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
-
LITE 的产能分配影响 AM D/MRVL
这不算好消息，但也没那么糟。 $LITE 的产能大概率已经分给了 $NVDA，这也是为什么 $AMD 的 CPO（大概率）会去用 $SIVE + $GFS，我是这么猜的。 $MRVL 大概会直接从 $SIVE 买货，然后多来源，可能再加上 $MTSI，并把封装层在内部设计掉。
英文原文
It's not good news. But it's not that bad either. $LITE likely has capacity allocated toward $NVDA, which is why $AMD CPO (likely) went with $SIVE + $GFS is my guess. $MRVL will probably just buy from $SIVE directly and multi-source, maybe $MTSI and design out the packaging side of things in-house.
-
MRVL 可能直接买激光器
很有可能现在 $MRVL 会直接从 $COHR、$SIVE 或 $LITE 采购激光器。 这只会让 Marvell 的 Celestial 路线图延后一些。 但我一直都预期封装这层最终会被垂直整合，只是没想到这么快。 激光 IP 要复杂得多。 激光公司为什么都值几十亿美元，是有原因的。
英文原文
Highly likely now $MRVL just buys the lasers directly from $COHR, $SIVE, or $LITE. It just causes delays to Marvell Celestial roadmap. But I've always had expectations on the packaging side of things to get vertically integrated down the road, just not this soon. Laser IP is much, much more complex. There's a reason why all the laser companies are worth tens of billions.
-
POET 很可能被设计掉
@AD14977 我一直说 $POET 是最有可能在后面被设计掉的那个。 他们只是买激光器然后封装……而 $MRVL 完全可以直接去激光源头买，虽然这会让他们的路线图有些延迟。
英文原文
@AD14977 I've always said $POET is the most likely to get designed out down the road. They just buy lasers and package them... and $MRVL can just go directly to the laser source and do the same with some delays to their roadmap.
-
POET 单一客户风险高
@satyadeva97 @PhotonCap 对 $POET 来说，这不算过度反应。Poet 之前很大程度上存在单一客户集中风险，虽然他们后来也开始接到一些更小的客户。
英文原文
@satyadeva97 @PhotonCap For $POET, not an overreaction. Poet was largely single source customer concentration risk, although they started getting other smaller customers.
-
MRVL 可能自研中介层封装
$MRVL 很可能会把那个 interposer 封装 IP 变成内部能力。我原本就预计他们会在后面把 $POET 设计出去，只是没想到这么早。 他们完全可以直接从 $LITE 或 $SIVE 这种公司买激光器，然后自己搭建整个流程。 激光设计比封装难得多。
英文原文
$MRVL will likely do that interposer packaging IP in-house. I did expect them to design out $POET down the road, just not this early. They can just buy lasers directly from someone like $LITE or $SIVE and build the process themselves. Laser design is much much harder than the packaging side of things.
-
MRVL 要把封装垂直整合
@BrokeBoyCap 哦，我猜 $MRVL 是在买了 $SIVE 或 $LITE 这类公司的激光器之后，把封装流程垂直整合掉。 我之前说的就是这个，后面那一层比上游激光容易多了。
英文原文
@BrokeBoyCap Oh, I'm guessing $MRVL is vertically integrating the packaging process after buying the lasers from someone like $SIVE or $LITE. That's what I said earlier it's a lot easier to do that layer than upstream lasers.
-
Towa 被市场忽略了
@reallldealll 我觉得市场忽略了 Towa，因为你们这些存储厂商都在创纪录地增加 capex。 https://t.co/iVmufEGXF8
英文原文
@reallldealll I think markets missed Towa given record capex spend from all your memory makers https://t.co/iVmufEGXF8
-
欧洲和日本的垄断都在起飞
像 $LPK 这种欧洲“垄断”（玻璃核心基板）最近也在一路起飞。 $ALRIB 技术上是双寡头，但也在这条线上……（量子 / MBE） 日本那边也有像 Towa 这种（HBM4 / 压缩成型）我持有的名字。 通常垄断会拿到更高倍数。 我之后会慢慢提到更多，只是得先把它们都找出来。
英文原文
European “monopolies” like $LPK (Glass Core Substrates) have also been going brrr lately. $ALRIB is technically a duopoly but it’s there too… (Quantum / MBE) Then there’s some over in Japan like Towa for (HBM4/compression molding) too that I own. Usually monopolies get higher multiples. I'll mention some more over time, just gotta find them all.
