供应链分析
产业链结构、上下游与瓶颈环节研究 · 共 1122 条 · 第 2 / 23 页
-
SIVE作为光子学激光瓶颈，获GFS/JBL等公开验证。
-> IP 收购 -> 只是等待 CPO 起飞 $SIVE 是光子学的激光瓶颈，并已被公开验证： - $GFS（根据演示，是 CPO 两家公开激光供应商之一，另一家是 $LITE） - $JBL（与 Sivers 构建的可插拔产品拥有“相当显著的护城河”） -
英文原文
-> IP acquisition -> Just waiting for CPO to take off $SIVE is a laser chokepoint for photonics and are publicly validated by: - $GFS (1 of 2 public laser suppliers with $LITE for CPO per presentation) - $JBL (“Relatively dramatic moat” for pluggables built with Sivers) -
-
反驳SIVE空头，称MRVL Celestial可能直接向SIVE买激光器。
有一个新的 $SIVE 空头，完全不知道自己在说什么。 1. $MRVL Celestial 很可能直接购买激光器，$SIVE 是激光供应商，而不是通过 $POET。 因为自2023年以来，它们就被识别为可能的直接客户。 一个类比是如果 O-Net / ……
英文原文
There's a new $SIVE short seller with 0 clue what they're talking about. 1. $MRVL Celestial is likely buying lasers directly with $SIVE as the laser supplier, not through $POET. As they've been identified as a likely direct customer since 2023. An analogy is if O-Net / https://t.co/XDrPS6OVNG
-
NVDA/西门子/FLNC为Vera Rubin N72开发参考电力架构。
$NVDA、西门子和 $FLNC 为 Vera Rubin N72 开发参考电力架构。 这发生在 FLNC 可能即将获得两个超大云厂商交易的时候。 如果你想知道它为什么涨36%，就是这个原因。
英文原文
$NVDA, Siemens, and $FLNC develop reference power architectures for Vera Rubin N72. This comes at the time FLNC has 2 likely hyperscaler deals coming up. Just in case why you’re wondering it’s up 36%. https://t.co/8f4gAhJeTE
-
认为光学股上涨来自极端前瞻收入增长而非其做多。
回复 @ethics3606：不是，很遗憾，我去年做多 $LITE 并不会像公众舆论说的那样把它推到700亿美元市值。 它们都由极端的前瞻收入增长驱动，比如 $AAOI 从2027年下半年开始每月4.71亿美元。 以及 $SIVE 总前瞻收入管线单季度增长77%，从……
英文原文
@ethics3606 No, unfortunately me going long on $LITE last year doesn’t send it to a $70B MC, unlike public opinion. they’re all driven by extreme forward revenhe growth, $AAOI $471m / month starting H2 2027. And $SIVE total forward revenue pipelines growing 77% in a single quarter, from
-
称CPO未来一年半TAM从0到810亿，光子学整体扩大10倍。
回复 @afasano23：CPO 在未来一年半里 TAM 从0扩张到810亿美元。 整体光子学行业未来一年半 TAM 扩张10倍。 哈哈。
英文原文
@afasano23 $0 to $81B TAM expansion for CPO in the next 1 1/2 years. 10x TAM expansion in overall photonics sector next 1 1/2 years. lol
-
推测Innolight等未垂直整合激光器的可插拔玩家可能使用SIVE。
回复 @WuKongWAGMI：我能想到的主要可插拔玩家里，只有少数没有像 $LITE 和 $COHR 那样垂直整合激光器。 所以 Innolight 是可能的。 $SIVE 用了复数暗示有多个玩家，所以很可能还有几个。
英文原文
@WuKongWAGMI There’s only a few major pluggable players that I can think of that don’t vertically integrate lasers like $LITE and $COHR. So Innolight is possible. $SIVE hinted at multiple players since this plural, so there’s likely a few more.
-
认为新可插拔光收发器客户是SIVE最大TAM扩张和收入驱动。
这是 $SIVE 最大的 TAM 扩张和收入驱动，市场还没有注意到。 Sivers 正在积极与新的可插拔光收发器公司合作。 在 Jabil 1.6T LRO 使用 Sivers 激光器实现“相当显著的护城河”之后。 “正在积极合作”，所以……
英文原文
This is the biggest TAM expansion + revenue driver with $SIVE, markets haven’t noticed. Sivers is actively working with new pluggable optical transceiver companies. After Jabil 1.6T LRO used Sivers lasers to achieve a “relatively dramatic moat”. “Actively working”, so https://t.co/NYGMP7tyV4
-
认为MSScorps是CPO检测良率事实垄断，只是CPO尚未放量。
回复 @AstuteKara：MSScorps 即使我不是每天谈它，也会表现不错。它们在 CPO 检测所需良率方面是事实上的垄断。 只是 CPO 还没有任何东西真正放量。
英文原文
@AstuteKara MSScorps will do just fine without me talking about it every day. They’re a de facto monopoly over CPO inspection needed yields. Just nothing CPO has scaled up yet.
