供应链分析
产业链结构、上下游与瓶颈环节研究 · 共 1309 条 · 滚到底自动加载更多
-
再排查 NVDA 架构变化下的台湾标的
@Leo66Gao 对，TWSE: 3324、TWSE: 3017、TWSE: 8996、TWSE: 2308、$VRT 这些都是我标出来的。 但我今天还想多做些尽调，以防 $NVDA 的架构变化会把其中一两个名字设计出去。
英文原文
@Leo66Gao Yeah TWSE: 3324, TWSE: 3017, TWSE: 8996, TWSE: 2308, $VRT, and others were the ones I flagged. But I wanted to do more DD today, just in case the $NVDA architecture changes might have designed or two out.
-
总结近期半导体与 AI 供应链的多项变化
这是我对最近半导体发展的一个 TLDR： 1. $TSM 在强推 CoPoS - VisEra / 其他公司可能会比预期更早起飞。 2. $AAPL 选择 $INTC 做半导体生产，这是个重大变化，因为他们通常会选 TSM。美国制造继续起飞，Intel 继续起飞。 3. $NVDA Vera Rubin 最近据说对散热架构做了很新的调整。 “台湾的热管理供应商正在成为 AI 硬件生态里增长最快的板块之一” - 上个月的说法。 “Vera Rubin server 架构预计会推动数据中心散热和系统设计发生根本变化” 我之后会专门看热管理生态，也许该开始研究一下？ 4. 2D NAND 短缺在 Samsung、Micron 和其他玩家退出后继续恶化。 Macronix、Windbond 都在起飞，这对 GigaDevice 和其他利基玩家也有影响。 5. “大科技公司据称愿意出资帮助 SK Hynix 建 fab 和 EUV” 说明 memory 短缺严重到 Mag7 都愿意出钱，那 $MU、SK Hynix、Samsung 都在起飞。 6. $TSM 2026 年 4 月净营收 126 亿美元，收入同比增长 30%，半导体继续起飞。 7. Anthropic 需要算力 -> SpaceX。 这意味着算力需求极其强，对 $NBIS 等都是“BRRR”级别利好。 但很有意思的是，他们绕开了 Neoclouds，选了 SpaceX。 8. “SKC 将在今年年底前加速为美国客户量产玻璃基板” “提前于原计划，已宣布今年年底前推进” 像 $LPK 这样的玻璃基板量产玩家，以及 SKC 这类相关公司都在起飞。 玻璃时间线被提前了，真的 heavy brrr glass。 9. AI 服务器需求和 GaN 争夺加剧，电源芯片短缺也在加深。 也许是时候看看电源芯片瓶颈了？ 10. Adata 说 DRAM 和 NAND 闪存合约价在 2026 年第二季度都会上涨 40% 以上。 这对 $MU、SK Hynix、Samsung、$SNDK 等都是正面。
英文原文
Just a TLDR of recent semi developments: 1. $TSM pushing hard CoPoS - VisEra/others might go brrr earlier than expected. 2. $AAPL goes with $INTC for semi production, which is a major shift cause they normally go with TSM. Made in America go like Intel go brrr. 3. $NVDA Vera Rubin reportedly makes changes to cooling architectures very recently. "Taiwan's thermal management suppliers are emerging as one of the fastest-growing segments in the AI hardware ecosystem" - From Last Month. "Vera Rubin server architecture is expected to drive a fundamental shift in data center cooling and system design" Will cover thermal ecosystem later, maybe it's time to take a look? 4. 2D NAND shortage spirals after Samsung, Micron, and rivals exit market Macronix, Windbond go brrr. implications for GigaDevice and other niche players. 5. "Big Tech reportedly offers to fund SK Hynix fabs and EUV" - Memory that badly bottlenecked that mag7 wants to pay for it, so $MU, SK Hynix, Samsung go brr. 6. $TSM 2026 net revenue $12.6B for April 2026. Revenue up 30%, Semis keep going brr. 7. Anthropic needs compute -> SpaceX. So implications for compute demand is extreme here which is BRRR $NBIS and others. But it's very interesting they sidestepped Neoclouds and went with SpaceX. 8. "SKC to Accelerate Mass Production of Glass Substrates for U.S. Clients by the End of the Year" "the end of the year, ahead of its original plan, it has been announced" Glass Core substrates players like $LPK for mass production and other related players like SKC go brrr. Glass timelines moved up. heavy brrr glass. 9. "Power chip shortages deepen as AI server demand and GaN battles escalate" Maybe time to look into the power chip bottleneck anon? 10. "Adata said DRAM and NAND flash contract prices will each climb more than 40% in the second quarter of 2026" Another positive for $MU, SK Hynix, Samsung, $SNDK, and others.