-
SIVE 可能是苹果供应链关键
现在很明显，$SIVE 很可能就是 $AAPL 的硅光“圣杯”供应商…… 市场只是还不知道而已（这就叫 Alpha）。 如果你想知道怎么把 Apple 映射到 Sivers： > “美国财富 100 强”客户。 > 每年 5000 万台规模的 RFQ（报价请求）。 > Apple 2022 年大约卖了 5390 万块 Apple Watch，年销量通常在 3800 万到 5000 万台之间。 > $SIVE / Carnegie research 发布了 135 个波长架构，这和 Apple 的应用规格匹配。 真的……全世界还有哪个美国财富 100 强公司能这么轻松地一年做 5000 万台消费穿戴设备？就连三星也就是 1000 万到 1500 万台。 这刚好和激光架构和应用场景对应上。 $SIVE 在 X 上还暗示过来自开发时间线的量产爬坡，我猜那是苹果（H2 2027 爬坡或 H1 2028），但也可能是 Ayar 或 Jabil。
英文原文
It's pretty clear by now that $SIVE is the likely holy grail silicon photonics supplier for $AAPL... Markets just don't know about it (this is called Alpha). If you're curious how to map Apple to Sivers: > "US Fortune 100" customer. > RFQ (request for quote) for 50 MILLION units a year. > Apple sold roughly 53.9 million Apple Watches in 2022 and averages between 38m-50m units annually > $SIVE / Carnegie research released 135 Wavelength arechitectures, which matches Apple application specifications. literally... what other US Fortune 100 company in world can casually produce 50 million units of consumer wearables a year? Even Samsung does 10-15M units. That coincidentally matches with the laser architecture and applications. $SIVE on their X did tease volume ramp from development timelines, which I guessed was Apple (H2 2027 ramp or H1 2028), but could be Ayar or Jabil.
-
CPU 热度盖过了存储
现在市场都在盯着 CPU，差点把存储给忘了。 但……SK 海力士、$MU、三星最近都在提高 capex，而且三星最近还开始了 HBM4 生产。 $TSM 也释放了全行业创纪录 capex 的信号。 但就像 $LPK 在玻璃核心基板上的逻辑一样…… HBM 阵营里其实也有不少结构性垄断，是市场可能忘掉了的，尤其是在日本这些地方。 它们大概率会从当前的 HBM4 capex 周期里受益。 韩国那边，像 Hanmi Semi（KRX: 042700）今天都涨了 27.6%+，所以我猜全球其他公司也很快会跟上。
英文原文
Markets are looking at CPUs right now and kinda forgot about memory. But... there's increased capex spend with Sk Hynix, $MU, Samsung around now, with Samsung starting HMB4 production recently. $TSM also signaled record capex across the board. But just like $LPK in glass core substrates... There's a decent amount of structural monopolies over in the HBM camp markets that I'm thinking about in places like Japan. That markets may have forgotten? They would largely benefit from current HBM4 capex cycles. Over in Korea, things like Hanmi Semi (KRX: 042700) have been taking off, up 27.6%+ today, so I'd guess the other companies around the world might play catchup soon.
-
CPO / 玻璃基板 / 光学测试分工
@wdh37 $SIVE 用于 CPO / 1.6T 激光 $LPK 用于玻璃核心基板 Shunsin 用于光学测试 / 封装。 我也喜欢很多别的名字，比如 $HPS.A 或 $SOI。
英文原文
@wdh37 $SIVE for CPO/1.6T lasers $LPK for glass core substrates Shunsin for optical test/packaging. Many others I like though like $HPS.A or $SOI.
