阿祖

做多Bonk 系代币 bonk GP useless @然后亏惨了

[NickSung @NickSung2017]

SIVE → POET → Lumilens → Top-3 （Amazon / Microsoft / Google 之一）
逻辑如下：

🚀Sivers 给 POET 提供 CPO 用的 CW DFB 激光器；
🚀POET 用自己的 Optical Interposer 技术做光引擎/模块，卖给 Lumilens；
🚀Lumilens 再把产品卖给 云厂商（Amazon / Microsoft / Google 之一）。

[Trade Whisperer @TradexWhisperer]

$MU $DRAM $SNDK Bank of America just put out their 2030 semi TAM.

$1.96T total.
$900B is memory. Nearly half.

And this is almost certainly outdated. It's not pricing in the Agentic AI boom yet.

[Serenity @aleabitoreddit]

$SIVE + Aeva likely coded from the new PR today and SIVE + Apple.

For Sivers lasers powering robotics / physical AI.

-> Sivers lasers powering Boston Dynamics Atlas via. LG Innotek for robotics
-> Nvidia Hyperion ecosystem w/ Sivers lasers for physical AI.

Are possible paths with Aeva.

Apple likely using Sivers also reiterated from the newer webpage:

“Optical sensing modules for wearables” from biometric and health monitoring.

They just can’t namedrop partners due to confidentiality, but consumer sensing is probably the biggest tell.

Physical AI and mass consumer production with Apple would be very material to revenue.

[Serenity @aleabitoreddit]

OpenLight (private) seems to get bigger and bigger every time I look at it.

If you’re curious about their public ecosystem outside of Advantest:

$JBL ( $SIVE partner ) for scaling PICs
$MRVL and $MXL for DSPs
$TSEM for the foundry.
$300394.SZ / TFC Optical (OSAT/subassembly)

Can often get exposure into private growth through public equities if you’re not an accredited investor.

Since optical players look very interconnected.

[Serenity @aleabitoreddit]

Fun new information discovery from Poet OSINT community:

Seems likely that $POET / $SIVE are going to power a Top-3 hyperscaler (either Amazon, Microsoft, Google).

Given a Linkedin update from Ankur Singla (CEO of Lumilens).

Who stated their customer is one of the top 3 hyperscalers with their post focusing on CPO/NPO.

With that clue, seems more likely the Sivers CW DFB light source path over other EML suppliers given it's CPO Scale Out/NPO.

If you don't remember, Sivers is the laser supplier to Poet. And Poet has purchase agreements with Lumilens.

Always fun to find major potential breadcrumbs in the wild before they're officially confirmed. (Disclosure, long Sive)

[Serenity @aleabitoreddit]

Think so. But in the meantime, I call my strategy: Diversified Losses.

With $AXTI, $SOI, $AAOI, and many others.

Had a massive drawdown recently,
CPO exposure was hit the hardest (Foci, Msscorp, etc) and adjacent names, feels bad.

Probably lesson personally, I had too much concentration in photonics vs. memory/other sectors without weighting/hedging properly.

With Soitec, there's been a few negative institutional reports that I'd disagree with.

Think AXT was hit harder in specific just because of float expansion/dilution concerns. AAOI, probably just brought down with the theme.

I can't give advice on buying, so completely up to you to make for cost averaging or entering positions.

But I do think we're still early in the Supercycle with photonics, there's bound to be corrections/crashes along the way up.

If my personal thesis is correct though, many of these names will have a major inflection point in midway through 2027 scaling up to 2028.

Markets don't typically wait to price things in advance, but some ideas might be a tad early or in the buildout given it's H2 2026 now.

Which is why it's important to build your own conviction.

[Serenity @aleabitoreddit]

Robotics is next.

Both deal count and investment amounts are skyrocketing per pitchbook March data (source: a16z)

Good thing is: the same AI DC exposure often has cross-exposure to humanoid ramp.

Like DRAM/NAND with memory (on humanoid inference/storage) or DFB lasers with photonics (FMCW LiDAR vision/sensing).