-
列出一串相关光学链条公司
@milesgillhespy Sumitomo Electric、Furukawa Electric、Mitsubishi Electric、$MTSI、$LITE
英文原文
@milesgillhespy Sumitomo Electric, Furukawa Electric, Mitsubishi Electric, $MTSI, and $LITE
-
预计美国上市后股权会从散户转向机构
@Philippe45164 我真觉得这会在美国上市后发生。 然后股份会从美国散户转向美国机构。
英文原文
@Philippe45164 I'm expecting that to happen after US listing tbh. Then shares go from US retail -> US institutions.
-
解释 CPO 前置布局下短期波动和做空比例
@TFillkcs80260 对，如果你在提前押 CPO 周期，波动会很大，因为收入还没真正进入报表。 我估计这一周又有大约 3% 的流通股被做空，所以现在可能在 9% 到 10% 左右。
英文原文
@TFillkcs80260 Yeah there's probably going to be a lot of volatility if you're frontrunning the CPO cycle since revenue hasn't hit the balance sheet yet. Think another ~3% of the float has been shorted this week so should be around 9-10% around now i think.
-
说明 CPO 的真正放量在 2027 年，Auros 还在认证中
MSSCorp 也是一个 CPO 标的……CPO 真正开始放量是在 2027 年。 2026 上半年并不会有太多量，所以在那之前大概率会有波动。但我觉得，提前等在这里去抢下一个巨大超级周期的起点，还是很有吸引力的。 至于 Auros，他们给 SK Hynix / Samsung 供货，产品还在认证中。我的 thesis 主要还是等到 H2 某个时候的量产爬坡。
英文原文
MSSCorp is a CPO names... CPO really starts to ramp up in 2027. You don't really have any volume H1 2026, so there's probably going to be volatility up until then. But I think the time wait is compelling enough to frontrun one of the largest supercycles at the very start. As for Auros, they supply to Sk Hynix/Samsung and are qualifying their products. My thesis was mainly about waiting for volume ramp sometime to hit H2.
-
SIVE激光器供应商客户覆盖多家AI/硅光子明星企业，估值对应的超大规模客户量惊人
$SIVE 向以下公司供应激光器：可能是 $JBL、Ayar、Lightmatter、$POET、$MRVL Celestial、Lightelligence、$AMD、O-Net/Enablence 以及其他公司。 这还不包括他们与 $AAPL 或 $AEVA 的硅光子(silicon photonics)项目，或者他们其他与CHIPS法案相关的工作。 以这个估值来看，他们所服务的超大规模企业(hyperscaler)数量相当惊人。
英文原文
$SIVE supplies lasers to likely $JBL, Ayar, Lightmatter, $POET, $MRVL Celestial, Lightelligence, $AMD, O-Net/Enablence, and others. Not including silicon photonics programs with $AAPL or $AEVA or their other CHIPS act related work. The amount of hyperscaler they feed into is pretty crazy at that valuation.