-
暗示台北GTC/Computex将揭示下一个AI瓶颈，可能在日本低估值公司。
在台北 $NVDA GTC/Computex 上： 我认为我们会听到下一个 AI 瓶颈。 它由一家市净率0.6倍、拥有180年历史的日本“土豆农场”公司拥有。 它的老板在夜市把那些土豆以每个160日元卖出。 但同样的土豆农场设备……
英文原文
At $NVDA GTC/Computex in Taipei: I think we’ll hear about the next AI bottleneck. That’s owned by a .6 P/B potato farming company in Japan, with a 180 year history. Their owner cooks those potatoes in night markets for 160 yen a piece. But that same potato farming equipment
-
说明这是用于InP衬底的赤磷。
回复 @GabrielNeo9：这是用于 InP 衬底的赤磷。
英文原文
@GabrielNeo9 This is red phosphorus for InP substrates
-
Computex是寻找新CPO敞口公司的机会，提到先进光电和大立光。
在台湾 Computex，对于新的 CPO 敞口公司来说，总是寻找想法的好机会： 例如 3362 先进光电，约4.77亿美元市值；以及大立光(3008)，约146亿美元市值。 PMLA（微透镜阵列）和 FAU。有点让我想起 $HIMX / Foci。 不过，采样/演示阶段和拥有活跃客户之间是有差异的，所以还需要在此基础上做更多研究。 这只是一个很好的研究起点。
英文原文
At Computex Taiwan, for new CPO exposure companies, always a good opportunity to find ideas: Eg. 3362 Advanced Optoelectronic ~$477M with Largan Precision (3008) ~$14.6B PMLA (microlens array) and FAU. Kinda reminds me of $HIMX / Foci There’s a difference between sampling/demo stage though and having active customers, so need do more research beyond this. Just a good starting research process
-
纠正玻璃核心基板时间线：2026下半年开始而非2028。
回复 @talethdabm89：那家研究公司的玻璃核心基板时间线错了。 应该是2026年下半年开始生产（SKC Absolics），不是2028年。TAM 数字到2040年看起来也不太对，所以我觉得最好不要引用。 尽管它不像 $LPK 和 GCS 那样覆盖完整价值链……
英文原文
@talethdabm89 The research firm got their glass core substrate timelines off. H2 2026 should be production start (SKC Absolics), not 2028. And TAM figures seems off toward 2040, so thought it was better not to cite it. Even though it’s not full value chain like $LPK and GCS in specific https://t.co/ixjeKpe3YH
-
列出黄仁勋“你们现在太富了”评论涉及的供应链伙伴。
“你们现在太富了” $NVDA ETF。 这些只是 Jensen 用那句话举杯致意的一部分供应链伙伴： TSMC（TSM）- 1.95万亿美元 Micron（MU）- 1.10万亿美元 Delta - 2022.8亿美元 Amphenol（$APH）- 1830亿美元
英文原文
The “You’re So Rich Now” $NVDA ETF. These were just some of the supply chain partners Jensen made a toast to with that comment: TSMC (https://t.co/ihVngJn3CV / TSM) - $1.95T Micron (MU) - $1.10T Delta (https://t.co/nxy6qVoi97) - $202.28B Amphenol ( $APH )- $183.00B https://t.co/khJTATSxIl
-
玻璃核心基板预计高速增长，但作者认为时间线应更早。
玻璃核心基板的爬坡真是惊人……（GNC/SEMI 预测） 从2028年起，预计复合增长率67.2%。 不过我不太同意图表，因为它应该是2026年下半年开始 / 2027年上半年爬坡，而不是2028年？ 但无论如何，整体趋势……
英文原文
What an unholy ramp for glass core substrates… (GNC/SEMI projections) From 2028 onward with 67.2% forecasted compound growth. Don’t really agree with the chart though, since it should be H2 2026 (start) / H1 2027 (ramp up) rather than 2028? But regardless overall trend for https://t.co/pRfmrZRRGY
-
量子商业化时间线可能在2030，当前投资或太早。
$IBM Starling 预计在2029年（名字挺酷）。 富士康在会议上提示量子商业化在2030年，这很有意思。2027年原型。 给 $QBTS、$RGTI、$IONQ 的兄弟们一些有用时间线。 量子不是我特别熟悉的东西，但从投资角度看，因为它很热门，也许2028年下半年、2029年上半年是提前布局量子的好时间，除非出现某种随机突破。 现在发生的事情看起来太早了。通常在原型阶段充当被稀释对象并不是好主意。 只是分享一些我看到的有趣笔记，关于量子计算商业化时间线大概在2030年。
英文原文
$IBM Starling was est. 2029 (cool name btw). Very interesting Foxconn is flagging quantum commercialization in 2030 from their conference. 2027 prototyping. Just some helpful timelines for the $QBTS $RGTI $IONQ bros out there Quantum is something I’m not as familiar with, but just in investing standpoints since it’s popular, maybe h2 2028, h1 2029 is a good time for quantum frontrunning unless there’s some random breakthrough? Whatever is happening now looks way too early. Typically not a good idea to serve as the dilution in the prototyping phases. But just sharing interesting notes I came across in terms of 2030 timelines for quantum computing commercialization timeline ests.
-
富士康称CPO交换机Q3开始、2026年1万台，顺欣可能是光学子公司。
根据富士康股东会： CPO 交换机产品预计第三季度开始。2026年1万台，明年爆发式增长。 目标/预计成为全球第一。 有人记得富士康哪家子公司负责其先进光学工作吗？ 咳咳……Shunsin（6451）。 很多……
英文原文
Per Foxconn shareholder meeting: CPO switch products expected to begin Q3. 10K units 2026, explosive grow to next year. Aims/expected to be #1 globally. Anyone remember which Foxconn subsidiary handles their advanced optical work? cough.. cough.. Shunsin (6451). Lot of https://t.co/kO55eyS2LJ
-
认为没有可信人士质疑Win Semi扩产能力。
回复 @LeeJem0X0：我没见过任何可信人士质疑 Win Semi 的扩产能力。 任何声称不可能的人都是在捣乱。你无法准确估计2027/2028年的产能和分配。 尤其是 Win 刚做了私募来扩大运营规模。
英文原文
@LeeJem0X0 I haven’t seen anyone credible doubt Win Semi’s capability to scale? Anyone claiming it’s impossible are trolling, you won’t be able to accurate estimate capacity + allocations for 2027/2028. Especially as win just did a private placement to expand operational scale.
