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
-
说 LITE 已卖到 2029 年，CPO 量产已确认。
@itsakek - $LITE 已经卖到 2029 年了。 - 2026 年下半年确认 CPO 量产。 $NVDA CPO / photonics 上游供应链直接起飞。
英文原文
@itsakek - $LITE sold out into 2029. - Confirmation CPO mass production H2 2026. $NVDA CPO/photonics upstream supply chain go brrr
-
给出 2027 年业绩模型和 EPS 推演。
好问题！大致是这样： - 2027 年营收：新台币 127.2 亿 - 毛利（26% 毛利率）：新台币 33 亿 - 运营费用（约新台币 11 亿） - 营业利润：新台币 22 亿 - 综合税费 + 非经营性：新台币 3.3 亿 - 净利润：新台币 18.7 亿 - 每股收益约新台币 17.61+，向 20+ 看 前些年因为在河内 / 北江扩建 CPO 设施，资本开支和折旧都很多，所以毛利率扩张、产能利用率到 80% 后会去到 26%。 模型基于 Shunsin 拿下 350-400 万个 SiPh/CPO 封装单元，ASP 为新台币 2,350。 350 万 × 2350 = 新台币 82 亿 传统业务（稳定后）：新台币 45 亿 8.2 + 4.5 = 2027 年营收 127 亿。并且还应该能随着 2028、2029、2030 年的光学预期继续扩张。
英文原文
Good question! Basically: -NT $12.72B revenue 2027 -Gross profit (26% margin): NT $3.3B -Opex (~NT$1.1B) -Operating income: NT $2.2B -Blended Tax + non-op: NT$330m -Net income: NT$1.87B ~NT$17.61+ EPS to $20+ Lot of capex/depreciation in former years because of scaling up CPO facilities in Hanoi/Bac Giang, so gross margin expansion and 80% utilization to 26%. Modeled based off Shunsin capturing 3.5-4M SiPh/CPO packaging units witH ASP NT$2,350 3.5M × NT2,350= NT $8.2B Legacy Business (Stabilized): NT$4.5B 8.2 + 4.5 = 12.7B rev for 2027. And should scale up based on optical projections for 2028, 2029, 2030.
-
说 6451 像是鸿海光学业务的低估去风险化标的。
@latent_value7 对……我觉得这就像一趟低市值、低 forward multiple 的鸿海光学业务去风险化搭便车。 我真的很喜欢 $6451.TW。
英文原文
@latent_value7 Yeah... looks like a de-risked foxconn optical segment piggy back ride to me at a low MC and low fwd multiples. I really liked $6451.TW.
-
用 IQE 的资产负债表和光子转型解释重估空间。
在我原来的 thesis 里，我提到 Landmark 当时是 38 亿美元市值，而它的 InP reactor 产能还比 $IQE 更少。 但 $IQE 虽然有负债，却有很多资产，只需要把这些资产重组到 photonics 方向上，再把资产出售做完，把资产负债表理顺，就能迎来重估。 我当时觉得涨到 10 亿美元以上是有可能的，我们拭目以待。
英文原文
In my original thesis, I cited landmark was a benchmark at $3.8B with less inp reactor capacity than $IQE. But $IQE had debt but a lot of assets, just needed to refactor all those toward photonics + finish asset sales to clear up balance sheet for re-rating. I thought 10x to $1B+ was possible, we'll see.