-
解释为什么 SIVE 总被瑞典媒体打压，但浮筹仍在转移
SIVE 每天都在重复上演这种剧情，已经像 deja vu 了，哈哈。 这也说明为什么 Sivers 很快会在纳斯达克上市，而不是继续留在本地市场。 每天开盘本地媒体都会来一篇攻击这家前沿公司的文章，然后股价跌 10%： -> “CPO 和 CW lasers 没什么新东西” -> “Sivers 要去跟资金雄厚的 $LITE 和 $COHR 竞争，肯定会因为资本劣势输掉” -> “为了上纳斯达克不得不稀释 2.5%” -> “年报延迟很可疑”（其实是为了上纳斯达克） -> $SIVE 体量不够，无法放大和 $JBL 的关系并扩张。 最后本地瑞典人都在卖，西方投资者 / 基金在接手流通股。 在 CPO 量产前就让西方持有这家公司，显然更好。 纯靠瑞典本地去做一家前沿公司，感觉希望很渺茫，因为本地文化就不太行。
英文原文
Every day it's like this on repeat with $SIVE, it's like dejavu lol. Makes sense why Sivers is getting listed on Nasdaq soon instead of staying in local markets. New local hit piece on their own frontier company happens every market open, stock drops 10%: -> "CPO and the CW lasers is nothing new" -> "Sivers is going up against well funded $LITE and $COHR and will lose because of capital" -> "They had to dilute 2.5% to get listed on Nasdaq" -> "Delayed annual report is sketchy" (to get listed on Nasdaq) -> $SIVE is not big enough to capitalize on $JBL relationship and scale. Local Swedish folks end up selling. Western investors/funds end up acquiring the float. Better for the West to own the company before CPO ramp starts. Creating a frontier company purely from Sweden seems hopeless given local culture.
-
解释 NVDA 对 IREN 的支持并不等于真正投资
$NVDA 什么都没投。他们拿到的是一个完全无风险的购买协议，自己没有承担任何投入。 这和可转债不同；可转债是他们借钱给 $IREN，然后按 70 美元转换。 很多人把融资结构的细节混为一谈，但如果他们真给 $NBIS 投了 20 亿美元，那就完全是另一回事。
英文原文
$NVDA hasn't invested anything. They got a completely risk-free purchase agreement without committing anything. This is different than a convertible note where they would loan $IREN money and to convert it at $70. People conflate all the nuances of financing structures, it's actually material if they put $2B into a company like $NBIS.
-
强调 NVDA 只是给了无风险认购文件
@OrkunEC Nvidia 基本上只是给了 $IREN 一个完全无风险的股份购买票据，自己根本没投钱。 这甚至都不是可转债，因为他们没有给 $IREN 任何现金。 他们让 $IREN 用自己的品牌去继续稀释并买 Nvidia GPU，自己却不会承担损失。
英文原文
@OrkunEC Nvidia basically got a completely risk-free share purchase note for $IREN and hasn't invested at all. This wasn't even a convertible note, where they gave $IREN any money. They're not losing anything for letting $IREN use their brand to dilute more and buy Nvidia GPUs.
-
认为 LPK 是玻璃基板超级周期的更干净玩法
$LPK 在过去两周涨了 80%。在大约 6.87 亿美元市值下，这成绩不算差。 它大概是下一个玻璃基板超级周期里最干净的玩法之一。 量产之后，每个客户可能要 50 到 100 台机器，2027 年的量产节奏会覆盖 $INTC、$GLW、SKC 等很多公司（因为它们已经拿到了大约 80% 的主要玩家）。 每台机器平均大概 200 万欧元。 仅仅 5 家玩家在 2027 年就可能带来 4 亿到 10 亿欧元以上的收入（甚至更多）？因为它们基本上是大家都在用的 chokepoint。按照约 67.6% 的综合毛利率看，等量产爬坡还是挺有希望的。
英文原文
$LPK up 80% in the last two weeks. Not too shabby at ~$687M MC? It's probably one of the cleaner ways to play the next Glass Substrate supercycle. 50-100 machines per customer at scale, with "start of 2027 as mass production" across likely $INTC, $GLW, SKC, and others (since they captured ~80% of the major players). Maybe ~€2M average per machine. €400M–€1B+ across just 5 players in 2027 (could be more)? Since they basically supply to everyone as a chokepoint. Off ~67.6% blended gross margins. Seems promising for volume ramp wait time.
-
认为算法看不懂供应链映射，只会继续做空
大概率还是纯算法行为，但昨天又有 0.76% 的流通股被做空了。 算法通常看不懂很多供应链映射细节，比如 $AMD 和 $GFS 一起做 CPO。 然后在 $GFS 的图片展示里，$SIVE 只被列成两个激光供应商之一。这就是 alpha，因为你知道别人不知道的东西。
英文原文
It's likely pure algorithmic, but another .76% of the float was shorted yesterday. Algorithms tend to miss a lot of the nuances of supply chain mapping like $AMD going with $GFS for CPO. Then $SIVE listed as 1 of 2 laser suppliers in $GFS image presentations. This is called alpha though since you know something others don't.