-
AI 相关供应链全面拥堵
最近新闻和瓶颈的 TLDR，全部都在起飞： 1. CPU 瓶颈 - $INTC CEO 说 AI 推理把 CPU 比例从 1:8 推到了 1:1。 CPU 起飞（$AMD、Intel、$ARM）-> $AMAT / $TSM / $KLAC 等也跟着起飞。 2. PGME / PGMEA 短缺。DuPont、Shiny Chemical、Daxin、San Fu、$DOW 等都要起飞？ 光刻胶瓶颈也起飞？ 3. 微控制器潜在瓶颈 + 涨价（Arterytek / Arterychip）正在压缩 AI 产能。 MCU 公司可能也要起飞？ 4. 总统本周援引了《国防生产法》，里面包括： - 变压器 - 输电组件 - 先进导体 - 电力电子 - 变电站 - 高压断路器 - 保护继电器、 电容组 - 电气硅钢 这些都被称为“严重短缺”。像 $AMSC、$PLPC、$POWL、$VICR、$ATKR、$HPS.A 之类的也要起飞。 5. $GOOGL 在扩容新的 TPU 服务器。谷歌把 AI 芯片拆成训练 TPU 和推理 TPU。 台湾很开心。联发科等也要起飞？ 6. 三星、Kingston 把 SSD 价格上调了 10% 以上。 SSD 价格还会继续起飞？ 7. T-玻璃玻纤短缺还在恶化？Nittobo 等也继续起飞？ 8. 用于蚀刻电路和阻燃的关键材料溴，价格已经涨到每公吨 12,000 美元。 以色列的 ICL Group 似乎掌控了全球 40% 的供应？ 我对这个不算特别熟，但也有点可疑的起飞？ 9. “外延厂 Landmark Optoelectronics 报告称产量仍远低于客户需求”。 嗯 $IQE 和其他公司也要起飞？ 10. “AI 数据中心碰到互连限制，正在推高光模块需求”。“瓶颈不再只是算力，而是算力如何连接。” 来自 $AAOI、$LITE、$COHR、Innolight 等的光子链条继续起飞？下一代来自 $SIVE、$POET、$MRVL、Win Semi 等也起飞？ 基本上就是 AI 半导体供应链全面起飞，因为 AI 超大规模云厂商需求带来了到处都是的短缺。
英文原文
TLDR of recent news + bottlenecks that go brr: 1. CPU bottleneck - $INTC CEO said AI inference pushed CPU Ratio From 1:8 to 1:1. CPUs go brr ( $AMD, Intel, $ARM) -> $AMAT / $TSM / $KLAC, etc. go brr. 2. PGME / PGMEA shortage. DuPont, Shiny Chemical, Daxin, San Fu, $DOW and others go brr? Photoresist bottleneck go brr? 3. Microcontroller potential bottleneck + price hikes (Arterytek/Arterychip) was weighing price hikes on AI capacity squeezes. MCU companies potentially go brr? 4. President invoked the "Defense Production Act" this week, it included: -Transformers - transmission components - advanced conductors - power electronics - substations - high-voltage circuit breakers - protective relays, capacitor banks - electrical core steel As "severe shortages". Stuff like $AMSC, $PLPC, $POWL, $VICR, $ATKR, $HPS.A go brr. 5. $GOOGL ramps new TPU servers. Google splits AI chips into training and inference TPUs. Taiwan happy. Mediatek and others go brr? 6. Samsung, Kingston lift SSD prices by over 10%. SSD prices keep going brrr? 7. T-glass fiberglass shortages keep getting worse? Nittobo and others keep going brrr? 8. Bromine, essential for etching circuits and flame retardancy, has surged to $12,000 per metric ton. ICL Group in Israel apparently controls 40% of the global supply? Not as familiar with this but questionable brrr? 9. "Epitaxy manufacturer LandMark Optoelectronics reporting output still far below customer needs". Uhh $IQE and others go brr? 10. "AI data centers hit interconnect limits, boosting optical module demand". "the bottleneck is no longer computing power alone, but how that power is connected." Photonics from $AAOI, $LITE, $COHR, Innolight and others keep going brr? next gen from $SIVE, $POET, $MRVL, Win Semi and others go brr? Basically AI semi supply chains go brr because there's widespread shortages everywhere due to AI hyperscaler demand.