-
量子点激光投资应优先选择MBE设备供应商如Riber，比下游纯激光公司更早受益于超大规模云厂商资本开支周期。
如果你想交易超大规模云厂商的资格认证周期……以及像QD Laser（TYO: 6613）这样年内涨幅达226%的量子点激光：IMO最安全的方式是在早期就关注他们未披露的MBE机器供应商，如$ALRIB。然后当他们从资格认证阶段转向批量订单时，再转向纯激光公司。我之前提到过，$MSFT量子的Riber未披露的超大规模云厂商客户。但QD Laser（越来越受欢迎）在量子点项目上也严重依赖Riber。我不看好进入批量生产阶段（比如$AIXA对比实际的批量生产商），因为设备厂商无法捕获下游收入。但如果你想从资本开支研发周期中获益……比如QD Laser吸收资本开支成本（支付给Riber）来建设产能，$ALRIB、$ASML、$AIXA这类公司的财务数据通常会更早反映在资产负债表上，所以你不需要等2-3年才看到生产订单。
英文原文
If you want to trade hyperscaler qualification cycles… And quantum dot like QD Laser (TYO: 6613) up 226% YTD: IMO safest way to do it are their unknown MBE machine suppliers like $ALRIB early on. Then pivot to pure play lasers later when they move from qualification -> volume orders. I mentioned earlier $MSFT quantum was Riber’s undisclosed hyperscaler customer. But QD laser (getting popular) is heavily reliant on Riber too for their quantum dot program. I’m not a fan when it comes time to volume production (eg. $AIXA vs the actual volume producers) since machines don’t capture downstream revenue. But if you want to benefit from capex R&D cycles… eg. QD Laser absorbing capex costs (paying Riber) to build capacity $ALRIB, $ASML, $AIXA type companies usually hits the balance sheet much earlier so you don’t need to wait 2-3 years for production orders.
-
担心味之素被激进股东逼涨 ABF 价格。
天啊，我真的希望 Ajinomoto 不要向这些激进股东让步。 WSJ 终于把旧消息翻出来了： “Palliser 入股 Ajinomoto，要求 ABF 价格上涨 30%+”（这是个关键卡点） 是的，我也对 $AXTI 做过类似的事……但那是美国 / 中国的问题，所以还算可以。 这也是为什么没人去照搬同样的剧本去对日本公司下手。 试图用西方对冲基金的掠夺式涨价来改变日本文化，是不被鼓励的。
英文原文
Holy ****, I really hope Ajinomoto doesn’t give into these activist shareholders. WSJ finally broke older news that: “Palliser Takes Stake in Ajinomoto, Urges 30%+ ABF Price Increase” (major chokepoint) Yeah, I did the same with $AXTI … but that’s US/China so it’s okay. There’s a reason why nobody copied that same playbook for Japanese companies. Trying to change Japanese culture with Western hedge fund predatory price hikes is not encouraged.
-
用 Empresas CMPC 和 Softys 的纸巾业务讲供应链梗。
@MoreyMuse Empresas CMPC 的远期市盈率大约是 7.8 倍。 它的 “Softys” 事业部在拉丁美洲制造纸巾和餐巾纸。 它以解决浴室和墨西哥豆卷饼之间的重大供应链瓶颈而闻名。 https://t.co/bL6Bpb7YvO
英文原文
@MoreyMuse Empresas CMPC trades at ~7.8x fwd p/e. Its "Softys" division manufactures tissue and paper napkins in latin america. It's known for solving the major supply chain bottleneck between the bathroom and chitotle pinto bean burritos. https://t.co/bL6Bpb7YvO
-
认为英伟达正在强力推动 CPO，相关时间点比市场预期更早。
$NVDA 正在非常、非常强力地推动 CPO。 从 $MRVL、$LITE 和 $COHR 的投入就能看出来。 不过有意思的是，CPO 的时间表也比预期早得多。 最新报道： “$ASE 据称将在 2026 年动工建设六座新工厂，CPO 量产预计将于今年开始” “他还首次透露，CPO 的量产预计将在今年启动” 现在可能正是提前布局 CPO 相关标的的好时机，比如 $SIVE（激光器）、Win Semi（晶圆代工）、$TSEM（晶圆代工）、$SOI（衬底）以及其他相关公司，在真正放量前先埋伏。 需求曲线会一路指数级上升，直到 2029 年以后，我相信很多预测都会被打穿。 总之，我之前在 OFC/GTC 前就一直说要提前布局 CPO / SiPh 标的。 结果时间点上我又判断对了？ 现在看起来 CPO 的放量要从 2026 年下半年开始，准备好吧。
英文原文
$NVDA is driving CPO, extremely, extremely hard. As seen with investments in $MRVL, $LITE, and $COHR. But, what's interesting is: CPO timelines are also happening way earlier than expected. New report: " $ASE Reportedly to Break Ground on Six New Plants in 2026, CPO Mass Production Expected to Begin This Year " "He also revealed for the first time that mass production of CPO is expected to begin this year" Probably a good time to frontrun CPO related names like $SIVE (lasers), Win Semi (foundry), $TSEM (foundry), $SOI (substrates), and others around now, before ramp really picks up. The demand curve goes up exponentially all the way past 2029+, and I'm sure they'll break many projections. But basically, I was talking about frontrunning CPO/SiPH names before OFC/GTC. Turned out I was right on timing again? CPO Ramp now looks like it's beginning H2 2026, get ready.