-
JBL证明SIVE作为1.6T LRO唯一激光供应商需求强劲。
回复 @halldj00：你已经有 $JBL 这个例子，确认了过量需求，并且 $SIVE 预计是其 1.6T LRO 的唯一激光供应商。 我对此并不担心。更多是看 Jabil 或 Ayar，然后看到一旦它们放量，规模如何回流到 Sivers。 而且我会……
英文原文
@halldj00 You already have $JBL as one example, confirming excessive demand with $SIVE expected sole source laser provider for their 1.6T LRO. Don’t really have concerns there. It’s more about looking Jabil or Ayar, then seeing that scale flow back to Sivers once those ramps. And I’d
-
JBL收购部分英特尔光收发器业务，使其能与SIVE合作。
回复 @onlyonexhj：$JBL 收购了 $INTC 光收发器部门的一部分，这就是他们能和 $SIVE 做现在这些事情的方式。
英文原文
@onlyonexhj $JBL acquired part of $INTC optical transceiver division, which is how they’re able to do what they’re doing with $SIVE.
-
SIVE财报电话会看多：JBL合作带来更多光收发器请求。
我花了更多时间看 $SIVE 财报电话会记录，非常看多： - $JBL 可插拔合作为 $SIVE 带来了更多光收发器请求。 所以我猜也许是 Innolight/Eoptolink 和其他大玩家。 - 在 Win Semi 之外还有更多激光产能……
英文原文
Had some more time go through $SIVE earnings transcript, it’s very bullish: - $JBL pluggable partnership led to more optical transceiver requests for $SIVE So maybe Innolight/Eoptolink and other large players are my guess. - More laser capacity on top of Win Semi with more https://t.co/ZaptyNiSv1
-
认为77%增长只是CPO超级周期起点，2027下半年才是真正放量拐点。
回复 @RyanU_1F42B：这是77%的增长，字面上来自2026年上半年 CPO 超级周期的开端。 我预计随着我们接近2027年下半年，也就是真正的 CPO 规模化拐点，这个数字会继续指数级复合。
英文原文
@RyanU_1F42B This is 77% growth, from literally the start of the CPO supercycle H1 2026. I’d expect that number to just keep on compounding exponentially as we approach H2 2027, which is the true inflection of scale up CPO.
-
Ayar与Wiwynn公告对SIVE CPO机架级部署潜力重大。
Ayar 今天与 Wiwynn 的公告，对 $SIVE 在 CPO -> 机架级部署方面可能非常重要。 Wiwynn 的云客户包括 $AMZN、$META、$MSFT。 而且他们一直在谈 $GOOGL TPU 部署。 我认为，仅供参考架构来看，每个机架大约需要512+个 SuperNova 光源。 所以如果 $SIVE 是主要激光阵列供应商（我们预计如此，因为 Macom + Lumentum 已从 Ayar 网站移除）。 即使是中等规模的机架部署，也会对收入非常有意义。 这只是 $SIVE / Ayar / Wiwynn 当前机架级商业化潜力，还不会体现在收入财务中。
英文原文
Ayar’s announcement today with Wiwynn is potentially very material for $SIVE regarding CPO -> rack scale deployments. As Wiwynn cloud clients include $AMZN, $META, $MSFT. And they’ve been in talks for $GOOGL TPU deployments. I think just for some reference architectures it’s around 512+ supernova light sourc a rack. So if $SIVE is the primary laser array supplier (which we expect, given Macom + Lumentum was removed from Ayar’s site). Even modest rack deployments would be very meaningful for revenue. This is just rack scale commercialization potential right now from $SIVE / Ayar / Wiwynn, which won’t show up in revenue financials yet.
-
MRVL财报验证CPO主题和Celestial放量，间接指向SIVE激光需求。
$MRVL 财报是更广泛 CPO 主题的看多信号（也包括 $SIVE 作为可能的激光供应商）。 - “Scale-up interconnect 是 AI 基础设施中最新、最具战略意义的新兴机会之一。” CPO 主题 go brrr。 - 确认 Celestial 被 T1 超大云厂商选中用于 scale-up。 我过去发现过 Celestial 的 $AMZN 认股权证……所以很可能是 Amazon。 - 明年带 Celestial 的 scale-up 光学收入应该超过此前约1.5亿美元预期的2倍。 前瞻收入爬坡预期 go brrr。 - Celestial 团队加上 $MRVL 光学团队是“全垒打”。 Marvell 把 Celestial 视为增长向量，上游 Celestial 供应商受益。 - $MRVL 现在专注于把 Celestial 推向高量制造。 这是放量指标。 如果你不记得，此前有 OSINT 映射显示 $SIVE 直接对应 Celestial，而不是通过 $POET。 所以 Celestial 的前瞻增长，是 Sivers 激光器的放量指标。
英文原文
$MRVL earnings were a bullish indicator on the broader CPO theme (and $SIVE as the likely laser supplier). - “Scale-up interconnect represents one of the newest and most strategically important opportunities emerging in AI infrastructure.” CPO thematically go brrr - Confirmation Celestial was selected by T1 hyperscaler for scale up. I’ve found Celestial $AMZN warrants in the past)… so probably Amazon. - Scale-up optics revenue next year should be more than 2x prior ~$150M outlook with Celestial Forward revenue ramp expectations go brrr. - Celestial team plus $MRVL optics team was a “home run”. Marvell sees celestial as growth vector, upstream celestial suppliers go brr - $MRVL is now focused on bringing Celestial to high volume manufacturing. Volume ramp indicator If you don’t recall, there was OSINT mapping $SIVE directly to Celestial, not through $POET. So Celestial forward growth is a volume ramp indicator for Sivers lasers.