-
Win Semi 让 SIVE 放量去风险
@Sopcaja 不，因为 $SIVE 的量产爬坡已经被 Win Semi 去风险了，这是我另一个多头。 Win 是 $AVGO 以及现有 SpaceX / 超大规模云厂商供应链的代工厂，所以它非常可靠。
英文原文
@Sopcaja No, because $SIVE volume ramp is de-risked by Win Semi, my other long. Win is the foundry for $AVGO and existing SpaceX/Hyperscaler supply chains, so it's heavily dependable.
-
ASX 加速 CPO
@Quantrarian52 对！我也看好 $ASX？而且他们的报告已经把 CPO 加速判断推到 2026 年下半年。
英文原文
@Quantrarian52 I’m a fan of $ASX? And converted their reports over accelerating CPO to H2 2026.
-
Shunsin 会在未来一年跑赢
@GeorgeMakro91 对！我短期时间框架上也会判断失误。 我 4 月 12 日提到 Shunsin，也就是 13 天前，当时把它看作富士康的 CPO 光学 / 测试业务。我对那笔也判断错了（sadge），但我预计它未来一年会跑赢。
英文原文
@GeorgeMakro91 Yep! I get definitely get things wrong on short term timeframes too. I mentioned Shunsin April 12th, so 13 days ago as Foxconn's CPO optical/testing arm. I'm down too on that (sadge), but I expect it to outperform over the next year.
-
上游卡点还没被完全梳理
这是真的。我不认为美国已经把所有上游卡点都完全梳理出来了。 这也是为什么中国还能逐步拿到那些美国超大规模云厂商所需的不同垄断，并且在过程中慢慢拿走 IP。 散户持股其实是防止敌意收购和外部对手控制的好办法。
英文原文
It’s true. I don’t think US completely mapped out all the upstream chokepoints yet. Which is why China was still able to acquire different monopolies required by US hyperscalers and steal IP over time. Retail ownership is a great way to prevent hostile takeovers and control from foreign adversaries.
-
变压器/开关设备短缺会持续
@nodeflowsol 变压器 / 开关设备的瓶颈大概会持续 1.5 到 5 年……而且几天前它们刚被认定为国家安全风险，随后总统也援引了《国防生产法》。 这些公司会因为联邦资金获得创纪录的 capex 流入和产能扩张。
英文原文
@nodeflowsol Transformer/switchgear bottlenecks are like 1.5-5 years… and it was designated as a national security risk few days ago + got the defense production act invoked by the President. Lot of these players will get record inflows for capex + capacity growth from federal funding.
-
光子链条整体都在涨
光子学起飞。 $MXL +76.2%（哈哈） $AAOI +15.03% $SIVE +12.69% $AEHR +7.07% $SOI +6.19% $LITE +4.96% 希望你没有卖掉以后去追别的卡点，匿名网友？ https://t.co/Ft2p94bDWR
英文原文
Photonics go brrr. $MXL +76.2% (lol) $AAOI +15.03% $SIVE +12.69% $AEHR +7.07% $SOI +6.19% $LITE +4.96% Hope you didn’t sell and try and chase other bottlenecks anon? https://t.co/Ft2p94bDWR
-
光子与存储都在加速
谁会想到把 Tangerine Peel 风格公司给资金…… 然后去做空光子和存储里增长最快的名字？ $SNDK 和 $AAOI 最近都在疯狂上涨。 我也看到像 AAOI 这样的名字会继续增长，因为光模块和美国制造供应链都在扩张。
英文原文
Who thought it was a good idea to give Tangerine Peel inspired companies the funds… To short the fastest growing names in photonics and memory? $SNDK and $AAOI have been on a hyperbolic run. And I see names like AAOI continuing to grow as optical transceiver + Made in America supply chains grow.