-
继续讽刺瑞典媒体对 SIVE 的恶意报道
我还在笑瑞典人有多讨厌自己的前沿公司。 他们天天给 $SIVE 写打压文章。 这篇就很有娱乐性：本地记者未经邀请就跑去一栋空着的 $SIVE 行政楼前。 因为他们无法理解 CEO 其实在硅谷，或者设计团队正在美国做 CHIPS Act 相关开发。再加上楼外面没停多少车，以及 CFO 不愿回答关于保密 hyperscaler 合同财务细节的问题。 于是他们又写了一篇莫名其妙的负面稿。 还反复写“做这种激光的公司不止一个，Sivers 远不是唯一一家”。 比如 $LITE、$COHR 这些 600 亿美元以上的公司。 又写早前那种“CPO 没什么特别，很多年都有了”的说法。 而 GS 的预测却是 CPO 在未来两年从 10 亿美元 TAM 增长到 910 亿美元。 他们甚至还把“Plans”写成引号，好像不相信 Sivers 在给 $JBL 的 1.6T LRO 供激光。 我认为 $SIVE 明年会变成一家 100 亿美元以上的公司，尤其如果他们像 $LITE / $COHR 那样把下游 IP 集成起来，多吃一点 CPO 模块 BOM。 他们根本不懂 hyperscaler 供应链、前瞻成长，以及员工数量不等于收入这件事。 当地瑞典持股向西方转移，这件事其实一直是好事，因为它曾经是一家由本地散户主导的公司。
英文原文
I'm still laughing how much Swedish hate their own frontier companies so much. That they write hit pieces every day on $SIVE. This one was entertaining: Local journalists show up to an empty $SIVE administrative building uninvited. Because they can't fathom the CEO is in Silicon Valley or design team is working on US Gov CHIPS act dev in the US. And because there weren't many cars parked outside + CFO wouldn't take questions about secretive hyperscaler deal financials. They wrote a random negative hit piece. By repeating "There are several who make lasers like these and Sivers are far from alone". Several like $LITE, $COHR, $60B+ companies. and reported earlier that "CPO is nothing special, it's been around for years." While GS projects CPO going from $1B -> $91B TAM over the next two years. Even put "Plans" in quotation marks because they didn't think Sivers is supplying lasers to $JBL 1.6T LRO. IMO, $SIVE ends up as a $10B+ company next year, especially if they follow what $LITE / $COHR did with downstream IP integration to capture more of CPO module BOM. Just don't think Swedish people understand hyperscaler supply chains, concept of forward growth, or the fact that employee count doesn't equate to revenue. Transfer of control from local Swedish -> West is always appreciated, as this was a majority owned local retail company before.
-
认为 CPO 是对所有光子公司都成立的新机会
$LITE 原本已经把 EML 和现有 pluggable optical transceivers 的价值定价进去了。 CPO 对所有光子公司来说，都是一个全新的机会，也是巨大的 TAM 扩张。 这是一次重大的架构变化，像 $SIVE 或 $SOI 这种过去没怎么受益的公司，也会明显受益。
英文原文
$LITE was priced in for EML and current pluggable optical transceivers. CPO is a completely new opportunity and massive TAM expansion for all photonics players. It's a massive architectural shift, and completely new optical companies benefit like $SIVE or $SOI that didn't as much before.
