概念词典 · 技术
ASIC 专用集成电路
为特定算法/任务定制的芯片，AI 时代主要指各大科技公司绕开 NVIDIA GPU 自研的训练/推理加速器。
别名： ASIC／ 专用集成电路／ 专用芯片
是什么
Application-Specific Integrated Circuit (ASIC) 是为特定算法或任务定制的芯片，区别于通用 CPU 与 GPU。在 AI 语境下，ASIC 几乎等同于 大厂自研 AI 加速器，目的是绕开 NVIDIA CUDA 生态、降低单算力成本。
代表 ASIC 项目：
| 项目 | 谁的 | 设计代工 |
|---|---|---|
| TPU | $AVGO Broadcom | |
| MTIA | Meta | $AVGO Broadcom（部分） |
| Trainium / Inferentia | Amazon | $MRVL Marvell |
| Maia | Microsoft | $AVGO |
为什么 ASIC 是博通 / Marvell 的现金牛
设计一颗高端 ASIC 需要：
- 高端 SerDes / HBM PHY IP
- 7nm/5nm/3nm 工艺经验
- 与 TSMC 的产能关系
这些都是 $AVGO 和 $MRVL 在传统通信芯片时代积累的。所以 Google / Meta / Amazon / Microsoft 自研 AI 芯片实际是 “外包给 Broadcom 设计”，“自研”叙事在财务上呈现为 Broadcom 的 ASIC 业务收入井喷。
Serenity 的关键判断
他对 ASIC 板块的核心论点：
ASIC 时代到来的速度取决于 TPU 商用化进度
如果 TPU 仅限 Google 内部使用，AVGO 的 ASIC 业务有天花板；如果 TPU 像 GPU 那样卖给所有人，整个 ASIC 板块定价权迁移，进入 NVIDIA 之外的”第二极”。
代表标的：
- $AVGO Broadcom：ASIC 龙头，受益于 Google + Meta + Microsoft 三家订单
- $MRVL Marvell：Amazon Trainium 主力设计代工
- $ALAB Astera Labs：ASIC 配套 retimer + smart cable
提到此概念的推文 (107)
按时间倒序 · 滚到底自动加载更多
-
博主回顾过往观点遭质疑后获市场验证的经历，强调市场是最终裁判。
我想你一定是新来的。我的许多观点起初都会遭遇强烈反对，尤其是那些更具原创性的观点。$AXTI 遭受无尽谩骂，甚至导致我被 $RDDT WSB 论坛禁言。人们以为它是家骗人的中国公司，但路透社、Epiwafer 财报及机构验证了其磷化铟(InP)衬底地位。$RPI 被所有人称为“梗股”。彭博、金融时报等媒体称其无基本面。分析师称其想法愚蠢且将崩盘。但财报显示其前瞻营收增长58%。他们似乎忘了之前的反对，现在却将其视为高增长AI硬件公司。$SIVE 也被叫作“梗股”，我遭到瑞典媒体大量攻击。人们不懂技术细节，莫名进行人身攻击。但如你所知，这可能是我最成功的观点，已获富达研究、摩根大通机构买入及 $JBL 至 $GFS 的正式合作验证。同样的仇恨也发生在：$AAOI $30时，管理层被称“骗局”；$LITE $300时，光子学被称“泡沫”；$RKLB $20时，被视为低营收泡沫公司，我因发帖被WSB临时禁言；$HOOD $20时，人们想起其冻结GME交易；$IQE $12时，被视为英国无光子学合作的垃圾公司；$SOI $44时，欧洲分析师称其“高估”；$NBIS $75时，$IREN 阵营称我散布俄罗斯宣传且无护城河；$INTC $115时，人们认为其无法与 $TSM 竞争；$MRVL $85时，人们认为其正失去ASIC份额给博通；$AEHR $35时，人们误读财报认为其无营收；$EWY $115时，人们哭诉KOSPI是泡沫且LNG/氦气/石油将颠覆存储贸易。例子不胜枚举……我常看评论，所以记得很多仇恨（我猜是嫉妒、博眼球或缺乏技术深度）。但现在，人们可以看到每个论点被反复验证。市场成功淹没了旧噪音。好消息是，市场是判断对错的最终仲裁者，而非X上的愤怒评论。
英文原文
I think u must be new here. Many of my ideas get intense backlash at the start, especially the more original they are. $AXTI - endless hate to the point I got banned from the $RDDT WSB forum. They thought it was some scam Chinese company, but Reuters, Epiwafer company earnings, and institutions validated their InP substrate position many months later. $RPI - everyone called it some meme stock. Literally Bloomberg, Financial Times, and others went out and called it a meme stock with no fundamentals. Analysts went and said the idea was stupid and said it was going to crash “as a fact”. Earnings came out? Blew away any projection with 58% fwd revenue growth. But they seem to have forgot all the backlash they threw out at me, while they’re citing it as a high growth ai hardware company no. $SIVE? Everyone called it a “meme stock”. I get bunch of hate from Swedish media all the way up. Bunch of people didn’t understand the technical nuances, and they keep throwing personal attacks for some strange reason. But as you know it’s probably my most successful idea and it’s validated so far by institutional buying from Fidelity Research, JP Morgan as well as formally announced partnerships from $JBL to $GFS . Same hate with: - $AAOI at $30, when everyone called management a “scam” or “shady” - $LITE at $300 when everyone called photonics a “bubble” - $RKLB at $20 when everyone thought it was a low revenue launch company that was a bubble. I actually got temp banned from WSB from posting about Rocketlab since moderators didn’t like the stock. - $HOOD at $20 when everyone recalled them freezing GME buys/sells - $IQE at $12 when people thought it was just some random crap $100m company over in the UK with “no actual photonics partnerships” - $SOI at $44 when European bank analysts thought it was “overvalued”and my thesis wasn’t anything new - $NBIS at $75 when everyone in the $IREN camp said I was spreading Russian propaganda and that they had no moat - $INTC at $115 when everyone thought they couldn’t compete with $TSM - $MRVL at $85 when everyone thought they were losing ASIC share to Broadcom. - $AEHR at $35 when everyone misread their earnings and thought they had no revenue - $EWY at $115 when everyone was crying KOSPI was a bubble and LNG/helium/oil would disrupt the memory trade Can go on and on… I do read a lot of the comments, which is why I remember a lot of the hate (probably either jealously, impression farming, or lacking the technical depth is my guess). But I think at this point, people can just see each thesis validated over and over again. And the success in markets drown out the old noise. Good thing is markets are the final arbiter of what’s right or wrong, not the angry comments or posts on X.