-
根据以往先例推测，Win Semi 可能会迎来新的长期客户。
@iamsattam 我引用的是过去的先例：当 Win Semi 宣布定向增发时，$AVGO 成了他们的主力客户。 我的猜测是，他们这次通过这份公告，等于又拿到了一个新的长期客户。
英文原文
@iamsattam I'm citing this as the old precedent on when Win Semi announced private placements, $AVGO became their lead customer. My guess was they just acquired a new long term customer by filing this.
-
猜测英伟达可能买入 Win Semi 股权，市场若解读成利空反而可能误判。
@Sim_Wizard 我的猜测是 $NVDA 可能在买 Win Semi 的股份，结合他们最近在光子学方面的动作和体量来看。 算法 / 市场上的人可能会把这解读成利空然后卖出，但我认为，考虑到 Win Semi 的历史，这其实更偏利好。
英文原文
@Sim_Wizard My guess was $NVDA buying Win Semi stake given their recent history in photonics + size. Algos/people might mistake this as bearish and sell, but I think it's actually bullish given the nuance of Win Semi's history.
-
认为 Win Semi 定增稀释约 9.4%，但结合历史先例和大客户逻辑，整体偏利好。
所以这次定增大概是通过发行 4000 万股新股，带来约 9.4% 的稀释，主体是 Win Semi（3105）。 换成别的股票我会很谨慎……但上一次 Win 这么做的时候，$AVGO 买入了 Win Semi 的股份。 而且博通后来成了他们的主力客户。 这次也许会是像 $NVDA 这样的一级客户，考虑到他们最近对 $MRVL、$LITE、$COHR 的投入。 另外还有 3 年持有期。 我之前说过，Win Semi 是光子学、人形机器人和太空（比如 SpaceX）领域的基础性公司，它很可能还会获得另一位一级客户并长期合作。 这很讲究细节，但实际上是偏利好的。
英文原文
So private placement is ~9.4% dilution via 40M new shares via Win Semi (3105). Any other stock I'd be cautious... but last time Win did this, $AVGO took a stake in Win Semi. And Broadcom became their lead customer. Maybe T1 semi like $NVDA, given their recent funding of $MRVL, $LITE, $COHR. There's also a 3 year holding period. I said before Win Semi is foundational to photonics, humanoids, and space (eg. SpaceX) and it's likely another T1 customer using them long term. It's nuanced but actually bullish.
-
称 Sivers 的激光业务被低估，是 Marvell 和 Jabil 的激光供应商。
@yy1625066197303 激光器就是在狂飙，基本上就是这样。我很喜欢 $SIVE 现在得到更多关注，感觉它作为 $MRVL、$JBL 的激光供应商被严重低估了。
英文原文
@yy1625066197303 Lasers go brrr basically, love $SIVE is getting more attention, feels extremely underrated as the $MRVL, $JBL laser supplier
-
认为 LITE 的产能订单已排到 2028 年，显示整个光子学链条需求强劲。
@thgstar2 $LITE 的产能订单已经排到 2028 年了，这说明整个光子学领域的需求都非常强。
英文原文
@thgstar2 $LITE capacity order filled into 2028, signaling massive demand for photonics across the board.