Right now most exposure is upstream component parts… or programs within large players like $AMZN or $TSLA.

So global IPO season H2 into 2027 for pure play humanoids/robotics companies is going to be fun.

[Rocky @Rocky_Bitcoin]

下半年，专注于投资机器人赛道和长持MLCC！

MLCC在机器人领域的使用量是AI服务器的10倍，三星电机5000亿供应合同，已经说明了一切！

平时有空多去深圳华强北，他们有第一手消息！🧐

[みみりん@投資系女子 @beauty_oe]

$SIVE
🌟 SIVEのサプライチェーン下流にTop3ハイパースケーラーが繋がっている可能性が非常に高くなった

・Lumileneの CEOが最初の顧客はTop-3 hyperscalerの1社と明言
• POETはLumilensと実質的な購入契約 + 共同開発をしている。
• SIVEはPOETの主要レーザーサプライヤー
※CPO向けのCW DFB レーザー

特に「CPO Scale Out/NPO」というキーワードあり。
$SIVE のCW系がEML系サプライヤーより優位になりやすいか？

Sivers ($SIVE)
　↓
　↓ CW レーザー供給
　↓
POET Technologies ($POET)
　↓
　↓光エンジン・パッケージ技術
　↓
LumiLens
　↓
　↓製品供給
　↓
トップ3ハイパースケーラー
（Amazon / Microsoft / Google のいずれか？)

[梭哈.AI @SUOHA_AI]

跟 Serenity 学美股：AI / 半导体 NAV 折价股票深度拆解

白毛股神Serenity 最近分享了一个 NAV 折价股票的列表，我认为这个选股思路非常值得研究

所以我把她提到的 8 家公司重新逐一交叉验证了一遍，发现含金量极高：

什么是 NAV 折价？

简单来说，当一家母公司持有子公司大量股权，而母公司自身的市值却显著低于这些持股资产的当前市场价值时，这部分差额就被称为 NAV 折价

---

1）WUS / 楠梓电子
持有沪电股份约 11.3% 股权，底层资产是 AI PCB / 高速 PCB 龙头

这里是最经典的 NAV 折价案例：折价深，而且 Palliser、Metrica 这类机构已经公开施压管理层，要求改善资本配置和股东沟通

我发过研究推文：https://t.co/NXo16X7nwR

2）ACMR / 盛美半导体
核心是美股母公司 ACMR 持有 A 股子公司盛美上海

盛美上海本身是半导体清洗设备核心资产，如果后续 H 股路径继续推进，就可能带来新的价格发现和估值重定价

3）Wistron / 纬创
持有 Wiwynn / 纬颖 35.46% 股权

Wiwynn 是 AI 服务器 ODM 核心玩家。Wistron 不是纯控股壳，更像是“自身 AI Server 高增长 + Wiwynn 股权支撑”的组合，所以质量很高

4）Sino-American Silicon / 中美晶
持有 GlobalWafers / 环球晶圆 46.64%
这是全球硅晶圆核心资产，NAV 折价很明显，但催化剂相对慢，更多要看半导体周期复苏、分红和资本配置

5）Priortech
持有 Camtek 约 21% 股权

Camtek 是半导体检测 / 量测设备公司，受益 HBM 和先进封装需求。折价存在，但更像 Camtek 的 holding company，不是最强事件驱动

6）Iljin Holdings
持有 Iljin Electric 42.99%

底层资产是电力设备、电缆和变压器，受益 AI 数据中心电力基础设施需求。数学折价很深，但韩国控股公司治理折价可能长期存在

7）Simmtech Holdings
持有 Simmtech Co.，方向是 PCB / 半导体封装基板

这里折价也很深，但要特别注意代码和口径：036710.KQ 是 Simmtech Holdings，222800.KQ 是 Simmtech Co.，不能把子公司市值误当成母公司市值