-
回顾2025年对光模块及半导体供应链龙头的早期估值判断。
有趣地回顾一下2025年的一些随机想法。 当时， $AAOI 市值为20亿美元， $LITE 市值为260亿美元， $AXTI 市值为5亿美元。 现在： - AAOI 市值为153.7亿美元 - Lumentum 市值为744.7亿美元 - AXT 市值为72.4亿美元 当时由于处于早期主题阶段，掌握的信息较少。 当然，专用集成电路(ASIC)方面存在细微差别。
英文原文
Fun throwback to random ideas back in 2025. Back then, $AAOI was $2B MC, $LITE was a $26B MC, $AXTI was $500M Now: - AAOI is $15.37B - Lumentum is $74.47B - AXT is $7.24B Was working off less information back then, given it was an early theme. Obviously nuances with ASIC
-
通过供应商关系映射推测AI芯片供应链瓶颈。
不客气，关于 $IQE。 -> $LITE 是 $GOOGL 的已知供应商。 -> $IQE 是 $LITE 的已知供应商。 -> $AXTI 是 $IQE 的已知供应商。 Google TPU + 超大规模云服务商 ASIC 的物料清单(BOM) 极度保密。 但你可以通过关系映射大致推测，尽管没人会明说 $AXTI 或 $IQE 供货给...
英文原文
Ur welcome with $IQE
-
长期更偏好英伟达上游瓶颈而非NVDA本身。
回复 @darkseidzz：嗯，长期看，相比做多 $NVDA，我更喜欢你所有的上游瓶颈环节，因为这些会被重估得最多（英伟达已经是世界最大公司）。 我很确定，超大云厂商 ASIC 最终会分流 $NVDA 的需求，比如 $GOOGL 的 TPU、$AMZN 的 Trainium 项目。
英文原文
@darkseidzz hmm, i prefer all your upstream chokepoints over $NVDA long term since those will be re-rated the most (nvidia already largest company in the world) pretty sure hyperscaler ASICs would eventually siphon off $NVDA demand like $GOOGL TPU, $AMZN trainium programs. wouldn't be too
-
称SIVE不是垄断，而是在ASIC/Merchant CPO架构路线具技术主导。
回复 @SidkMena：这不是垄断。只是它在 ASIC / merchant CPO 路由架构上拥有技术主导地位。 如果其他可插拔玩家也用 $SIVE 创造新的 1.6T 架构，我也不会惊讶。
英文原文
@SidkMena It’s not a monopoly. Just technical dominance over ASIC/Merchant CPO route architectures . And wouldn’t be surprised if other pluggable players created novel 1.6T architectures using $SIVE too.
-
称SIVE是最近最明显、仍在重估的CPO激光瓶颈。
回复 @_king142：$SIVE 可能是最近最明显、仍在重估中的激光器瓶颈，因为它似乎出现在每个 $NVDA NVLink CPO 生态伙伴以及 merchant/ASIC CPO 供应链里。 也许下一个是 CPO 中的无源光组件供应链。
英文原文
@_king142 $SIVE was probably the most recent visible laser chokepoint that’s still being rerated since seem to be in every $NVDA NVLink CPO ecosystem partner + merchant/asic cpo supply chains. Maybe passive optical components supply chains in CPO next.
-
没预期英伟达今年入股Marvell，但看好其ASIC/连接收入机会。
回复 @WEB3_furture：好文章！我其实没想到 $NVDA 今年会与 $MRVL 合作并入股。 但无论如何，我本来就预计 Marvell 会因为即将到来的 ASIC/连接收入机会而成为一个有吸引力的想法。
英文原文
@WEB3_furture Great article! I actually didn't expect $NVDA to partner and take a stake in $MRVL this year. But I expected Marvell to be a compelling idea anyway from their upcoming ASIC/connectivity revenue opportunites.
-
说明讨论的是ASIC/merchant NVLink CPO生态伙伴而非英伟达自身。
回复 @siflower：那是 $NVDA 直接相关。我说的是 ASIC/merchant NVLink CPO 生态伙伴。 在这些伙伴里，$SIVE 看起来像是所有人的供应商。
英文原文
@siflower That's $NVDA directly. I'm talking about the ASIC/merchant NVLink CPO ecosystem partners. Where $SIVE looks like they're the supplier to everyone.
-
关注ASIC报告和AlChip通过并购扩展TAM的机会。
我发现一份1月的 ASIC 研究报告挺有意思。 我不太常提 AlChip，但如果他们想像 Marvell 那样做并购，会有很多 TAM 扩张机会。
英文原文
ASIC research report from Jan that I found interesting. I don’t really mention AlChip much but there’s a lot of tam expansion opportunities if they want to M&A like Marvell. https://t.co/oaSMBgO72C
-
回顾AAOI/AXTI早期被FUD，如今类似CPO转折正在SIVE/Foci上重演。
当我在 $AAOI 约20-30美元做多时： 我以为 $AMZN 和 $MSFT 正在为其 ASIC 项目认证特定光收发器。 结果它更可互换、更可大规模生产。 无论如何，很高兴我关于 $LITE、$AAOI、Innolight、$COHR、$AXTI 的 thesis 表现这么好。 记住：当 $AAOI 在20-30美元时，X 上所有人都说“管理层诈骗”…… 或者在 $AXTI 12美元时说“诈骗公司”。 FUD 非常极端。 现在感觉又像似曾相识……做多下一轮 CPO 架构转变，比如 $SIVE 或 Foci？然后收到同样评论。 我们会看看我的 CPO 多头是否会像 Shunsin 对 Innolight、或 $SIVE 对 $LITE 那样兑现。 无论如何，那些名字现在都已经上涨数百到数千个百分点。
英文原文
Back when $AAOI was ~$20-30 when I went long: I thought $AMZN and $MSFT were qualifying specific optical transceivers for their ASIC programs. Turns out it’s more interchangeable/mass producible. Regardless, glad my thesis on $LITE, $AAOI, Innolight, $COHR, $AXTI played out so well. Keep in mind: everyone on X was saying “scam management” with $AAOI back at $20-30… Or “scam company” with $AXTI at $12. FUD was pretty extreme. Feels like dejavu again… going long on the next CPO architectural shift like $SIVE or Foci? And getting the same comments. We’ll see if my CPO longs play out the same way like Shunsin does with Innolight or $SIVE does with $LITE. Either way all of those are now up hundreds to thousands of percent.
