff of photonics arm (markets don't like investing in legacy companies). - - - - - > This will likely work in 2026, bc CPO/photonics is hot, now and not like 2024-early 2025. -> Raise enough to automate wafer probers and fund the module NPI. -> Fab-lite test model: develop probe cards and test inhouse, but push capex to OSATs. -> Buy SiPh packaging IP (or just continue JDM with O-Net and $POET) -> Keep their Glasgow InP fab for in-house testing for iteration/raw epitaxial. -> Outsource lithography, etching, etc. stuff to Win Semi + others (don't do capex heavy things yourself) -> Then just $SIVE can just hand tested lasers and packaging IP to $FN And Poof. You have a independent, mass producible, high margin, ELS module as ELSFP becomes standard. Then they can do other stuff like buying DSP designs and start consolidating the tech stack. Instead of cheap $10 laser dies x 6 for a $50-$100 laser array, you have $400 hot swappable modules. Basically their starting point is owning one of the chokepoints moat in photonics on the laser level... Then they can go downstream like what $LITE did with EML and optical transceivers. TLDR: Sivers ( $SIVE ) holds a rare, independent chokepoint in InP lasers at the exact moment the world is transitioning to CPO. There's only a few independent companies in the world that can do this like $SIVE (eg. Sumitomo, $LITE, $COHR). But of course they need the funding to achieve TAM expansion and the right vision.
-
伊朗无人机袭击阿联酋铝厂后，提出铝产业链和铝期货的交易思路。
伊朗刚刚无人机袭击了阿联酋的一家大型铝厂。 如果你和所有人一样第一反应都是：我怎么从这件事赚钱？ 思路如下： $CNEX - 铝生产中的冶炼环节。全球原铝价格上涨时，它们会受益。 $AA - 铝土矿开采、氧化铝精炼和铝冶炼。他们是从矿山 -> 精炼厂 -> 冶炼厂的一体化纯标的。 当然还有：做多 CME/COMEX 铝（ALI）。 大多数人都在做多石油，但如果伊朗用无人机袭击把一切都打乱了，铝实际上也是战争的一个大受益者。
英文原文
Iran has now just drone striked the major UAE Aluminum Plant. If your first thought like everyone else was: How do I profit off this? Here's how: $CNEX - Smelting process for aluminum production. They benefit when primary aluminum prices hike from global prices. $AA - Bauxite mining, alumina refining, and aluminum smelting. They capture mine -> refinery -> smelter as a pure play way for Aluminum. And of course: Long CME/COMEX Aluminum (ALI). Most people are long oil, but Aluminum from Iran drone striking everything, is a large beneficary of the war.
-
认为超大规模云厂商需要的是当下的电力与土地，neocloud 有短期护城河窗口。
@futureproof_inv 超大规模云厂商需要的是今天的电力……不是五年后的电力。 这就是为什么像 $NBIS 或 $IREN 这样的 neocloud 正好卡在一个甜蜜点：它们有一小段时间去建立可防御的护城河，并成为下一个 AWS。 短期内的电力/土地本身并不是可防御的护城河。
英文原文
@futureproof_inv Hyperscalers need power today... not in 5 years time. That's why neoclouds like $NBIS or $IREN hit that sweet spot where they have a short time period to grow a defensible moat and become the next AWS. Short term power/land is not a defensible moat.
-
仍然极度看多光子学，只是对存储略偏多，并把 doomsday portfolio 当对冲。
也不是完全这样，我对光子学还是超级看多，只是现在对存储只算略微看多。 我确实把一些标的的仓位砍掉了一点，好把杠杆降下来。 如果未来几年光子学已经卖断货，那么超大规模云厂商哪怕削减 CapEx，它依然还是卖断货，因为它在 BOM 里的占比本来就很低。 但存储在 BOM 里占比就大得多。 末日组合就是一种对冲。
英文原文
Not exactly, still mega bullish on photonics, only slightly bullish on memory now. I did cut some concentration in some names to go off margin. If you’re sold out for next few years for photonics, hyperscalers possibly cutting capex still makes it sold out since it’s still a low part of BOM. Memory is a huge amount of BOM though. Doomsday portfolio is a hedge.
-
认为 SOI 估值便宜且在衬底层有垄断地位，重估会慢慢发生。
@Ud197601 @AdityaInvests90 $SOI 的市净率大概只有 1.3 倍，所以估值已经很低了。 我觉得它会缓慢重估向上，但它在衬底层确实拥有相当明显的垄断地位。 当然，像 $TSEM 那种一周涨 70% 的走势还是把我吓了一跳。
英文原文
@Ud197601 @AdityaInvests90 $SOI trades close to 1.3x book value. So it’s already at depressed valuations. I think it’s a slow re-rating up but it holds a material monopoly on the substrate level. Of course like $TSEM that 70% 1 week rally caught me off guard.
-
提醒别人忽略了股市已经在跌，如果指数是红的这些名字会更高。
@hero688 大家都忘了，哪怕指数已经在跌，这些名字还是这样。 如果指数是红的，这些票会高得多。 https://t.co/lsV1cgfBfs
英文原文
@hero688 People are forgetting that this is despite indexes crashing already. If they were green, these names would be a lot higher. https://t.co/lsV1cgfBfs