-
用 Coherent 财报再次验证 CPO 的时间线
$COHR 财报笔记： Coherent 的 CEO 基本上又一次印证了 GS 研究里的判断，CPO 会成为非常大的收入驱动。 转录里有一句："One of the most important long-term growth opportunities for Coherent"。 对 $SIVE、$LITE 和其他 CPO 相关名字的时间线含义也很清楚： “初始 scale-out CPO 收入预计在 2026 年下半年开始，而 scale-up CPO 收入将在 2027 年下半年跟进”。 “Coherent 正在和多个其他客户合作，并预计 CPO 解决方案会被广泛采用。” 这也再次确认了采用程度。 scale-up CPO 是全盘最大的价值驱动点，会在明年下半年开始。我们现在还在 2026 年上半年，正进入下半年。 所以现在就是在任何东西落到报表前，提前抢跑 CPO 标的的理想时点。 这就是我说的“alpha”——在市场还看不见财务数据时，把未来先定价进去。 市场通常会提前 8 到 12 个月看未来……所以如果大规模量产在 2027 年 7 月开始（也可能再提前）， 而现在是 2026 年 5 月，那一大段定价会在 2026 年 7 月到 2027 年 7 月之间逐步发生…… 也就是在 2026 年下半年到 2027 年上半年之间，先把量产协议和前生产爬坡这些东西价格化。 在市场真正开始定价最大量能驱动之前 2 个月提前布局，就是我所谓的在下一轮 CPO 超级周期里“抢跑”。 我们知道它会来。现在只是在等。
英文原文
$COHR earnings note: Coherent's CEO basically reaffirmed GS research note about CPO being a massive revenue driver: Transcript: “One of the most important long-term growth opportunities for Coherent” In terms of timeline implications on $SIVE, $LITE, and other CPO adjacent names, just confirms timelines: "Initial scale-out CPO revenue is expected to begin in the second half of calendar 2026, with scale-up CPO revenue following in the second half of calendar 2027" "Coherent is working with multiple other customers and expects CPO solutions to be widely adopted." - This also reaffirmed adoption level. Scale-up CPO is the biggest value driver across the board that happens start of H2 next year. We're still in H1 2026 entering H2 2026. So this is the ideal time to frontrun CPO names before anything hits the balance sheet. This is known as "alpha"... pricing in things market fully don't know yet by looking at financials. Markets are typically forward looking 8-12 months... So if massive volume ramp happens July 2027 (might get pushed forward again)... And it's May 2026, a lot of that gets priced in between July 2026 and July 2027... between volume agreements H2 2026, and pre-production ramp H1 2027. 2 months before markets price in the largest volume drivers is what I call "frontrunning" the next CPO Supercycle. We know it's coming. Just a matter of waiting.
-
认为 LPK 是最喜欢的玻璃基板敞口
不，我现在最喜欢的敞口大概是 $LPK，因为它们给所有人供货。 我的意思是，如果你做多玻璃核心基板，Samsung、$INTC 和其他公司可能会更大一些。 SKC 可能是这个领域里市值最低的龙头。 然后如果它们跑通了，HB 大概会顺带受益，并且最终给其他玩家提供和良率相关的东西。
英文原文
No. $LPK is probably my favorite exposure right now since they supply to everyone. I'm just saying if people are long glass core substrates, Samsung, $INTC, and others are probably a bit bigger. SKC is probably the lowest MC leader in that space. Then if you see them succeeding, then HB gets a piggyback + probably ends up supplying yields related stuff to the other players.
-
表示自己在 LPK 上的仓位最高
@tw_crypto_ 对，我在 $LPK 上的仓位最高……感觉像 HB / LPK 这种更上游的供应商，比 SKC 更去风险。
英文原文
@tw_crypto_ Yeah I have highest concentration in $LPK... feels like more upstream suppliers like HB/LPK are more derisked than SKC.
-
认为做多 SKC 生态不算差
做玻璃核心基板的话，做多 SKC 生态好像也不算太差？ SKC（011790）—— 在 CHIPS Act 资金支持下，是玻璃核心基板的领导者之一，市值约 55 亿美元。 HB technologies（078150）—— 自动光学检测 / 良率，借着 SKC Absolics 做二次增长，市值约 2.5 亿美元。 SKC 的铜箔 / 电动车旧业务拖累似乎已经在见底。它的财报看上去像一团糟，但其实已经被市值定价进去，而且比那些有 60 亿美元以上稀释的美国公司要好。 HB 风险更低，估值也不高，而玻璃的良率会变得非常重要。 我在这两只上都只有很小的仓位。 本来还想再等一等，但市场看起来已经在把玻璃定价成下一个可能的 memory / photonics 交易。
英文原文
For Glass Core Substrates, doesn't feel too bad going long on the SKC ecosystem? SKC (011790) - Is one of the Glass Core Substrate leaders with CHIPS ACT funding at ~$5.5B MC HB technologies (078150) - automated optical inspection / yields piggy backing off of SKC Absolics at ~$250M. SKC's copper foil / EV legacy drag seems to be bottoming. Their financials seem like a dumpster fire, but it's kinda priced into MC + better than US companies with $6B+ in dilution. HB seems lower risk, low p/e and yields for glass is going to be pretty important. I do have very small exposure to both. Was going to wait a bit longer but seems markets are pricing in Glass as the next possible memory/photonics trade.