-
称SIVE是CPO kingmaker，围绕Lightmatter/Celestial/Ayar/O-Net/Jabil等供应激光。
$SIVE 字面上是 CPO 的 kingmaker： 从早期的 Lightmatter、Celestial、Ayar、Lightelligence，到现在这些估值50-100亿美元以上的领导者。 再到 Marvell、Alchip、GUC 这样的 ASIC 生态。 O-Net 与 Sivers 为 CPO 量产 ELS。 Jabil 与 Sivers 为可插拔光收发器量产可插拔产品。 所有这些很可能都围绕 Sivers 激光，并以它为设计核心。 市场才刚刚开始看到这家激光公司有多重要。
英文原文
$SIVE is literally the kingmaker for CPO: From ~ Lightmatter, Celestial, Ayar, Lightelligence early on. Now the leaders with massive $5-10B+ valuations. To the ASIC ecosystems like Marvell, Alchip, GUC. With O-Net mass producing ELS with Sivers for CPO. And Jabil mass producing pluggables with Sivers for pluggable optical transceivers. All likely surround and are designed around Sivers lasers. Markets are only starting to see the precipice of how important this laser company is.
-
梳理玻璃基板光刻胶、ABF瓶颈和关键矿物供应链。
一些随机笔记： 关于玻璃基板：“YC Chem 据称成为业内首家供应玻璃基板光刻胶的公司”（KR 112290）。 - 目标年底量产，通常量产 + 首发对股价有利。 “目前正与三家以上公司洽谈玻璃基板材料供应”，所以还有更多认证。 “Samyang NC Chem 也在开发玻璃基板光刻胶材料”（482630）。 - 正在送样，明年量产，是另一个玻璃基板名字。 “$NVDA 可能通过长期协议、预付款或战略合作提前锁定高端基板产能，目标是避免 CoWoS 和 HBM 过去出现的供应限制重演。”这与 $TSM COUPE 基板有关。 “Commercial Times 强调，AI GPU、ASIC 和 CPO 需求可能在2027年前触发另一轮高端 ABF 基板短缺。” 所以还有更多基板瓶颈。 另外还有：“中国同意解决美国对稀土和关键矿物短缺的担忧，特别是钇、钪、钕和铟。” 如果你想找多头想法，可以研究钇、钪、钕和铟供应链。 现在一切都在回调，但如果你对更多随机想法感兴趣，这是值得关注的东西。
英文原文
Just some random notes: For glass substrates: "YC Chem has reportedly become the first in the industry to supply photoresists for glass substrates" (KR 112290) - targeting EOY mass production, usually mass production + first to market is good for stock prices. "is currently in discussions with more than three companies regarding the supply of glass substrate materials" so more qualifications. "Samyang NC Chem is also developing photoresist materials for glass substrates" (482630) - Sampling, mass production next year another glass substrate name " $NVDA may seek to secure high-end substrate capacity in advance through long-term agreements, prepayments, or strategic partnerships, aiming to avoid a repeat of the supply constraints previously seen in CoWoS and HBM." This is in relation to $TSM COUPE on Substrate. "As Commercial Times highlights, demand from AI GPUs, ASICs, and CPO could trigger another supply shortage of high-end ABF substrates by 2027." So more substrate bottleneck stuff. Then there's: "China's agreement to address US concerns over rare earth and critical mineral shortages specifically yttrium, scandium, neodymium and indium" So good to look into yttrium, scandium, neodymium and indium supply chains if you want long ideas. Everything is kinda correcting right now, but just something to keep a lookout for if ur interested in more random ideas.
-
用 napkin math 算 Nextronics 的 CPO 价值
所以这是我在做多 Nextronics（8147）时做的便签纸计算。 他们是 $NVDA 的 CPO 供应商，供应 CPO connectors 和 cage thermal modules。 我按 2028 年的前瞻市盈率大概算了 2 倍，这也是我认为它有望在 2028 年重估到约 20 亿美元以上市值、具有很强风险回报的原因。 只看它的 CPO 暴露： -> CPO connector 大概 15 到 25 美元 -> ELS thermal cages 估计约 50 美元 每个 switch 18 个单元：18x50 = 900 美元 CPO Connectors：每个 switch 72 个 optical engines，72 x 15 = 1080 美元 （如果 $NVDA 扩到 Spectrum-X switch，这个数会到 1920 美元） Total Nextronics Content：大约 1980 美元（保守算成 2000 美元） 这意味着 BOM 只占 rack 的 0.08%，大概只占 switch 的 1.5%。 机构眼里这太小了，所以大概率被忽略。 但对 Nextronics 这家约 2 亿美元的公司来说，这很重要吗？ 当然，绝对很重要。 我算的时候把 Nextronics 在 Nvidia connector/cage 市场的份额打了 50% 折扣，因为存在多供应商。 并且我用的是 GS 的预测，再假设 $AVGO、$MRVL、ASIC CPO 生态的规模只有 $NVDA 的 30%。 净利润率：22.4%（GM 38% 时）到 24.0%（GM 40% 时）。 但如果按其他预测，只看 rack shipment： 2026：CPO revenue 约 1010 万，净利润（22.4%）约 226 万 + 基础业务 1250 万 = 1470 万（540k 单元是 connectors，40k 单元是 cage，且已除以 50%） 2027：CPO revenue 约 1.72 亿，净利润（22.4%）约 3853 万 = 5103 万（约 800 万单元 connectors，103 万单元 cage） 2028 scale up 扩张：CPO revenue 约 4.5 亿，净利润约 1.0093 亿，基础业务约 1130 万（约 4000 万单元 connectors，298 万单元 cages，例如 Nvidia ELS 量是 1990 万） 所以前瞻市盈率大概是 2026 年 15.4 倍，2027 年 4.45 倍，2028 年 2 倍。 当然，放量后综合毛利可能会下降，也可能会有别的玩家把份额稀释到 25% 左右，预测也可能比 GS 更高或更低。 但不管怎样，哪怕我偏差 50%，这仍然极具不对称性。 2028 年通常是 CPO 玩家大重估的一年，2026 年其实还很早。 希望我算得对，但 20x 的前瞻市盈率会对应 22.6 亿美元市值。 即便我们把： -> 市占率降到 15%。 -> 净利润率压到 14%。 -> connector ASP 降到 10 美元。 按 20x 倍数，股价仍然能有大约 4.5 倍回报，市值超过 10 亿美元。 我们走着看吧。（NFA，只是投机性财务建模）
英文原文