-
认为 LITE 的 backlog 和 AAOI 的自制能力都指向极强需求，AAOI 的市值目标开始显得可行。
我不确定大家是不是还没完全理解： $LITE 的 backlog 订单已经排到 2028 年，这说明需求极强，而且产能明显不足。 随后在超大规模客户需求外溢的二阶效应下： 猜猜美国被预测会拥有最大的 800G/1.6T 产能的是谁？ $AAOI。 他们自己制造 InP 激光器，自己设计收发器，还自己组装。 如果 $AAOI 能把产能爬坡执行到位，那大概率都会直接转化为营收，因为产品已经全部卖光了。 我从 50 亿美元市值推到 400 亿美元市值的目标价，现在看起来越来越可能了？
英文原文
I'm not sure people understand yet: $LITE backlog order fill into 2028 signals extreme demand. And a lack of capacity. Then by second order effect of hyperscaler demand spillover: Guess who is projected to have the largest 800G/1.6T capacity in America? $AAOI. They fab their own inp lasers, design their own transceivers, and assemble it. If $AAOI can execute on capacity ramp, that likely all translates into revenue due to everything being sold out. My $40B MC price target from $5B is starting to look more and more likely?
-
分析光子学超级周期下的CPO/激光技术路线机遇，SIVE作为小市值激光标的具有独特供应链位置。
光子学“超级周期”有许多细微的架构变化。如果你在去年错过了$LITE，今年可以通过CPO/激光来抢先布局下一个周期。$SIVE是我相对小众的激光敞口标的。他们参与在$JBL的1.6T LRO、$MRVL的Celestial、O-Net到超大规模业者。这对于一家市值约3.5亿美元的公司来说极为罕见。如果要我类比：类似于$AAOI在2025年的情况，或现在的$AEHR——正处于批量生产前的资质认证周期。正因如此，散户只看今天财务数据（负eps或capex支出）时会误解公司情况。而且与批量生产/capex支出相关的dfb激光扩产执行不确定性，已通过Win Semi被去风险化。然而看起来从2026年下半年开始会持续爬坡到2029年以后。可能在时间上稍早一些……但我认为这是对未来即将到来的趋势的一个不错研判。稀释是真实风险。但很多担忧已经被定价到约3.5亿美元市值里，尤其当$AVGO或AlChip这样的公司可以收购$SIVE用于垂直整合或干扰Marvell的CPO计划时。
英文原文
There's many nuanced architectural changes for "supercycles" in photonics. If you feel like you missed $LITE from last year, just frontrun the next cycle with CPO / lasers. $SIVE was my lesser known pick for laser exposure. They're in $JBL 1.6T LRO, $MRVL via Celestial, O-Net -> Hyperscalers. Which is extremely abnormal for a ~$350M MC company. If I had to draw parallels: it's similar to $AAOI 2025 or $AEHR now with current qualification cycles before volume. Which is why retail misunderstands it when you look at purely financial numbers today with negative eps or capex spend. And a lot of that execution uncertainty with volume production / capex spend for scaling dfb laser is de-risked from Win Semi. However, it looks to just keep ramping past 2029+ starting from H2 2026. Might be a little early... But I think it's a decent read on what's up and coming. Dilution is a real risk. But a lot of fears are priced into its ~$350M MC, especially when a company like $AVGO or AlChip can just buy $SIVE for vertical integration or to troll Marvell’s CPO program.
-
判断 SIVE / Win 会受新一轮 CW 激光周期带动。
$SIVE / Win 会受 2027 年下一轮架构超级周期里的 CW 激光推动。我现在只是像去年布局 $LITE 那样，提前在 H1 先埋伏。 你大概也知道，他们已经进了 $MRVL 的 Celestial、$JBL 的 1.6T LRO，以及 O-Net，所以到 2029 年会有很大的放量。 像 $AAOI、$LITE、$COHR、Innolight 这些则是当前的光收发器周期。
英文原文
$SIVE / Win benefits from CW lasers from the next architectural supercycle in 2027. I'm just frontrunning that now H1 like $LITE last year. You already kinda know they're in $MRVL Celestial, $JBL 1.6T LRO, and O-Net so thats huge ramp into 2029. Stuff like $AAOI, $LITE, $COHR, Innolight are current optical transceiver cycles.