-
MicroLED技术就绪度低，$SIVEF/$AAOI等CW光模块企业更值得布局
MicroLED在高盛报告所述的任何CPO相关批量增长之前，仍然感觉还很遥远。 Quantum Dot（如$ALMU/QD）甚至没有被提及，因为它的前景在数年之外。 在调整中，"溢价"会消失，但估值底部由收入支撑。 另一方面，$SIVEF或$AAOI即将进入批量生产阶段（<800万），直接处于CW领域。 我更倾向于在经过多年研发/扩张之后立即获得上涨空间... 而不是等待数年从研发到样品再到认证...然后才到达批量增长的转折点。 MicroLED的"技术就绪度"很低。
英文原文
MicroLED still feels far away before any CPO related volume ramp per Goldman Sachs note. Quantum Dot like $ALMU / QD isn't even mentioned since it's way way out by years. In a correction, "premiums" disappear, but valuation floors are backed by revenue. $SIVEF or $AAOI on the other hand are volume ramping soon <8M and directly in the CW space. Would prefer immediate upside after years of R&D/expansion... Than waiting years at the beginning for R&D into sampling into qualification... Then inflection point of volume ramp. "Technology readiness" is low for MicroLED.
-
LITE财报后市场误读CPO利好导致光学股被错杀，CPO超级周期才刚开始。
我认为很多人会很惊讶地发现，股票并不会直线上涨。 今天： 从 $LITE 到 $SIVE 再到 $AAOI，激光类公司分别下跌 -8.0%、-10.2% 和 -4.38%。 台湾涉及共封装光学(CPO)概念的公司如 MSSCorps、Shunsin 以及相关公司 Win Semi 跌停 -10%。 $AXTI 和 $SOI 分别下跌 -7% 和 -10.2%。 这是在 $LITE 发布财报之后。 我非常确定，市场错过了财报电话会议中关于 CPO 概念的强烈利好信号，但对于 $SIVEF 或 Shunsin 这样的 CPO 概念股来说，这是极度看涨的，然而算法交易却卖出了所有光学股。 这仍然是整个 CPO 超级周期的非常早期的阶段。 在任何量能上升之前。
英文原文
I think many people are surprised to learn that stocks don’t move in a straight line up. Today: Laser companies from $LITE to $SIVE to $AAOI are down -8.0%, 10.2%, and -4.38%. Taiwan limit down -10% on CPO names like MSSCorps, Shunsin, and adjacent like Win Semi. $AXTI and $SOI are down -7% and -10.2%. After $LITE reported earnings. Pretty sure markets missed the heavy nuance that this was extremely bullish for CPO names like $SIVEF or Shunsin from the earnings transcript, but algos sold off everything optical. This is still the very beginning of the entire CPO supercycle curve. Before any volume ramp.
-
LPK 继续垂直上涨，玻璃生态很热
@tw_crypto_ 对，$LPK 在垂直线往上走，玻璃生态很热闹。今天 photonics 虽然跌了，但 memory + glass 能把它平衡回来。
英文原文
@tw_crypto_ Yeah $LPK is going in a vertical line up, fun times for the glass ecosystem. Photonics is down today, but memory + glass helps balance it out.
-
说明 MSSCorps 做的是 CPO 检测
@itsadnyc MSSCorps 做的是 CPO 检测……
英文原文
@itsadnyc MSSCorps is CPO inspection...
-
观察到玻璃基板相关标的今天被机构一起买入
我感觉今天机构就是在买所有和玻璃核心基板有关的东西？ 像： - Philoptics，涨停 - HB technologies，涨停（SKC 供应商） - YCCHEM，涨 20% 而欧洲那边还有 $LPK。 不知道我是不是漏掉了什么重大消息。
英文原文
I feel like institutions are just buying everything glass core substrate related today? Stuff like: - Philoptics, limit up - HB technologies, limit up (SKC supplier) - YCCHEM +20% Then there's $LPK over in Europe. Wonder if I missed any major news.