So here's the napkin math I did on Nextronics (8147) when I went long. They're the $NVDA CPO supplier for CPO connectors and cage thermal modules. And I modeled around 2 FWD p/e for 2028, which is why I think risk-reward is very compelling for a potential 10x rerating to ~$2B+ MC in 2028. Just for their CPO exposure: -> CPO connector runs roughly $15 to $25 -> ELS thermal cages, maybe ~$50 from est. 18 units per switch: 18x50 = ~$900 CPO Connectors: 72 Optical Engines per switch 72 x $15 = $1,080 (If $NVDA scales their Spectrum-X switch, it goes to $1,920 for CPO connectors). Total Nextronics Content: ~$1,980 (rounded to $2k for calculations) in conservative case. Implied BOM % of rack: 0.08%. Maybe ~1.5% of switch. This looks microscopic to institutions so it probably is ignored. Is it material to Nextronics, a ~$200m company? Yes, absolute massive. For calculations: Applying 50% haircut to Nextronics' share of the Nvidia connector market/cage market because of multi-source. And I’m using GS projections, and assuming $AVGO, $MRVL, ASIC CPO ecosystem is 30% size of $NVDA. Net Income Margin: 22.4% (at 38% GM)- 24.0% (at 40% GM). But going off other projections from just, a rack shipments: 2026: CPO revenue ~10.1M, net income (22.4%) ~2.26M + $12.5M base = $14.7M (540k units for connectors, cage, 40K units, already divided by 50%) 2027: CPO revenue: ~$172M, net income (22.4%): ~$38.53M = $51.03M (~8M units for connectors, ~1.03M units for cage) 2028 scale up expansion: CPO revenue: $450m, net income: $100.93M, ~$11.3M base (~40M units for connectors, 2.98M unit for cages, eg. Nvidia ELS volume is 19.9M) So implied fwd p/e 15.4x for 2026, 4.45x for 2027, 2x for 2028. Of course at scale, blended margins might go down, there might be other players bringing market share down to like 25%, etc. and projections might be more or less than GS. But regardless seems highly asymmetrical even if I'm off by a whole 50%. 2028 is usually the massive re-rating for CPO players, 2026 is still really early. Hope my math is right, but 20x fwd p/e multiple would be $2.26B MC. Even if we drop: -> market share to just 15%. -> compress their net income margin down to 14%. -> connector ASP to $10. At a 20x multiple, the stock would still achieve a ~4.5x return to a $1B+ market cap. We'll see if this is right or not. (NFA, just speculative financial modeling)
-
SIVE 业务很分散
哦，$SIVE 肯定没问题，因为它们已经高度多元化了。 它们大概基本上服务所有人： -> $AAPL 硅光 -> $JBL（可能最大） -> $AMD CPO 项目配合 $GFS， -> Ayar -> Alchip/GUC 和其他 ASIC 设计商 -> O-Net ELS 还有其他客户。虽然这会在实质上让 2028-2029 年后期的预测稍微往后推一些（比如 Q4 5 亿美元、2029 年 10 亿美元）。 但如果它们的光源最终以某种方式进了 Marvell，我也不会惊讶。
英文原文
Oh $SIVE is definitely fine, since they're vastly diversified. They likely basically serve everyone: -> $AAPL Silicon Photonics -> $JBL (probably largest) -> $AMD CPO program with $GFS, -> Ayar -> Alchip/GUC and other ASIC designers -> O-Net ELS amongst others. Though this does materially slow later 2028-2029 projections a bit (eg. $500m Q4 Celestial, $1B 2029). But wouldn't be surprised if their light source ends up in Marvell one way or another.
-
CPU 开始像存储一样短缺
CPU “严重短缺” 现在看着 CPU 一天天变得越来越像存储，挺好笑的，像 $SNDK 那样不断涨价。 “自 3 月以来，服务器 CPU 已上涨 10% 到 20%” $INTC 和 $AMD 计划在 Q3 继续涨价。 “CPU 供应很可能在 2026 和 2027 年继续紧张”，其中“Intel 的服务器 CPU 短缺可能长达 6 个月，AMD 则为 8 到 12 周”。 $TSM 也在因这波短缺而增加 capex。 “恐慌的主要原因是 CPU 和 AI ASIC 需求同时爆发；目前库存包括 Intel、AMD 主流代际 CPU，以及 $NVDA 即将推出的 Vera CPU，全部都在 3nm 工艺上。” 看起来我们又绕回来了？
英文原文
CPU "serious supply shortage" It's still funny to see CPUs become more like memory day-by-day with the price hikes like $SNDK. "Server CPUs have jumped between 10% and 20% since March" with $INTC and $AMD planning to further price hikes Q3. "CPU supply is likely to remain tight through 2026 and 2027" with "server CPU shortages reaching up to six months for Intel and 8-12 weeks for AMD" $TSM is also increasing capex in response to this shortage. "The main reason behind the fear is the simultaneous outbreak of CPU and AI ASIC demand; inventory currently includes Intel, AMD mainstream generation CPU, and $NVDA's upcoming Vera CPU, all using 3nm process Looks like we've come full circle?