-
认为美国光子学三巨头将受 AI 需求激增带动。
我们现在正处于光子学超级周期。 美国这三家 $COHR、$LITE、$AAOI 预计会受需求激增推动而表现不错。 最新报道： “Lumentum 正在经历来自人工智能行业的需求激增，其订单管线已经排满到 2028 年”
英文原文
We're now in the photonics supercycle. The trio from $COHR, $LITE, $AAOI in America are expected to perform well from a surge in demand. New Report: " Lumentum is experiencing a surge in demand from the artificial intelligence sector, with its order pipeline now fully booked through 2028 "
-
认为变压器 / 开关设备是 DC 瓶颈，需求和 backlog 都很强。
遗憾的是，我的票并不是每周都能翻三位数。 $HPS.A 目前也就涨了大约 17%。 Hammond 是我在变压器 / 开关设备这个 DC 瓶颈上的选择，因为它市值更低，但在干式变压器市场的份额很高。 $POWL 也是我经常遇到的另一只不错的开关设备多头；我最后没有买，但还是顺手提一下。 变压器 / 开关设备大概被低估了，作为瓶颈很重要，但很多型号未来 2 到 5 年都已经卖光了。 也正因为如此，它们正在造成大范围的 DC 延迟，而从 backlog 可见度和极端需求来看，这些标的应该会是不错的结构性多头。
英文原文
Unfortunately, not all my names go up triple digits in a week. $HPS.A is only up ~17% so far. Hammond was my pick for the transformers/switchgear DC bottleneck since it had a lower MC but high dry transformer market share. $POWL was another good switchgear long that I came across often; I ended up passing but just throwing that out there. Transformers/Switchgear are probably under-appreciated as a bottleneck but many types are sold out for the next 2-5 years. Because of this, they’re causing widespread DC delays and should be a good structural long from backlog visibility + extreme demand.
-
亚马逊因Anthropic模型基于Trainium训练消息反弹，带动MRVL、AAOI等相关股
顺便解释一下你们可能想知道的$AMZN今天为何反弹：新消息显示Anthropic的Mythos及其最新模型很可能是基于Amazon Trainium训练的。这对亚马逊生态系统来说意义重大，从$MRVL到$AAOI都受益于Anthropic。就像今年早些时候谷歌TPU生态因Gemini而反弹，包括$LITE一样。
英文原文
Just in case you’re wondering why $AMZN rallied today: New news that Anthropic Mythos and their latest models were likely trained on Amazon Trainium. This is big for Amazon’s ecosystem from $MRVL to $AAOI due to Anthropic. Just like how Google’s TPU ecosystem rallied earlier this year like $LITE from Gemini.
-
把 VNP 视为 AXT 供应链替代方案。
@lamottown 我喜欢把 $VNP 当成 AXT 供应链的替代标的，这个名单只限美国，而 5n 是加拿大上市公司。
英文原文
@lamottown I like $VNP as the AXT supply chain alternative, this is a US only list + 5n is a Canadian listed company.
-
解释 ALRIB 与微软量子相关，但市场目前可能还没给出合理定价。
@MarginCallMick 这只是围绕 $MSFT Quantum 和 $ALRIB 的新信息发现。市场怎么给它们和超大规模客户的关联定价，我也控制不了。 现在很多事情本来就是算法驱动的。
英文原文
@MarginCallMick This is just new information discovery around $MSFT Quantum and $ALRIB. Can't control how market prices in that connection to a hyperscaler. A lot of things are just algorithmically driven nowadays.