-
认为 LITE 其实是在说 InP substrates 不够
我觉得这只是 $LITE 的公司措辞，意思其实就是： “我们没有足够的 InP substrates”。 还记得我去年说过，随着 photonics 放量，$AXTI 会继续作为关键瓶颈而增长吗？ https://t.co/DfeQa3haIK
英文原文
I think this is just $LITE corporate wording for: “We don’t have enough InP substrates”. Remember I sad last year that as photonics ramps up, $AXTI will just keep growing as a critical bottleneck? https://t.co/DfeQa3haIK
-
认为 LITE 外购 CW lasers 让 AAOI 和 SIVE 受益
是的，$LITE 承认自己被迫从竞争对手那里购买 CW lasers，这对 $AAOI 来说是巨大利好。 因为他们自己就有 CW laser fab，而且目标是在 2027 年把产能提升 350%。 $LITE 财报电话会最重要的 nuance 是：$NVDA 已经在上游给 CW lasers 造成了巨大瓶颈。 很可能 $COHR 和 $LITE 都签了多年的 EML 协议，所以它们被锁死了，没法生产足够的 CW lasers。 我持有的一些原来被认为是“Tier 2”的名字，比如 $AAOI 和 $SIVE，现在很可能因为没有被 Nvidia 完全分配产能而地位上升。
英文原文
Yes, $LITE admitting they're forced to buy CW lasers from their competitors is a massive win for $AAOI. Since they have their own CW Laser fab (and target 350% increase in capacity by 2027). The massive nuanced takeaway from $LITE earnings call: $NVDA caused a massive upstream bottleneck for CW lasers. It's likely $COHR and $LITE have multi-year EML agreements, so they're locked up and can't produce enough CW lasers. A lot of the former "Tier 2" players I own like $AAOI and $SIVE, are likely elevated in status now given they're not fully allocated to Nvidia.
-
发现 LITE 外购 CW lasers 对 SIVE 是重大利好
这是个新信息：$LITE 被迫从外部竞争对手那里买 CW lasers。 很可能是因为他们和 EML 签了 2027 到 2028 的合同，自己产能不够…… 所以 $SIVE 会成为一个很大的受益者，因为其他 hyperscaler 供应商会转向别的 CW laser 供应商，还有亚洲那些大玩家，比如 Sumitomo + $COHR。 我们看看市场能不能把 1 加 1 连起来。
英文原文
This is new information discovery that $LITE is forced to buy CW lasers from external competitors. Likely because they have EML contracted from 2027-2028 and can't produce enough CW... So $SIVE is a massive beneficiary of other hyperscaler suppliers going to other CW laser suppliers + the big ones in Asia like Sumitomo + $COHR. We'll see if markets can put 1+1 together.
-
LITE 的 CW laser 瓶颈验证了 SIVE 的上游地位
哇……巨大的 CW laser 瓶颈。 有意思的是，$LITE 被 CW lasers 卡住了，所以它们要去公开市场买： “现在我们在自己的供应链里，确实看到一点紧张，必须去外部市场拿 lasers，也就是 CW lasers。” $SIVE / Win、Sumitomo、Furukawa 等都在一个结构性卖断货的市场里…… 很可能是因为像 $LITE 这样的公司必须继续生产 EML。 这也开始解释了为什么 Ayar 在网站上只把 $SIVE 列为唯一 laser 供应商，还把 $LITE / $MTSI 删掉了。
英文原文
Woah... Massive CW laser bottleneck. It's interesting $LITE is bottlenecked by CW lasers so they buy them on the open market: "Now we’ve seen a little bit of tension in our own supply line, externally to get lasers from the external market, CW lasers." $SIVE / Win, Sumitomo, Furukawa, and others operate in a structurally sold out market... Likely because companies like $LITE have to keep producing EMLs. Starting to make more sense why Ayar made $SIVE the only laser supplier on their website and removed $LITE / $MTSI?