-
认为SIVE/SOI/XFAB可能获得EU Chips Act 2资金。
是的，如果看到 $SIVE、$SOI 和 $XFAB 获得 EU Chips Act 2 资金，我不会惊讶。 很高兴我在那里所有多头（除了 $IQE）都列在蓝图中，我之前研究时还没看到！ 不过正式公告看起来推迟到下周了。
英文原文
Yeah would not be surprised to see $SIVE, $SOI, and $XFAB funding with EU Chips Act 2. Glad to all my longs there (aside from $IQE) listed on the blueprint, I didn’t see that earlier in my research! Looks like formal announcement got pushed back to next week though. https://t.co/2SdG4Gocql
-
EU Chips Act 2预计聚焦光子学，SIVE是欧洲领先光子/激光公司。
提示一下，EU Chips Act 2 是今天，5月27日。 这次预计会专门针对光子学。 欧洲正好有一家位于瑞典的领先光子/激光公司叫 $SIVE。 它也获得了美国 CHIPS Act 资金，并且对西方供应链至关重要。
英文原文
FYI, EU Chips Act 2 is today, May 27th. This one is expected to be targeted specially around photonics. There happens to Europe’s leading photonics/laser company called $SIVE over in Sweden. That also received US CHIPS act funding and is critical to Western supply chains. https://t.co/Ijhnp3wIoq
-
暗示有一个与NVDA相关但不像NVTS直接被提到的功率半导体标的。
有一个非常有吸引力的名字被人点出来了。 我最后为了功率半导体敞口建仓。 它与 $NVDA 高度相关，但不像 $NVTS 那样被直接提到。 有人能猜到吗？
英文原文
There’s one very compelling name someone called out. That I ended up taking positions in for power semi exposure. Heavily tied to $NVDA but not directly mentioned like $NVTS. Can anyone guess?
-
认为稳懋3105将成为激光扩产瓶颈，并提到顺欣、上诠、MSSCorp。
Win Semi（3105）几乎从不出现在光子学分析师报告中。 但明年它们很可能会作为激光器扩产的重要瓶颈出现。 很高兴看到我的台湾多头里，Shunsin（6451）开始起势。 Foci（3363）、MSSCorp（6830）也应该开始得到……
英文原文
Win Semi (3105) is almost never mentioned in photonics analyst reports. But they’ll probably show up as an important bottleneck for scaling lasers next year. Glad to see Shunsin (6451) start picking up steam from my TW longs. Foci (3363), MSSCorp (6830) should start getting https://t.co/BOYJl4tiTq
-
JBL称1.6T LRO三到十个月内量产且需求过量，SIVE可能是唯一激光供应商。
$JBL 在炉边谈话中直接宣布…… 其 1.6T LRO 将在3-10个月内量产，并且需求过量。 $SIVE 很可能是这一特定光收发器的唯一激光供应商。 Ayar 最近融资5亿美元用于放量，而 $SIVE 是主要/……
英文原文
$JBL literally announced in their fireside chat… Mass production of their 1.6T LRO with excessive demand in 3-10 months. $SIVE is likely sole source laser supplier for this specific optical transceiver. Ayar raised $500M for volume ramp recently, and $SIVE is the primary / https://t.co/e9zTspSgFk
-
中文长文：列出多个实质性垄断瓶颈，并讨论SIVE、HIMX、Foci等design-in/out风险。
几个值得重点关注的“实质性垄断”标的： - MSSCORP（6830）：在检测和 CPO 良率把控上构筑了极深的专利护城河。 - $SOI：主导绝缘体上硅（SOI）衬底市场。 - NGK（5333）：稳拿薄膜铌酸锂（TFLN）晶圆核心技术。 - $AXTI：把控磷化铟（InP）衬底等上游关键材料。 像讯芯（Shunsin）这类公司其实很难被轻易颠覆，毕竟背靠富士康，而富士康本身就深深扎根于众多核心供应链的腹地。 $SIVE 的逻辑也极其相似。他们已经成功打入了众多顶尖 CPO 架构的设计体系，抱紧了 Ayar、Lightelligence（壁仞供应商）、Lightmatter、Celestial 等行业领军者。 相比之下，个人认为 $HIMX 或 Foci 未来面临被踢出局的风险最大，很可能被台积电光学部门采钰（Visera 6789）这类巨头垂直整合。不过未来两三年，借助 CPO 相关 FAU 和无源器件，它们眼前依然有巨大赚钱机会。
英文原文
几个值得重点关注的“实质性垄断”标的: - MSSCORP (6830)：在检测和 CPO 良率把控上构筑了极深的专利护城河。 - $SOI：主导绝缘体上硅 (SOI) 衬底市场。 - NGK (5333)：稳拿薄膜铌酸锂 (TFLN) 晶圆核心技术。 - $AXTI：把控磷化铟 (InP) 衬底等上游关键材料。 像讯芯 (Shunsin) 这类公司其实很难被轻易颠覆，毕竟背靠富士康，而富士康本身就深深扎根于众多核心供应链的腹地 🏭 $SIVE 的逻辑也极其相似。他们已经成功打入 (design in) 了众多顶尖 CPO 架构的设计体系，抱紧了 Ayar、Lightelligence (壁仞的供应商)、Lightmatter 以及 Celestial 等 众行业领军者的大腿 相比之下，个人认为 $HIMX (奇景光电) 或 Foci (上诠) 未来面临被踢出局 (design out) 的风险最大，很有可能会被台积电的光学部门采钰 (Visera 6789) 这类巨头直接垂直整合。不过话说回来，在未来两三年内，借助 CPO 相关的光纤阵列 (FAU) 和无源器件，他们眼前依然有 波巨大的赚钱机遇
-
UDN报道TSM COUPE和AI光学市场，Foci和MSSCorp可能受益。
看，一篇新的 UDN 文章谈 $TSM COUPE + AI 光学市场。 机构投资者看好台湾公司： FOCI（3363）和 MSSCorps（6830）预计受益。 不过……这个“机构”是不是应该改成 X 上的散户投资者？ 关于 FOCI：“其 FAU 产品至少已与台积电合作开发三年，规格已推进到1.6T、3.2T，后续量产日程成为市场观察重点。” 关于 MSS：“切入硅光子和 CPO 检测分析，主打‘光损耗检测’技术……并推出硅光子测试平台，锁定研发、工程验证、失效分析和少量多样化测试应用。” 这仍然是两家极小的公司……像 MSS 这样的名字甚至没有被高盛提到，尽管新闻声称有相关性。 所以感觉是 X 很早发现了这个，主流机构看到这类确认后可能即将进入。
英文原文
Oh look, a new UDN article about $TSM COUPE + AI optical markets. Institutional investors are optimistic that Taiwanese companies: FOCI (3363), and MSScorps (6830) are expected to benefit But... that "Institutional" should be changed to retail investors on X? For FOCI: "Its FAU products have been developed in cooperation with TSMC for at least three years, and the specifications have been promoted to 1.6T, 3.2T, and the subsequent mass production schedule has become the focus of market observation." For MSS: "cut into the silicon photonics and CPO detection analysis, the main "light loss detection" technology ... and launch of the silicon photonics test platform, lock in research and development, engineering verification, failure analysis and a small number of diverse test applications." These are still two extremely small companies... and things like MSS weren't even mentioned by GS, despite the news claims about relations. So feels like X found this early, mainstream institutions are likely about to enter after seeing this sort of confirmation.