-
AMD 的 CPO 供应链
哇……如果 AMD 走 Ayar 路线，并在 $GFS 上推进，那么 Sivers 看起来就是 $AMD CPO 项目的主要激光供应商。 有意思的是：$LITE 和 $MTSI 曾经被悄悄地、按顺序从 Ayar 的网站上移除了。 （特别感谢我的一位关注者 Setian 发来私信。） 所以 $SIVE 很可能先成为 Ayar 一代产品的主要激光供应商……进而成为经由 Alchip 或其他 ASIC 设计公司流转的所有 Ayar 客户的供应商。 Ayar 上个月还为了量产融资 5 亿美元，而 $SIVE 已经被设计进去。 这个发现对 $SIVE 来说结构上非常重大，因为它削弱了 $LITE 等公司作为主要供应商的叙事。 尤其也削弱了市场对 $SIVE 在 CPO 里只占很小激光份额的说法，如果它们很可能就是主要激光供应商的话。
英文原文
Woah... Sivers looks like the primary laser supplier to $AMD's CPO program if AMD goes the Ayar route at $GFS. The interesting thing is: $LITE and $MTSI were silently removed from Ayar's website sequentially over time. (Special thanks to one of my followers Setian for the DM.) So $SIVE likely became the primary laser supplier for Ayar first-gen... and by extension for all of Ayar's customers routing through Alchip or other ASIC design firms. Ayar also raised $500M last month for VOLUME PRODUCTION, where $SIVE is designed in. This find looks structurally massive for $SIVE as it undercuts narratives about $LITE and others being primary suppliers. And especially about $SIVE having only a small % laser share for CPO if they're likely to be the primary laser supplier.
-
主题框架
CPO / 1.6T。 基板（玻璃核心、InP、SoI）。 CPU / ASIC。 变压器。
英文原文
CPO / 1.6T. Substrates (Glass Core, InP, SoI). CPUs / ASICs. Transformers.
-
GOOGL ASIC 新闻对 SIVE 不够计价
我很惊讶 $SIVE 没有因为 $GOOGL 和 Marvell 的 ASIC 新闻而涨得更高。 甚至连 $MRVL 隔夜都涨了 6.24%+。 市场大概还不知道上游激光源到底来自哪里，对吧？ https://t.co/0Hgfo0dQLo
英文原文
Surprised $SIVE isn’t up higher from the $GOOGL ASIC news with Marvell. Even $MRVL is up 6.24%+ overnight. Markets still probably don’t know where the upstream laser source comes from yet? https://t.co/0Hgfo0dQLo
-
Marvell与Google洽谈TPU开发，Sivers有望成为光子光源供应商
$MRVL正在洽谈为$GOOGL开发TPU（张量处理单元），涵盖两种不同的芯片型号。 如果谈判顺利，$SIVE很可能成为Google供应链的光源供应商。 尤其因为其定制ASIC（专用集成电路）可能推动Celestial IP在MPU（微处理器单元）架构中实现规模扩展（光子结构，Photonic Fabric）。这与$POET和Sivers两家公司的布局都相吻合。 定制ASIC需要数年时间，因此来自Google的实质性营收可能要等到2028年下半年。 供应商可能采用多元采购策略，但拥有Google作为潜在终端客户是一个超越2028年的结构性多年催化剂。 市场通常具有前瞻性。
英文原文
$MRVL in talks to develop $GOOGL TPU for two different chip models. High possibility $SIVE likely became the light source for Google's supply chains if talks go well. Especially as their custom ASIC likely pushes Celestial IP for scale up (Photonic Fabric) in the MPU's architecture. And this maps to both $POET and Sivers. Custom ASICs take years, so there's probably no material revenue from Google until H2 2028. Vendors are also probably multi-sourced, but having Google as a likely end user is a structural multi-year catalyst past 2028. Markets are typically forward looking.
-
认为 Enplas 仍有双重瓶颈，值得在中小市值里关注。
@Letterkees 我还是认为 Enplas 会在 6 个月内翻倍。 通常你不会看到服务 hyperscaler 的 <10 亿美元公司长期还停留在那个市值区间，而且它们在 photonics / ASIC-GPU 测试上握着两个不同的瓶颈。
英文原文
@Letterkees I still think Enplas doubles within 6 months. Normally you don't see <$1B companies serving to hyperscalers stay in that MC range for too long and they own two different chokepoints in photonics/ASIC-GPU testing.
-
认为 Enplas 的双瓶颈与客户结构很有意思。
Enplas（6961）在大约 9.85 亿美元市值时看起来挺有意思。 它们握着两个瓶颈： 1. MLA（微透镜阵列）的主导供应商。比如 SiPH switch、1.6T、3.2T 光子产品。 2. IC 测试 socket 的寡头供应商（AI 芯片测试） 手头现金大约 1.55 亿美元，没有债务。股东权益/资产比大约 89%（一般 50%+ 就算稳），所以我觉得下行风险很低。 而且它们推测的客户是： 1. 光子（MLA）：高度可能是 Innolight、Eoptolink、Furukawa、Intel（SiPh）。大概率还有 $COHR、$LITE。 2. GPU/ASIC（测试 socket）：大概率是 $TSM、ASE，以及这些类型用来测试 $NVDA GPU、$GOOGL TPU（Google Ironwood 很可能就是这个方向）。 “为主要 GPU 制造商以及 hyperscaler 的 ASIC 相关项目扩大量产订单。” 然后在 OFC： OFC： “我们将展示支持当今 AI 快速技术进步的 800Gbps 和 1.6Tbps 收发器以及 CPO（共封装光学）的新产品。” 本质上你得到的是一家同时服务一线半导体、晶圆厂和 hyperscaler 的公司，并且它能同时受益于 MLA 光子段带来的 1.6T，以及之后 OFC 产品带来的 CPO TAM 扩张。 这是我想分享的一笔持仓，因为它在不到 10 亿美元市值时就显得挺酷。
英文原文
Enplas (6961) at ~$985M seems kinda interesting. They hold two chokepoints: 1. Dominant supplier for MLAs (micro lens arrays). Eg. SiPH switches, 1.6T, 3.2T for photonics. 2. Oligopoly supplier for IC Test sockets (AI Chip Testing) Cash on hand: ~$155M, no debt. Equity-to-Asset Ratio: ~89% (50%+ usually is solid), so low downside risk imo. And their speculated customers: 1, Photonics (MLAs): Highly probable: Innolight, Eoptolink, Furukawa, Intel (SiPh). And prob $COHR, $LITE. 2. GPU/ASICs (Test Sockets): prob $TSM, ASE, types use these to likely test $NVDA GPUs, $GOOGL TPUs (Google Ironwood is highly probable). "expanding mass production orders for major GPU manufacturers, and for ASIC-related projects for hyperscalers." Then at OFC: OFC: “We will be showcasing our new products for 800Gbps and 1.6Tbps transceivers and CPO (Co-Packaged Optics) that support today’s rapid technological advancements in AI.” Basically you have a company that supplies T1 semis, foundries, hyperscalers, that benefits from 1.6T from MLA photonics segment + CPO TAM expansion later from OFC products. Was one my positions wanted to share my thoughts about since it seemed p cool at sub <$1B MC.