-
澄清新标的是英伟达 CPO 生态相关的量子跟进。
@devandraan 不是的，这个新名字和 $NVDA 的 CPO 生态有关。这只是关于一个量子标的的后续跟进。
英文原文
@devandraan It's not, new name is related to $NVDA CPO ecosystem. This is just a follow-up regarding a quantum name.
-
认为微软量子可能在买 Riber，其量子设备业务很有看点。
$MSFT 看起来像是在秘密买入 Riber < $ALRIB >，用于 Microsoft Quantum。 更广泛的市场和算法大概还不知道这种关联。 MBE 对量子计算、量子点激光器、VCSEL 和硅光子都很重要。 但如果 Microsoft Quantum： -> 开始向这家欧洲设备制造商采购 那长期影响会很大。 量子仍然非常早期，早期产业的风险也很实在。 但如果这家公司看起来像量子领域的 $AIXA，而且作为和 $VECO 并列的 MBE 双寡头之一，前瞻市盈率大约 27 倍并且还在盈利，那它就挺有意思的。 从公开的 LinkedIn 线索来看，美国超大规模客户正在使用 Riber 的 MBE 设备来制造前沿量子芯片。 至少这能验证这家公司作为一项技术的重要性。 而且也让它看起来更有机会在未来拿到量子计算和量子点资本开支周期。
英文原文
$MSFT looks like a secret buyer of Riber < $ALRIB > for Microsoft Quantum. The broader market and algorithms likely don’t know this connection yet. MBEs are important for quantum computing, quantum dot lasers, VCSELs, and silicon photonics. But the long-term implications are sizable if Microsoft Quantum: -> Are starting to buy from this European equipment manufacturer Quantum is still very early and risks to early stage industries are material. But since this company looks like an $AIXA for Quantum and trades at a profitable ~27x est. forward P/E, as a duopoly for MBE with $VECO, it looks interesting. From open LinkedIn intelligence, having a US hyperscaler use Riber MBE equipment to fab frontier quantum chips. At the very least helps validate the company’s importance as a technology. And makes it appear well positioned to capture quantum computing and quantum dot capex cycles in the near future.
-
通过LinkedIn追踪发现$MSFT量子项目的未披露供应商，分析其在量子MBE设备领域的双寡头地位及财务状况。
所以彩蛋来了： $MSFT的Quantum项目似乎是一家未公开披露的超大规模云厂商... 而它秘密地从一家公司采购——根据LinkedIn追踪，这家公司的名称听起来像是和"Prime Ribs"有关。 -> Microsoft Quantum发布了照片 -> 放大左上角并放大 -> Prime Ribs机器验证确认？ 在量子机器的MBE（分子束外延）领域，他们与$VECO形成了双头垄断（可以类比$AXIA在光子学领域的地位，但这是针对量子领域）。 他们今天发布了财报： -> 已经盈利。财报发布后约25倍PE。 -> 资产负债表非常健康 -> 用于周边领域，比如硅光子学（例如Fujitsu的量子点分拆公司） -> IntelliEPI（除$MSFT外）是知名的买家 一些粉丝主动联系我了解这家公司的名字（DD发现功劳不归我）。免责声明：我认为他们的分析很有启发性，并且持有仓位。 -> 全部功劳归于@latent_value7发现了$MSFT的关联 -> 功劳归于@TheSarge_很早就让我注意到了这个名字 提醒一下：尽管我分享这些精彩发现，我不建议任何人根据我的帖子建仓。 我只是很惊讶，有人在X平台上发现了一家超大规模云厂商从前沿量子项目的小型供应商那里秘密采购。
英文原文
So the easter egg: $MSFT Quantum appears to be an undisclosed hyperscaler... That is secretly buying from a company that sounds with Prime Ribs according to LinkedIn tracking. -> Microsoft Quantum posted photos -> Zoom in to top left and magnified -> Prime Ribs machine checks out? It's a duopoly in the MBE space with $VECO for quantum machines (think $AXIA for photonics but for Quantum). They reported earnings today: -> It's profitable. ~25 P/E ratio after ER release. -> Very healthy balance sheet -> Used for adjacents, like silicon photonics (eg. Fujitsu's quantum dot spinoff) -> IntelliEPI (aside from $MSFT) well known buyer Some followers reached out about this name (not taking any credit for DD or finding it). Disclosure: I found their ideas interesting and have positions. -> Full Credit to @latent_value7 finding $MSFT links -> Credit to @TheSarge_ for bringing the name to my attention early on A reminder: Even though I share these cool ideas, I don't recommend anyone take positions based on my posts. I'm just impressed that people on X found that a hyperscaler buying from a tiny company for their frontier quantum program.