-
Sivers通过Aeva成为Boston Dynamics的激光供应商，CPO技术是物理AI开发核心
有趣的事实：同一家公司经常出现在不同的供应链中。一个可能的例子是：$SIVE 作为上游激光供应商，通过以下路径供应给 Boston Dynamics：Sivers -> $AEVA FMCW（CW DFB 激光器）-> LG Innotek -> Boston Dynamics。我个人最初其实是因为4D AI而看好Aeva的。偶然发现Sivers是他们4D FMCW激光雷达的高置信度激光供应商。所以你实际上可以在光子学领域获得机器人曝光，而用于超大规模数据中心AI的同款CW激光器也在使用。虽然近期收入增长可能是$SIVE通过Aeva为$NVDA自动驾驶汽车相关架构供应激光器批量增长。人形机器人可能要到2028年？你总能得到更多间接曝光，比如MU的内存或$INTC的边缘CPU，但当然也有更直接的曝光方式。我觉得之前已经覆盖过很多名字了，比如$VPG或Harmonic Drive。但可笑的是，CPO玩家如$SIVE其实是前沿物理AI开发的核心部分。
英文原文
Fun fact: Lot of the same companies are often used across different supply chains. One likely example is: $SIVE as the upstream laser supplier to Boston Dynamics via: Sivers -> $AEVA FMCW (CW DFB lasers) -> LG Innotek -> Boston Dynamics. I actually personally liked Aeva for 4D AI first. Just so happened to find out Sivers was their high confidence laser supplier for 4D FMCW lidar. So you actually get robotics exposure with photonics while the same CW lasers used for hyperscaler AI DCs. Near term revenue ramp though it's probably $SIVE supplying laser volume ramp for $NVDA self-driving car related architectures though Aeva. Humanoids are probably later in 2028? You can always get more indirect exposure like MU with memory or $INTC with edge CPUs, but of course there's more direct exposure out there. Think I've already covered a lot of names in the past like $VPG or Harmonic Drive. But hilariously enough CPO players like $SIVE are a core part of frontier physical AI development.
-
黄仁勋预计2030年AI资本开支每年3-4万亿，应持有AI王国关键瓶颈。
根据 $NVDA 黄仁勋预测，到2030年 AI 资本开支预计将达到每年“3到4万亿美元”。 你还不够看多。 保持敞口并拥有 AI 王国钥匙，可能是个好主意： -> $AXTI 控制光子学材料建设。 -> $SOI 通过硅光子控制 AI 建设。 -> $SIVE 控制 CPO 的激光瓶颈。 -> $IQE 控制西方光子学外延片供应链。 这些公司最初都很小，但预测中的万亿美元资本开支正逐渐向它们上游流动。 其他行业也有许多类似例子： -> AI 资本开支流向 $NBIS 等 Neocloud。 -> AI 资本开支流向 $MU 和 $SNDK 等存储。 过去20年许多“商品”材料或“科学项目”，现在突然出现指数级 TAM 扩张。 我们正在见证人工智能 + 物理 AI 带来的下一次工业革命。
英文原文
AI capex spend is expected to go to "$3 to $4 trillion annually" by 2030 from $NVDA Jensen Huang projections. You're not bullish enough. And it might be a good idea to stay exposed + own the keys of the AI Kingdom: -> $AXTI controls the materials buildout with photonics. -> $SOI controls the AI buildout with silicon photonics. -> $SIVE controls laser chokepoints for CPO. -> $IQE controls Western epiwafer supply chains for photonics. All these started off as tiny companies, yet the trillions of projected capex gradually upward to them. There's many more in other industries as well. -> AI Capex flows to Neoclouds like $NBIS. -> AI Capex flows to memory like $MU and $SNDK. And many of the "commodity" materials or "science projects" for the past 20 years now a sudden shift in exponential TAM expansion. We're witnessing the next industrial revolution with Artificial Intelligence + Physical AI.