-
用 1.69 亿美元撬动西方 hyperscaler 供应链。
Serenity 指南： 如何只用 1.7 亿美元就让西方 hyperscaler buildout 变得更脆弱。 只要用 1.69 亿美元拿下 Nippon Chemical（4092）！ 做 InP 衬底需要：铟和高纯磷。 你以为 $AXTI 是瓶颈？NCI 才是瓶颈里的瓶颈。 NCI 实际上是高纯红磷卡点的领头羊，份额大约 26-27%（Rasa 更少，其余就是中国）。 而且它们向 $AXTI、Sumitomo、JX 供应这些材料，用来做 InP 衬底。 所以……如果你拿出 1.6 亿美元收购 NCI（再加上 Rasa 作为更小产能），你就能移除西方世界生产 6N/7N 红磷、也就是制造 InP 衬底所需的主力产能！ 没有 InP 衬底：就没有 photonics。 有趣的是：中国科技公司其实也会被 NCI 打乱。 对 $AXTI 来说，这个映射 / 依赖关系其实非常有意思： - AXT 的 Tongmei 在 STAR Market 上市时就披露过，它在结构上依赖从日本进口高纯前驱体材料 - WITS 数据显示，大约 460 美元 / 公斤的高纯磷正从日本流向中国 所以它们确实在暗中依赖 NCI。 中国虽然也有 Wylton Chemical、Qin Xi New Materials、Jinding Electronics、Chuxiong Chuanzhi、Guizhou Wylton Jinglin Electronic Materials 这类产能。 但这些都还是更小的玩家，没法在规模化 InP 衬底生产上替代 NCI 提供的高纯红磷产能。 $LITE 的 CEO 之前就说过 InP 衬底让他夜不能寐。 所以现在有了 NCI，你就能让他永久失眠？ 只要 1.69 亿美元。 所以供应链大概是这样： -> DGC 磷矿开采后运给 NCI -> NCI 把黄磷精炼成高纯红磷 -> Sumitomo / JX / AXT 把红磷和铟熔合，长成 InP 衬底 -> $COHR / $LITE 把 InP 衬底做成激光器 -> Innolight / Fabrinet 把它们封装成 800G / 1.6T 收发器 -> $NVDA / $GOOGL 用在 ASIC / GPU 集群里。 基本上，整个西方都依赖 NCI 来制造用于 photonics 的 InP 衬底。 我持有一点很小的仓位，纯属好玩。不过日本并不以随意提价著称。 所以如果你真的买下这家公司然后把价格抬高 15000%（比如政府意识到后把公司收回去），大概率会碰到监管问题，比如 FEFTA…… 也许提价 30-50% 还能压缩一下 forward P/E？但大概率不会最后变成 $AXTI 那样。 不管怎样，这家公司按大概 1.6 亿美元市值定价，本身就是一个巨大的国家安全风险。 至于基本面，它们市净率 0.54、forward P/E 11.4，所以其实也可能是被低估了。 TLDR： -> 它是下一个 $AXTI 吗？不是。 -> 它是不是西方 hyperscaler buildout + photonics 里一个未知的结构性瓶颈和关键漏洞？是。 -> 还有没有重估空间？ 只要回到 1 倍市净率，立刻就有 80-85% 的上涨空间。如果再给它更高的 forward P/E，可能更多。 总之，用 1.69 亿美元找到 hyperscaler 供应链里的一个主要失效点，挺好玩的。
英文原文
A Guide by Serenity: How to Cripple the Western Hyperscaler buildout with just $170m. Just take over Nippon Chemical (4092) with $169m! For InP substrates, you need: Indium and High Purity Phosphorus. Thought $AXTI was a bottleneck? NCI is the bottleneck of the bottleneck. NCI is actually the leader of the high purity red phosphorus chokepoint holding 26-27% of the market share (Rasa has less share, then the rest is China). And they export to $AXTI, Sumitomo, JX that need it to make InP substrates. So… if you have $160m to spend to acquire NCI (plus Rasa as smaller capacity), you can remove the leading Western world’s production of 6N/7N red phosphorus needed to make InP substrates! And without InP substrates: no photonics. Fun fact: China’s tech companies would get pretty disrupted with it too by NCI. For $AXTI, the mapping/reliance is actually pretty interesting: - AXT's Tongmei outlined its structural reliance on importing high-purity precursor materials from Japan on their STAR Market listing - WITS data showing ~$460/kg high-purity phosphorus flowing from Japan into China So they secretly do depend on NCI. China does have capacity like Wylton Chemical, Qin Xi New Materials, Jinding Electronics, and Chuxiong Chuanzhi, Guizhou Wylton Jinglin Electronic Materials as well. However, they’re all smaller players so can’t make up for high purity red phosphorus capacity provided by NCI for InP substrate production at scale. $LITE CEO already said inp substrates keeps him up at night. So now with NCI, you can give the guy permanent insomnia? For just $169M. So here's what the supply chain looks like: -> DGC phosphate rock mine and ships it to NCI -> NCI refines Yellow Phosphorus into High Purity Red Phosphorus -> Sumitomo / JX / AXT melt the Red Phosphorus with Indium to grow InP Substrates -> $COHR / $LITE fab InP substrates into Lasers -> Innolight/Fabrinet package them into 800G/1.6T transceivers -> $NVDA / $GOOGL use them for ASIC/GPU clusters. And basically, the entire West depends on NCI to make InP substrates for photonics. I hold some very small positions, just for fun. However, Japan is not well known for price hiking. So you’d probably run into regulatory problems eg. FEFTA if you bought the company and hiked prices 15000% (like government seizing back the company once they realize)… Maybe 30-50% hikes is possible to compress fwd p/e? But very likely wont end up like $AXTI. Regardless, this company is a massive, massive national security risk priced at ~$160m. As for fundamentals, they’re trading at .54 book value and a forward P/E of 11.4 so it’s probably undervalued anyway. TLDR: -> Is it the next $AXTI? No. -> Is it an unknown structural bottleneck + critical vulnerability of the Western hyperscaler buildout with photonics? Yes. -> Is there still room for re-rating? Just reverting to Book Value of 1 is immediate 80-85% upside. Maybe more if you give it multiples past 11 fwd p/e. Regardless, it’s fun to find a major point of failure in the hyperscaler supply chains for $169m.