-
Rockley 退出后 SIVE 补位
@veggiepoultry @StormDirac Rockley 掉下去了，这很可能就是 $SIVE 成为 $AAPL 供应商的原因。
英文原文
@veggiepoultry @StormDirac Rockley went down, which is likely the reason why $SIVE became the $AAPL supplier.
-
AAPL 是量产爬坡的暗示客户
@Embezzl3r 对，这么看确实合理，把 $AAPL 看成他们暗示过的量产爬坡超大规模云客户。
英文原文
@Embezzl3r Yeah makes sense about seeing their reactions to $AAPL as their hinted hyperscaler customer for volume ramp.
-
META 可能是终端客户
@xm_build 这甚至都算不上泄密…… 你知道 $SIVE 在给 $JBL 的 1.6T LRO 供货，而当你指出 $META 很可能是最终终端客户时（因为 Meta 是 $INTC 的硅光客户，而 Jabil 已经收购了这个链条）， 人们就会生气，然后开始做空。这个文化挺有意思的。
英文原文
@xm_build It's not even leaks though... You know $SIVE is supplying to $JBL 1.6T LRO and when you point out $META is the likely end user (because Meta is the $INTC SiPH customer Jabil acquired). People get upset and start shorting. It's an amusing culture.
-
把 SIVE 激光供应链映射到各家客户
我现在已经对瑞典文化感到很有趣了。 当有人把 $SIVE 激光的供应链映射到下面这些名字时，他们的散户 / 记者每天都在哭： -> $MRVL CPO -> $AMD CPO -> $JBL 1.6T LRO -> $NVDA -> $AMZN、$META、$GOOGL、$MSFT -> 腾讯、百度、阿里用于 ELS。 然后他们还会哭得更厉害…… 当别人展示 $AAPL 是 Sivers 激光器的秘密超大规模云客户时（因为苹果下一代 Apple Watch 开发需要每年 5000 万台产能）。 给超大规模云厂商供货……会让本地人难受成这样，甚至想自毁吗？ 有个“瑞典专业人士”出来说 CPO 没什么新东西、也没什么特别，所以 Sivers 也没什么特别，这真的很可笑。 与此同时，$NVDA 正在给从 $MRVL 到 $LITE、$COHR 的每个 CPO 项目都砸 20 亿美元。 X 最酷的地方在于，你每天和全球各地的人互动，都会学到新东西。
英文原文
I'm amused by Swedish culture at this point. That their retail/journalists are crying every day when someone posts a supply chain mapping of $SIVE lasers to: -> $MRVL CPO -> $AMD CPO -> $JBL 1.6T LRO -> $NVDA -> $AMZN, $META, $GOOGL, $MSFT -> Tencent, Baidu, Alibaba for ELS. Then they go crying even further... When someone shows $AAPL is secret hyperscaler customer of Sivers lasers for their next-gen Apple Watch development (which requires 50,000,000 units annually). Does supplying to hyperscalers... Hurt locals's feelings that much to the point they try and self-destruct? It's hilarious that a "Swedish professional" went out and said CPO is nothing new and nothing special, so Sivers is nothing special. Meanwhile $NVDA is just funding every CPO program $2B from $MRVL to $LITE to $COHR left and right. Cool thing about X is you learn something new every day interacting with global audiences.
-
AAPL 高概率也在用它们
@Matrix_B0SS $AAPL 高概率也在用它们做下一代 Apple Watch，所以这就是 7/7。
英文原文
@Matrix_B0SS $AAPL high certainty uses them too for next-gen apple watches, so that's 7/7.