-
分享周末研究的热管理和功率半导体供应链主题股票清单
随便分享一些我周末研究的随机主题： -> 功率半导体、集成电压调节器、多相控制器、CDU、冷板、歧管、泵、阀门、热交换器、背门系统、冷水机。 $POWL、$NVT、010120 KS、267260 KS、2308 TT、$VICR、$POWI、宁德时代、阳光电源 300274 C -> 热管理相关（ALFA SS、ASTK DC、大金 6367 JP）、$CARR、MTRS SS） 这些只是我之前聊过的之外的一些，比如 $NVTS 或 $HPS.A。 它们都有点已经被定价了，但现在是个好的起点，可以帮你找到那些还没被定价的更重要玩家。 然后你会问：这些产品最有可能依赖哪一家公司？ 我个人对能源/散热/电力这个交易领域不太熟悉，我很高兴承认这一点，因为一切都是学习的过程。 但通常即使我发现了一些有趣的东西，我也觉得它不如继续持有CPO多头更有说服力。学习更多总是很有趣的。 一直在寻找SRAM相关的敞口，我觉得私募市场比 $CBRS 更好。
英文原文
Just random themes im looking im having a good time looking into over the weekend: -> power semis, integrated voltage regulators, multiphase controllers, CDUs, cold plates, manifolds, pumps, valves, heat exchangers, rear-door systems, chillers. $POWL, $NVT, 010120 KS, 267260 KS, 2308 TT, $VICR, $POWI , CATL, Sungrow 300274 C -> thermal stuff (ALFA SS, ASTK DC, Daikin 6367 JP), $CARR, MTRS SS) These are just ones aside from the ones I've talked about eg. $NVTS or $HPS.A. They're all kinda more priced in but now but it's a good starting point in terms of finding more important players that's not priced in. Then you try and look: which single company do most of these probably rely on? I'm personally less familiar with the energy/cooling/power trade, which I'm happy to admit since everything is a learning process. But normally even if I find something interesting, i don't find it more compelling than staying in CPO longs. Always it's always fun learning more. Was always looking for SRAM-adjacent exposure, think private markets are better to go then $CBRS?
-
认为SIVE员工少但博士比例高，能构建极有价值IP。
瑞典人总是说： “$SIVE 员工这么少，怎么能做到所有这些？”相较于 $LITE？ 结果是……如果你整个员工队伍约四分之一都是博士，你就能构建世界上最有价值的 IP。
英文原文
People in Sweden always like: “How does $SIVE have so little employees and can do all of this” vs. $LITE? Turns out… You can build the most valuable IP in the world if ~1/4th your entire workforce are PHDs. https://t.co/oVlJJ9OEaZ
-
美国CHIPS Act资金显示美国政府强力支持SIVE，涉及Golden Dome和RTX防务供应链。
市场有点意识到美国政府正在大力支持 $SIVE，正如直接美国 CHIPS Act 资金所显示。 因为你能在 $YSS Golden Dome 供应链（来自 allspace）和 $RTX 防务供应链（CHIPS Act）中看到 Sivers。 不过是的，Ericsson 和 Sivers 可能是从美瑞技术 MOU 中受益更明显的两个。
英文原文
Markets are kinda realizing US Gov is heavily backing $SIVE as you can see with direct US CHIPS Act funding. Since you see Sivers in $YSS Golden Dome supply chains (from allspace), and $RTX defense supply chains (chips act). But yeah Ericsson and SIvers are probably the two more noticable ones benefiting from US x Sweden tech MOU.
-
提到Intel EMIB先进封装供应链中AP Memory 6531。
关于 $INTC EMIB 先进封装供应链： 如果你以为上 AP 历史或 AP 化学之后不会更糟： 显然有家公司叫 AP Memory（6531）。市值48亿美元，已经获得认证。 如果有人想看一下。我现在在吃火锅，没有足够时间做尽调。
英文原文
For $INTC EMIB advanced packaging supply chains: If you thought it couldn’t get worse after taking AP History or AP Chemistry: There’s apparently a company called: AP Memory (6531). $4.8B MC that got qualified. In case anyone wanted to take a look. I’m eating hot pot rn so not enough time for DD.
-
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.
-
列出多家台湾功率半导体低市值潜在玩家。
功率半导体现在似乎很受欢迎。报告中标出了这些名字： Pan Jit（2481）约16亿美元 Eris Tech（3675）约5.32亿美元市值 Advanced Power Electronics（8261）约5.65亿美元市值 Inergy Technology（6693）约2.02亿美元市值 Potens（7712）约1.70亿美元 这些功率半导体制造商可能扮演重要角色。 市值都很低，也许其中几个有高上行空间，所以我现在可能会对每一个做更多研究。
英文原文
Power Semis seem to be very popular now. These names got flagged in reports: Pan Jit (2481) ~$1,6B Eris Tech (3675) ~$532M MC Advanced Power Electronics (8261) ~$565M MC Inergy Technology (6693) ~$202m MC Potens (7712) ~$170M for power semiconductor makers that might play an important role. All really low MC, maybe high upside in a few, so I'll probably do some more research on each one now.
-
AMD投资台湾生态100亿美元以上，利好封测、ODM、PCB及二阶光学受益者。
正如我刚说的： $AMD 向台湾生态投资100亿美元以上（用于保障产能/扩展基础设施）。 AMD 正与 $ASX 和 SPIL 合作。 以及 ODM 伙伴如 $SANM、Wiwynn（6669）、Wistron（3231）和 Inventec（2356）。 PTI、Unimicron、AIC、Nan Ya PCB、Kinsus 也被点名为直接受益者。 当然，像 Foci 或 MSScorps 这样的二阶受益者也会受益。 总的来说，AMD 新闻利好台湾供应链。
英文原文
Just as I say this: $AMD invests $10B+ into Taiwan ecosystem (for securing capacity/scaling infra) AMD is collaborating with $ASX and SPIL. As well as ODM partners like “ $SANM, Wiwynn (6669), Wistron (3231) and Inventec (2356). PTI, Unimicron, AIC, Nan Ya PCB, and Kinsus were also name dropped as direct beneficiaries. And of course second order beneficiaries like Foci or MSScorps also benefit. But bullish on Taiwan supply chains all around from $AMD news.
-
认为6月1日NVIDIA台北Computex/GTC演讲可能利好台湾光学生态。
如果你想提前布局下一个重大催化： $NVDA 6月1日 Computex/GTC Taipei 主题演讲，应该会对台湾光学生态非常看多？ 就像我在 Nvidia 美国活动时做多 $TSEM 和其他公司一样。 Shunsin、Foci、Nextronics、MSScorps，以及我 Nvidia 光子学生态组合里的很多名字，可能会在那里被关键提及或获得间接催化。
英文原文
Just in case you want to frontrun the next major catalyst: $NVDA June 1 Computex/GTC Taipei keynote should be heavily bullish Taiwan optical ecosystem? Just like when I went long on $TSEM and others in Nvidia’s US event. Shunsin, Foci, Nextronics, MSScorps, and a lot of my Nvidia photonics ecosystem port might be critical mentions or have indirect catalysts there.