-
认为 SIVE 在 CW 激光赛道被低估，且随着 CPO 和光学扩产，估值有望抬升。
当你把从 $COHR 到 $AAOI 的 CW 激光龙头公司名单都过一遍时， 就会发现 $SIVE 以 3 亿美元市值还站在那里，真的很夸张……其他玩家几乎都已经是几十亿市值了。 而它还是 $JBL 1.6T LRO 光收发器和大约 $MRVL Celestial 光子项目的光源供应商（如今还受到 $NVDA 20 亿美元投资的加速）。 从 Furukawa 到 Yuanjie，整个行业最近都被重新定价了。 但我确实认为市场漏掉了这只，因为他们预计从 H2 开始量产爬坡（来自 $POET 的业绩提示），而 CPO 和光学扩产会带来多年指数级 TAM 扩张。 下行风险在于资产负债表，但我认为他们的光子业务实体未来大概率会在纳斯达克上市，而且作为超大规模客户的光源供应商，价值远高于当前估值。 而且这很可能会在不久后带来估值溢价。 我们看看结果是不是这样吧。
英文原文
When you go through the list of the leading CW laser companies from $COHR to $AAOI. It’s very impressive $SIVE stands out at $300m still… compared to the every other player in the billions. This is despite being the light source for $JBL 1.6T LRO optical transceivers and ~ $MRVL Celestial photonics program (now accelerated by $NVDA $2B investment). The sector from Furukawa to Yuanjie has been re-rated recently. But I do think markets missed this one, as they start volume ramp est. H2 (from $POET earnings) with multi year exponential TAM expansion from CPO and optical scale up. Downside risk is balance sheet, but I do think the likely Nasdaq listing for their photonics entity + scaling as the hyperscaler light source far exceeds current valuations. And will likely drive valuation premiums in a matter of time. We’ll see if this turns out right or not.
-
通过公司实地拜访和线索，推测微软量子与光学供应链有关。
我去总部问他们关于 $NVDA GPU 集群和 $LITE 光学扩展瓶颈的问题。 我让他们给我看看新的 Mythos 模型，结果他们给我看了一间工厂。 很明显 OpenAI 和 xAI 的人才更强。
英文原文
Walked into the HQ to ask about Claude Opus bottlenecks with $NVDA GPU clusters and $LITE optical scale out. I asked to see the new Mythos models, they showed me a plant. It’s clear OpenAi and xAI has better talent. https://t.co/LXlyghhdeJ
-
认为 SOI 的 CPO / SiPh 放量需要时间，但下半年会逐步追上。
@Niatama195432 $SOI 的 CPO / SiPh 放量还需要一点时间。自从我最初的 thesis 以来，它也就涨了大约 11% 左右，不过这取决于入场价。 当前收发器周期里像 $AAOI 和 $LITE 这样的公司已经完成了重估，但我预计其他标的会在 H2 追上来。
英文原文
@Niatama195432 $SOI takes a bit of time for CPO/siph ramp. It's only up ~11% or so since my original thesis, but depends on entry point. Lot of the current transceiver cycle players like $AAOI and $LITE have re-rated currently, but I expect the others to catch up H2.