-
Ayar 的激光更贵
从我看到的情况来看，那是不同类型的激光。给 Ayar 用的更贵，因为 AI 数据中心要求更高，比如还要做光纤对准，所以成本也更高。 Apple 这边只需要推到大约 100mw 的组合功率，相比 AI 光子数据中心要简单得多，而且他们还能压缩激光尺寸。 另外，因为限制更宽松，晶圆良率也应该更高。 以 Apple 的规模来看，单位经济性也便宜得多。
英文原文
From what I’ve seen, it’s different lasers, ones for Ayar are more expensive since AI DCs have higher requirements, eg. there’s also fiber alignment involved, which hikes the cost. Apple should only push out 100 mw combined, compared to AI photonic DCs, and they can compress laser size. Also wafer yields should be higher since looser restrictions compared to AI DCs. Cost at Apple scale is also much much cheaper in terms of unit economics.
-
TAM 扩张不只 Apple 一家
@Photonix1337 这已经低估了 TAM 的扩张，因为三星、华为、Google 以及其他公司都在复制 $AAPL。 Apple 还能把同样的技术用到其他产品线里。
英文原文
@Photonix1337 That’s underselling TAM expansion with Samsung, Huawei, Google, and others all copying $AAPL. Alongside the same technology for other product lines Apple offers.
-
Win Semi 负责扩产
Win Semi 负责扩产……这也是我看多 $SIVE 的原因，这样他们就不用自己烧 capex 去处理这件事。 世界上能做这类事的公开上市公司没几家，而且它们都已经是几十亿美元市值了。 $JBL 专门选 Sivers 是有原因的。Ayar 也是。和 $LITE、$MTSI 等相比都一样。 只看市值就低估这家公司的人真的太多了。
英文原文
Win Semi handles expansion… that’s why I was bullish on $SIVE so they don’t need to burn through capex handling this. There’s only a few publicly traded companies in the world that can do this and they’re all in the tens of billions. There’s a reason $JBL selected Sivers in specific. Same with Ayar. Compared to $LITE, $MTSI and the others. People really underestimate the company just by looking at the MC.
-
Win Semi 替 SIVE 做量产爬坡
@B_Funk_ Win Semi 负责给像 $SIVE 这样的公司做量产爬坡。Win 体量很大，而且是 SpaceX 供应链和 $AVGO 的代工厂。
英文原文
@B_Funk_ Win Semi hands volume ramp for companies like $SIVE. Win is massive and is the foundry for SpaceX supply chains and $AVGO.
-
AAPL 使用 SIVE InP 激光阵列
秘密是，$AAPL 🍎 正在用 $SIVE 的 InP 激光阵列来大规模生产消费硬件（高置信度）。 按估计和图表来看，发布时间大概在 2027 年下半年。 想象一下 Apple 每年以 5000 万台以上的规模生产消费硬件…… 还是用一家市值 8.5 亿美元的激光供应商，哈哈。 然后 Apple 一旦这么做，$GOOGL、三星以及其他公司也会跟着上。
英文原文
The secret is that $AAPL 🍎 is using $SIVE InP laser arrays for mass production of consumer hardware (high confidence). Launch date around late H2 2027 from estimates and the graph provided. Imagine Apple mass producing consumer hardware with 50M+/units a year… Off a $850m MC laser supplier lol. Then with whatever Apple does, you’ll get $GOOGL, Samsung, and the others following along next.
-
SIVE 正在对接 AAPL
@ashmyro 我已经知道 $SIVE 是下一个很可能的 $AAPL 供应商，这很重要。 我自己最关注的是它们给 AI 数据中心供货的 DFB 激光，因为那会是它们未来的核心营收驱动力。
英文原文
@ashmyro I'm already aware $SIVE is the next likely $AAPL supplier, which is huge. I've been focusing on their DFB lasers going to AI datacenters the most personally, since that's their core revenue drivers in the future.