-
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.
-
认为 AEHR 在小市值下被低估，正被多家超大客户做资格认证。
$AEHR 在大约 11 亿美元市值下看起来极有吸引力。 Aehr 开始让我想起早期的 $TER，又有点像财报前的 $AAOI。 如果把时间线和猜测客户列出来： 2 月 11 日：Sonoma 为超大客户的 AI ASIC 处理器拿到生产胜利。（大概率是 $GOOGL、$AMZN、$META） - 可能是谷歌？Aehr 收购了 Incal，而 Incal 曾被猜测会被谷歌用于 TPU。 2 月 26 日：来自 AI 主力客户的 1400 万美元。（大概率是 $AMD、$NVDA） - 这里更像是 $AMD，用于 Instinct MI300/MI400。 3 月 3 日：一台 FOX-XP 系统对应的主硅光客户。（大概率是 $INTC siph） - 很可能 $INTC 就是他们的主客户。 3 月 31 日：来自一家新的大型硅光客户的首单。（大概率是 $AVGO、$MRVL、$CSCO） - 新客户（排除了 Intel），大概是这些正转向 800G/1.6T 硅光收发器的公司之一。 （这些都属于非常保密的 BOM 级推测） 不管怎样，这条时间线对 $AEHR 来说就是在不断加码。 也许就是下一次财报。也可能再过两季。 但感觉量产订单只是时间问题。
英文原文
$AEHR looks extremely promising at ~$1.1B MC. Aehr is starting to remind me of an early $TER, mixed with pre-earnings $AAOI. If we look at the timeline and speculated customers: Feb 11th: Sonoma production win for Hyperscaler's AI ASIC processors. (likely $GOOGL, $AMZN, $META). - Probably Google? Aehr bought Incal, who was speculated to be used by Google for their TPUs. Feb 26th: $14 million from AI lead customer (likely $AMD, $NVDA) - Probably $AMD here for Instinct MI300/MI400. March 3rd: Lead silicon photonics customer for one FOX-XP system (likely $INTC siph) - Very likely $INTC has been their lead customer. March 31st: Initial order from major new silicon photonics customer (likely $AVGO, $MRVL, $CSCO ) - New customer (rules out Intel), prob one of these transitioning to 800G/1.6T silicon photonics transceivers (All speculative, very confidential BOM) Regardless. This timeline is just bottling up for $AEHR. Could be next earnings. Or two quarters from now. But feels like a matter of time before we see mass orders.
-
整理Mirae Asset的CPO价值链全图，涵盖激光源到EDA各环节上市公司
来自Mirae Asset的CPO价值链概览： 激光源： Coherent < $COHR > Lumentum < $LITE > Furukawa Electric (TYO: 5801) Yuanjie Semiconductor (SHA: 688498) Innolight Technology / 中际旭创 (SZSE: 300308) PIC代工： TSMC < $TSM > GlobalFoundries $GFS Samsung Electronics (KRX: 005930) Tower Semiconductor < $TSEM > EIC、驱动IC： Broadcom < $AVGO > Marvell < $MRVL > NVIDIA < $NVDA > ELS、光引擎： Innolight / 中际旭创 (SZSE: 300308) TFC / 苏州天孚光通信 (SZSE: 300394) O-Net Technologies Eoptolink Technology (SZSE: 300502) FAU（光纤阵列单元）： Senko Advanced Components（私企） Sumitomo Electric (TYO: 5802) TFC (SZSE: 300394) FOCI Fiber Optic Communications (TWO: 3363) FAU、对准工具： ficonTEC（私企） All Ring Tech (TWO: 6187) ADST（私企） FAU、引擎组装： Fabrinet < $FN > Hon Hai / Foxconn (TWSE: 2317) ASE Technology < $ASX > FOCI (TWO: 3363) OSAT、先进封装： ASE Technology < $ASX > Amkor < $AMKR > Kyocera (TYO: 6971) Powertech / PTI (TWSE: 6239) Shinko Electric (TYO: 6967) Fabrinet < $FN > 连接器、套管： Senko Advanced Components（私企） Sumitomo Electric (TYO: 5802) US Conec（私企） T&S Communications (SZSE: 300570) Molex（私企） Browave (TWO: 3163) 光纤： Corning < $GLW > Sumitomo Electric (TYO: 5802) Nittobo / 日东纺绩 (TYO: 3110) 电/光测试： Keysight < $KEYS > Teradyne < $TER > FormFactor < $FORM > Chroma ATE (TWSE: 2360) Multilane（私企） 交换、系统： NVIDIA < $NVDA > Broadcom < $AVGO > Marvell < $MRVL > Google < $GOOGL > EDA： Synopsys < $SNPS > Cadence < $CDNS > Ansys < $ANSS > 对这些名单中的一些名称感到困惑，他们可能混淆了一些公司，比如把Innolight和激光源公司（如$MTSI、Sumitomo、$SIVE、Luxnet）与实际终端模块混在一起了（除非有什么非公开材料或者我遗漏了）。 但对于对这个领域感兴趣的人来说，这是一个很好的高层概览。 （该推文引用了 @aleabitoreddit 的推文，引用内容仅供理解语境）: CPO领域Mirae分析师笔记： 规模化方向： CPO ASIC：$AVGO、$MRVL 光收发器：$COHR、$LITE、中际旭创 DSP/PAM4：$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、$COHR OCS网关：KDDI 规模提升方向： SiPh代工：imec、$GFS、$TSM SiPh调制器：$NVDA（内部MRM）、$INTC ELS：NTT、Furukawa、$LITE、$COHR 太赫兹互连：研发阶段？ CPO测试："新玩家进入增加" 微透镜/光学系统："新玩家进入增加" TLDR：规模提升的CPO即将到来。我认为分析师笔记遗漏了一堆上游公司名称，并把ELS和光源混为一谈了。 但对于了解他们认为谁是领先玩家来说，这是个很好的高层概览。
英文原文