-
认为SIVE可能是JBL特定1.6T LRO唯一供应商，因架构独特。
回复 @ShakespeareDad：鉴于架构独特性，$SIVE 很可能是 $JBL 这款特定 1.6T LRO 的唯一来源。 它不是标准 CW 激光器。
英文原文
@ShakespeareDad $SIVE is likely sole source for this specific 1.6T LRO from $JBL given the architectural uniqueness. It’s not a standard cw laser.
-
JPM炉边会显示JBL 1.6T LRO量产早于预期，SIVE激光架构形成护城河且需求近乎无上限。
$SIVE 与 $JBL 的 1.6T 光收发器量产： 根据 JPMorgan 炉边谈话，现在比预期更早。 他们宣布的内容： > $SIVE 激光架构暗示“相当显著的护城河”。 > 其 1.6T 存在极端需求，此前市场并不知道具体量级。 > 时间线快于预期，将收入实现窗口提前。 声明内容： 1. Jabil 的 1.6T LRO “将在未来1到4个月进入不同认证”。 “认证可能需要2到6个月。” 考虑到现在是2026年上半年5月，量产和收入实现可能在3到10个月内开始。 以6.5个月中点看，就是2026年底。 很多此前估计是2027年下半年。 3. 架构上，它“比当前1.6T 功耗曲线低约11千瓦”。 超大云厂商会喜欢听到这个，而这就是通过 $SIVE 作为关键光子学 chokepoint 证明的竞争差异化和相当显著护城河。 4. “在这一点上，不是份额问题。真正的问题是跟上整个市场的自然增长。” 这再次显示巨大需求已经超过供应。 含义是，关键变成 Sivers + Jabil 能一起生产多少，因为它们生产的任何东西都会被买走。 这种表述对 Sivers 相对于当前市值的收入极其重要。 市场此前不确定 Sivers x Jabil 的确切需求量和商业时间线。 Jabil 刚刚公开确认，使用 $SIVE 激光的 1.6T LRO 需求本质上没有上限。
英文原文
$SIVE mass production for 1.6T optical transceivers with $JBL: Is now earlier than expected per JP Morgan Fireside chat. Here's what they announced: > "Relatively dramatic moat" implied with $SIVE laser architectures > Extreme demand for their 1.6T, which was previously unknown in terms of volume. > Faster than expected timelines, pulling revenue realization window forward. The statement: 1. Jabil’s 1.6T LRO: "Goes into different qualifications across the next 1 to 4 months" “The quals can take anywhere between 2 to 6 months" Given its May H1 2026, mass production and revenue realization could begin in anywhere between 3 months to 10 months. So late 2026 with 6.5 month midpoint. Lot of former estimates were H2 2027. 3. Architecturally it's "which is about 11 kilowatts dramatically lower than current 1.6T power profiles" Hyperscalers would like to hear this, and this is the competitive differentiation + relatively dramatic moat proven with $SIVE as the critical photonics chokepoint. 4. "At this point, it's not about share. It's really about keeping up with the organic growth of the entire market." Again this shows that the enormous demand has outstripped supply. The implications are that it’s more of a matter of how much Sivers + Jabil can build together, as anything they make would be bought. This type of statement is just incredibly material for Sivers revenue relative to their current marketcap. The market was previously uncertain about the exact volume demand and commercial timeline from Sivers X Jabil. Jabil just publicly confirmed that the demand for their 1.6T LRO with $SIVE lasers is essentially uncapped.
-
出口管制InP会伤及IQE/LITE等下游，而管制人形机器人瓶颈连带伤害较小。
我在思考对 Rasa/NCI 的出口管制，但如果总体上暂停中国 InP 衬底生产，会对 $IQE、$LITE 等下游造成很多附带损害。 但通过 Ulvac 和其他瓶颈暂停中国人形机器人生产，并不会造成同等程度的附带损害。
英文原文
I was thinking about export controls on Rasa/NCI but that would cause a lot of downstream collateral damage on $IQE, $LITE, etc. if you halted China's InP substrate production in general. But halting China's humanoid production with Ulvac and other chokepoints doesn't really do as much collateral harm.
-
称中国曾管制日本上游InP衬底以阻断西方AI光子供应链，建议美日反制。
回复 @TigerMeditation：兄弟，中国在1月我发 $AXTI 帖后，字面上对日本上游 InP 衬底实施出口管制，以阻断西方 AI 光子供应链。 也许美国/日本该用便宜内裤关税以外的方式反击了？
英文原文
@TigerMeditation Bro China literally export controlled Japan on upstream InP substrates back in Jan, after my $AXTI post, to halt Western AI photonic supply chains. Maybe it's time for US/Japan to fight back, in ways other than tariffs on cheap underwear??
-
建议美日将Ulvac列入出口管制，以打击中国人形机器人稀土磁体设备瓶颈。
一个随手想法： 美国政府/日本应该把 Ulvac（6728）列入出口管制名单，让它们停止向中国发货。 这样就能扰乱中国整个人形机器人计划（因为它是“军民两用应用”）。 作为对上游 InP/稀土相关出口管制的报复。 对袜子出口加45%关税没什么用，但扰乱中国机器人计划，确实有用。 Ulvac 是日本在钕磁体上的主要 chokepoint。 “Ulvac 长期通过其中国子公司制造稀土磁体真空熔炼炉。” 所以也许该让这件事停下来？ 如果它们停止出货，你可以打断宇树和其他中国人形机器人量产供应链。 只有当中国允许稀土出口重新流动时，出口才能继续。
英文原文
Just a shower thought: US Gov/Japan should put Ulvac (6728) on the export control list so they stop shipping to China. That way you can disrupt China's entire humanoid program (since it's "Dual Use Military Applications"). As retaliation over upstream InP/rare earth related export controls. Adding 45% tariffs on sock exports won't do much, but disrupting China's robotics program, sure. Ulvac is Japan's major chokepoint over neodymium magnets. "Ulvac has long manufactured rare-earth magnet vacuum melting furnaces through its Chinese subsidiary". So maybe put that to rest? If they stop shipping, you can break Unitree's and other humanoid mass production supply chains in China. And only if they allow rare earth exports to start flowing again, then exports can continue.