-
CPO面临全行业供需失衡，多处产能瓶颈将显现。
$LITE CEO在财报电话会中措辞是"CPO存在巨大的供需失衡"。 我认为即便是$SIVE、$GFS、Win、Shunsin，以及所有CPO相关公司，都会面临需求过剩的局面……从而在多个环节引发全行业瓶颈。 但话说回来，$AMD的CPO项目可能不得不选择其他激光供应商，比如$SIVE，因为$NVDA锁定了产能。
英文原文
$LITE CEO on their transcript worded it as "massive supply demand imbalance on CPO" I think even $SIVE, $GFS, Win, Shunsin, and everyone CPO related will have too much demand... leading to an industry wide bottleneck on multiple chokepoints But yes, $AMD CPO program likely had to go with other laser suppliers like $SIVE as a result of $NVDA securing capacity.
-
认为 LITE 已确认 CPO 超级周期和 TAM 扩张
$LITE 的 CEO 基本上确认了 GS 的 TAM 扩张和 CPO 超级周期。 从接近 0 到 910 亿美元。H2 2026 到 2028： “我认为我们已经说过，CPO 会有巨大的供需失衡” 需求大于供给。 “最大的单一增长驱动，scale-up CPO” 巨大的收入驱动就是 CPO。 “现在还非常处于早期”。 我们正站在起点。 这正是为什么我会持有 $SIVE、MSSCorps、Shunsin、$TSEM、$SOI 等，去获得对 CPO 超级周期起点的高 beta 暴露。
英文原文
$LITE CEO basically confirmed GS TAM expansion + CPO Supercycle. From ~close to nothing -> $91B. From H2 2026 to 2028: “ I think what we’ve said is that we will have a massive supply demand imbalance on CPO” Demand > Supply. “Largest single growth driver, scale-up CPO” Massive revenue driver is CPO. “is very much in its infancy”. We’re at the beginnings. This is exactly why I have positions in $SIVE, MSScorps, Shunsin, $TSEM, $SOI and others for high beta exposure to the start of the CPO supercycle.
-
Sivers处CPO供应链前沿，910亿美元TAM蓝海刚起步，H2超级周期可期。
只是近期信息发现的回顾 + 与$SIVE的可能映射： -> $JBL 1.6T -> Lightmatter -> Ayar -> $MRVL Celestial -> Lightelligence -> $POET -> $GFS生态 -> $AMD CPO -> O-Net / Enablence -> $AAPL 硅光子学 _ -> $YSS Golden Dome/DoD -> $RTX / $ERIC (太空) -> Bae Systems -> $AEVA $JBL到Ayar供应给超大规模云商如$MSFT、$GOOGL、$AMZN、$META。 Lightelligence到O-Net可能供应给亚洲超大规模云商如腾讯、字节跳动和百度。 在此之上……根据高盛报告，CPO的整体TAM（总可寻址市场）从0增长到$91B。 而Sivers恰好是CPO领域的前沿（也从0开始）。 这绝对是高贝塔和高波动的。 但如果Win卷随着$SIVE一起增长，我看到他们明年都能成为$10B+的公司。 这只是CPO超级周期H2 2026开始之前非常早期的阶段（H1）。
英文原文
Just a recap of recent information discovery + likely mapping with $SIVE: -> $JBL 1.6T -> Lightmatter -> Ayar -> $MRVL Celestial -> Lightelligence -> $POET -> $GFS ecosystem -> $AMD CPO -> O-Net / Enablence -> $AAPL Silicon Photonics _ -> $YSS Golden Dome/DoD -> $RTX / $ERIC (Space) -> Bae Systems -> $AEVA With $JBL to Ayar feeding into hyperscalers like $MSFT, $GOOGL, $AMZN, $META. With likely Lightelligence to O-Net feeding into Asian Hyperscalers like Tencent, Bytedance, and Baidu. On top of that... the overarching TAM with CPO from the GS report goes from 0 -> $91B. And Sivers happens to be the bleeding edge for CPO (also starting from 0). This is definitely high-beta and volatile. But if Win volume ramps alongside $SIVE, I see them both becoming $10B+ companies next year. This is just extremely early on (H1) before the CPO supercycle starts H2 2026.