-
Aehr获超大规模客户量产订单确认，从资格认证转入大规模生产阶段。
$AEHR财报： - 当前积压订单：季度末为3870万美元，"有效积压订单"为5090万美元 - FY26下半年指引：维持对下半年2500万至3000万美元收入的预期 对财报的看法： 可以直接忽略当前财报，主要指标看的是超大规模客户在H2的产量爬坡。 产量爬坡信号确认： 1. "我们看到主要超大规模客户对我们的Sonoma系统发出了大量预测订单，用于其定制AI处理器ASIC的高产量生产老化测试" 2. "我们预计该客户近期将追加大量系统的后续生产订单，计划在Aehr的2027财年交付" 这正是市场预期的$AEHR从资格认证转向大规模量产的过程，类似$AAOI，我们得到了这个确认。 H2可能是进一步确认某大客户产量爬坡将延续到2027年。
英文原文
$AEHR earnings: - Current Backlog: $38.7 million as of the end of the quarter, with an "effective backlog" of $50.9 million -H2 FY26 Guidance: Reiterated expectations for $25 million to $30 million in revenue for the second half of the year Thoughts on earnings: Can just ignore current earnings, main indicators are around hyperscaler volume ramp H2. Volume ramp indication confirmations: 1. "We are seeing significant forecasts from our lead hyperscale customer for our Sonoma systems for high-volume production burn-in of their custom AI processor ASICs" 2. "We expect a significant near-term follow-on production order from this customer for a large number of systems to be shipped during Aehr's fiscal year 2027" This is what is expected with $AEHR transitioning from qualification to mass volume like $AAOI and we got that confirmation. H2 is probably more confirmation around siph customer volume ramp into 2027.
-
超大规模建设延迟溢出效应利好已锁定产能的Neocloud公司如$NBIS。
英伟达($NVDA)的黄仁勋说： “$NBIS会照顾好你的。” （引用内容）我对上个月这份报告思考得越来越多…… 它对Neocloud板块比如$NBIS（甚至$CRWV）可能比市场预期的要利好得多。 因为超大规模云服务商建设延迟的溢出效应。 这也在一定程度上解释了为什么$META签署了270亿美元的协议。 那些已经为其建设项目锁定产能/组件订单的公司，可能会成为大赢家，因为其他竞争对手停滞不前。
英文原文
Jensen Huang from $NVDA: “ $NBIS will take care of you. “ https://t.co/lIRrnAC3fc
-
给出光子学 CW 激光瓶颈的个人高 beta 排序。
下一轮光子学 CW 激光瓶颈： 我个人的高 beta 敞口排序： 1. $SIVE：3.02 亿美元 2. $AAOI：83.5 亿美元 3. Yuanjie（688498）：135.5 亿美元 4. $MTSI：174 亿美元 5. $LITE：561 亿美元 6. $COHR：492 亿美元 7. Suzhou Everbright：59.5 亿美元 8. LuxNet（4979.TWO）：17 亿美元 9. Henan Shijia（SHA: 688313）：62 亿美元 10. Furukawa Electric（TYO: 5801）：160.6 亿美元 11. Sumitomo Electric（TYO: 5802）：451.13 亿美元 12. Mitsubishi Electric（TYO: 6503）：712 亿美元 13. $AVGO：1.53 万亿美元 最后我自己拿了最高敞口的几个（但因为利益冲突，避开了港股标的）。
英文原文
For the next photonics CW laser chokepoint. Personal high-beta exposure tierlist: 1. $SIVE: $302m 2. $AAOI: $8.35B 3. Yuanjie (688498): $13.55B 4. $MTSI: $17.4B 5. $LITE: $56.1B 6. $COHR: $49.2B 7. Suzhou Everbright: $5.95B 8. LuxNet (4979.TWO): $1.7B 9. Henan Shijia (SHA: 688313): $6.2B 10. Furukawa Electric (TYO: 5801): $16.06B 11. Sumitomo Electric (TYO: 5802): $45.113B 12. Mitsubishi Electric (TYO: 6503): $71.2B 13. $AVGO: $1.53T Ended up taking the highest exposure picks personally (but avoided HK listed names due to conflict of interest).