CPO Value Chain Summary from Mirae Asset: Laser Source: Coherent < $COHR > Lumentum < $LITE > Furukawa Electric (TYO: 5801) Yuanjie Semiconductor (SHA: 688498) Innolight Technology / Zhongji Innolight (SZSE: 300308) PIC Foundry: TSMC < $TSM > GlobalFoundries $GFS Samsung Electronics (KRX: 005930) Tower Semiconductor < $TSEM> EIC, Driver IC: Broadcom < $AVGO > Marvell < $MRVL > NVIDIA < $NVDA > ELS, Optical Engine Innolight / Zhongji Innolight (SZSE: 300308) TFC / Suzhou TFC Optical Communication (SZSE: 300394) O-Net Technologies Eoptolink Technology (SZSE: 300502) FAU (Fiber Array Unit): Senko Advanced Components (Private) Sumitomo Electric (TYO: 5802) TFC (SZSE: 300394) FOCI Fiber Optic Communications (TWO: 3363) FAU, Align Tools: ficonTEC (Private) All Ring Tech (TWO: 6187) ADST (Private) FAU, Engine Assembly: Fabrinet < $FN > Hon Hai / Foxconn (TWSE: 2317) ASE Technology < $ASX > FOCI (TWO: 3363) OSAT, Advanced Packaging: ASE Technology < $ASX > Amkor < $AMKR > Kyocera (TYO: 6971) Powertech / PTI (TWSE: 6239) Shinko Electric (TYO: 6967) Fabrinet < $FN > Connector, Ferrule: Senko Advanced Components (Private) Sumitomo Electric (TYO: 5802) US Conec (Private) T&S Communications (SZSE: 300570) Molex (Private) Browave (TWO: 3163) Fiber: Corning < $GLW > Sumitomo Electric (TYO: 5802) Nittobo / Nitto Boseki (TYO: 3110) E/O Testing: Keysight < $KEYS > Teradyne < $TER > FormFactor < $FORM > Chroma ATE (TWSE: 2360) Multilane (Private) Switch, System: NVIDIA < $NVDA > Broadcom < $AVGO > Marvell < $MRVL > Google < $GOOGL> EDA: Synopsys < $SNPS > Cadence < $CDNS > Ansys < $ANSS > Confused by some of names of the list, they might have conflated a few names like Innolight with laser source like $MTSI, Sumitomo, $SIVE, Luxnet, with the actual end module (unless there's something that's not public material or I missed)? But just for people interested in the landscape, this is a good high-level overview.
-
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 $SIVE remains independent by owning enough shares?
-
Mirae分析师梳理CPO产业格局，作者认为遗漏上游公司但框架有参考价值
CPO产业格局 Mirae分析师笔记： 横向扩展： CPO ASIC：$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.
-
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.
-
分享小账户分散化投资组合，涵盖AI半导体、光子学、机器人、加密等领域的高增长和杠铃策略标的
快速增长赛道： $AAOI - 光收发器2027年下半年收入增长10倍 $NBIS - 2026年第四季度收入增长10倍 $ARM - 全新AI CPU带来5倍收入增长 $MRVL - 来自$MSFT Maia芯片项目2-3倍收入增长 $AVGO - 长期持有超大规模云厂商ASIC $LITE - 长期持有OCS/谷歌TPU Win Semi - 对前沿行业的代工敞口 $TSEM - 长期持有光子业务，订单积压 SK Hynix - 存储业务敞口，极端营业利润增长 同时配置一些杠铃策略，远离超大规模云厂商资本支出（除亚马逊外）： $VNP - 长期持有西部供应链的稀土 $NEO (TCX) - 机器人供应链 $AMZN - 机器人/AI降低运营成本 $CRCL - 稳定币多头 $RDDT - 高得离谱的利润 $GLD - 安全对冲 $IBIT - 2028年减半 $CVX 看涨期权 - 石油对冲 也许还有长期配置（你知道迟早会来）： $INTC/$AMKR - 美国制造供应链 $SOI - 硅光子学/共封装光学(CPO)基板 $RKLB - 对太空行业的长期看涨期权 然后挑选一两支小盘股博暴涨： $SIVE - CW激光关键点或$I QE - Landmark重组重估，这是我最喜欢的两个 还有其他我提过的，比如$AEHR用于测试或$VPG用于Optimus 我自己在$AXTI等股票上的主动管理模式与其他人应该做的完全不同风险敞口配置。 在这种宏观环境下全仓投入高贝塔标的不太是最佳选择。
英文原文
Faster compounds: $AAOI - 10x revenue ramp from optical transcivers h2 2027 $NBIS - 10x revenue ramp Q4 2026 $ARM - 5x revenue growth from their new AI CPU $MRVL - 2-3x revenue growth from $MSFT Maia Ramp. $AVGO - Long hyperscaler ASIC $LITE - Long OCS / Google TPU Win Semi - Foundry exposure to frontier industries $TSEM - Long photonics, backlogged SK Hynix - Memory exposure, extreme operating income ramp With some barbell exposure away from Hyperscaler capex aside from Amazon: $VNP - Long term rare earths for Western Supply chains $NEO (TCX) - Robotics Supply chains $AMZN - Robotics/AI cutting opex $CRCL - Stablecoin long $RDDT - Ridiculously high profit $GLD - Safe Hedge $IBIT - Halving 2028 $CVX Calls - Oil Hedge And maybe long term (you know it's coming): $INTC / $AMKR- Made in America supply chains $SOI - Silicon Photonics / CPO substrates. $RKLB - Long term call on Space industry Then pick one or two small cap moonshots: $SIVE - CW Laser Chokepoints or $IQE for Landmark rerating on restructuring were my two favorites. There's others I've mentioned like $AEHR for testing or $VPG for Optimus. How I actively manage my own stuff from $AXTI and others is a lot different risk profile than what others should do. Going full port into high-beta in this macro environment is not the best idea.
-
说超大规模 ASIC 正在吞掉推理市场，如果只剩训练会有麻烦。
@yuntungshieh 超大规模云厂的 ASIC 正在把推理市场吃掉。 如果你被稀释得只剩训练业务，那可能就麻烦了。 另外我还有别的想法，但不想公开发出来。
英文原文
@yuntungshieh Hyperscaler ASICs eating up inference market. If you’re diluted down to just training could be trouble. Also there’s something else but don’t want to post it publicly