-
梳理Starlink/Amazon Kuiper LEO PCB供应链和下一阶段光学星间链路瓶颈。
Starlink / Amazon Kuiper LEO 想法： -> Compeq（2313）：“已获得包括 Starlink 和 Amazon Kuiper 计划在内的领先运营商订单” - 96亿美元市值。 “分析师称，PCB 供应商涌入可能形成‘玩家众多、蛋糕有限’动态，提高新进入者壁垒。” - Unitech Printed Circuit Board - Chin-Poon Industrial、K Tech - ACCL “都预计在赢得合同时面临越来越大困难。” 接下来会发生什么： - 光学星间链路（OISL）。 瓶颈想法： - 高频所需的下一代 HDI 板越来越依赖特殊材料，导致上游供应链交期更长。 所以值得研究这个上游领域。
英文原文
Starlink / Amazon Kuiper LEO ideas: -> Compeq (2313): "has secured orders from leading operators, including Starlink and Amazon's Kuiper program" - $9.6B MC "Analysts say the influx of PCB suppliers is likely to produce a "many players, limited pie" dynamic, raising barriers for newer entrants" - Unitech Printed Circuit Board - Chin-Poon Industrial, K Tech - ACCL "are all expected to face growing difficulty in winning contracts." What's coming next: - optical inter-satellite links (OISL). Bottleneck ideas: - next-generation HDI boards required for high-frequency are increasingly dependent on specialty materials, leading to longer lead times across the upstream supply chain So good to look into this upstream area.
-
中国掌握人形机器人量产关键牌，美国应加大主权稀土供应链投资。
谁能想到中国在人形机器人量产上握着所有牌？ 如果美国真的看好 $TSLA 或 Figure 机器人计划的未来。 也许是时候开始向主权稀土供应链投入更多资金了。 我们现在做的还不够。
英文原文
Who could have thought China was holding all the cards over humanoid mass production? Really if America sees a future in $TSLA or Figure robotics programs. Maybe it’s time to start pouring more funding sovereign rare earths supply chains. Whatever we’re doing now isnt enough. https://t.co/VQXh9otLMi
-
Apple Watch 2026下半年健康监测升级或用TASC光电二极管，后续无创血糖监测可能关联SIVE激光。
这只是对时间线的确认： 2026年下半年是高端 Apple Watch 的“重大重新设计”，用于血压相关功能，使用 TASC 光电二极管（2340）。 下一部分可能是 $SIVE 与 $AAPL 的项目： “此后，Apple 的下一项健康监测能力预计将聚焦无创血糖监测。” 所以……下一轮更新很可能是 Sivers 的高度定制激光，也许在2027年下半年或2028年。 仍然有点早，但如果时间线/含义保持完整，市场可能会重视 Apple 带来的多年收入可见度。
英文原文
This is just confirmation on timelines: H2 2026 is the "major redesign of high end Apple Watches" for blood pressure related stuff using TASC photodiodes(2340). This next part is $SIVE likely project with $AAPL: "After this, Apple's next health monitoring capabilities are expected to focus on noninvasive blood-glucose monitoring." So.. the next cyle refresh is likely Sivers for hyper-customized lasers. Maybe H2 2027 or 2028. Still a bit early, but if timeline/implications are in tact, markets might appreciate multi-year revenue visibility with Apple.
-
梳理玻璃基板光刻胶、ABF瓶颈和关键矿物供应链。
一些随机笔记： 关于玻璃基板：“YC Chem 据称成为业内首家供应玻璃基板光刻胶的公司”（KR 112290）。 - 目标年底量产，通常量产 + 首发对股价有利。 “目前正与三家以上公司洽谈玻璃基板材料供应”，所以还有更多认证。 “Samyang NC Chem 也在开发玻璃基板光刻胶材料”（482630）。 - 正在送样，明年量产，是另一个玻璃基板名字。 “$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 $TSM COUPE on Substrate. "As Commercial Times highlights, demand from AI GPUs, ASICs, and CPO could trigger another supply shortage of high-end ABF substrates by 2027." So more substrate bottleneck stuff. Then there's: "China's agreement to address US concerns over rare earth and critical mineral shortages specifically yttrium, scandium, neodymium and indium" So good to look into yttrium, scandium, neodymium and indium supply chains if you want long ideas. Everything is kinda correcting right now, but just something to keep a lookout for if ur interested in more random ideas.
-
SIVE获得美国CHIPS Act 660万美元，与BAE合作用于美国防务基础设施现代化。
最新：$SIVE 获得美国 CHIPS Act 660万美元。 这是第二年项目的660万美元。 与 BAE Systems 合作：“使用 Sivers 技术现代化美国防务基础设施”。 像 $SIVE 这样小的公司获得 CHIPS Act 资金，极其看多。 因为它把 Sivers 定位为一个关键、由美国政府托底的国家安全 chokepoint。
英文原文
Just in: $SIVE gets $6.6M from the US CHIPS ACT. This is a $6.6M Year 2 Program. In collaboration with BAE Systems: "Modernizing U.S. Defense Infrastructure Using Sivers Technology" This is incredibly bullish to have such a small company in $SIVE to get CHIPS ACT funding. As it positions Sivers as a critical, US-government backstopped chokepoint for national security.