8）Bit Digital
持有 WhiteFiber，同时还有 ETH 资产、可转债、稀释和业务切换问题

Serenity 说它“不 clean”我觉得是对的，这个不适合作为纯 AI NAV 折价案例

所以这 8 家不是简单地“谁折价最大谁最好”

---

NAV 折价本身不是投资理由，它只是一个筛选入口

我的理解是，真正值得研究的结构是：
深度折价 + 母公司或子公司独立高增长 + 明确的价值解锁催化剂

---

比如 WUS 就很符合这个框架：

母公司市值相对其持有的沪电股份股权价值存在明显折价，而沪电本身又卡在 AI PCB / 高速 PCB 需求爆发的赛道里，同时还有 Palliser、Metrica 这类机构公开施压管理层，要求改善资本配置和股东沟通

所以重点不是单纯“它便宜”，而是便宜背后有没有成长，有没有人推动管理层释放价值，有没有可能让市场重新定价

按这个框架看，我目前最关注的是 WUS、ACMR、Wistron

WUS 是事件驱动最强，ACMR 是路径最清晰，Wistron 是基本面质量最高

---

再次感谢 @aleabitoreddit 提供的宝贵信息！！

仅个人学习分享，不构成投资建议，请DYOR

[Serenity @aleabitoreddit]

IMO photonics theme + CW laser chokepoint is goated.

It's legit like markets have short term memory loss and forgot how $LITE went from $3B -> $65B+ from 2024 to now.

Because $NVDA caused EML bottlenecks, and forced architectural changes.

We're literally seeing the same thing today with CW lasers + 1.6T/CPO shifts with Nvidia signing LTAs everywhere.

Now, $AMD + other CSPs are hunting for remaining scraps with large LTAs for CW lasers + optical components.

GS Research's ~9-10x $154B optical TAM in 2028 and near $0 -> $91B CPO TAM in just 2 1/2 years.

Don't just magically disappear from a month of trading volatility.

$AAOI sitting at ~$13B, $SIVE sitting at ~$3B, and other CW laser players look strategically very valuable.

And next year I think we'll look back and say "Why didn't I learn my lesson the first time with EML from Nvidia and pick up CW laser adjacent names!"

Then there's likely gonna be some new mini trend 1-2 years from now like microled or quantum dot and we're gonna see the same thing repeat.

Think Sumitomo's projections with CW laser share + silicon photonics being majority / dominant architecture should be correct.

I'm personally just focusing on that bottleneck as you've seen with $SOI, $TSEM, $SIVE, and others.

[Cyberpunk Sense 👑 @napoleon21st]

Speculation has arisen that the new sensor by Sivers $SIVE has been prototyped for Apple $AAPL

1. You can see here a picture of a watch that quite resembles the Apple Watch.

2. When I consult Gemini it says this technology is a perfect match for the requirements Apple has had for its next generation of Apple Watch for health tracking of blood glucose and more

3. In past corporate communications and across tech-investor forums, Sivers’ leadership hinted at an engagement with a massive US consumer electronics firm requiring immense volume quotes (speculated by analysts to be Apple for a health-sensing pipeline)

If this is indeed becomes used by the next Apple Watch Sivers $SIVE will rerate very quickly towards a 10-20 billion dollar valuation.

It is very likely tomorrow Swedish time is an explosive market day

[Serenity @aleabitoreddit]

As for $LPK: Maybe $3B-$5B seems reasonable when they fully volume ramp if I had to guess.

Feels more asymmetrical to me personally since it's just a waiting game and they have the customers for glass substrates.

Small machine supplier chokepoints usually cap out from TAM though and don't go to $20B+ unless you're ASML or hold many chokepoints like KLAC.

Not many dominant "monopolies" like these out there right before volume ramp this small:

Maybe you have stuff like:
- Aixtron (MOCVD) ~$8B
- Towa (compression bonding) ~$2B
- Techwing (memory handler/cube probes) ~$1.5B
- MSScorps (CPO inspection), but pre-ramp. $850M
- Riber (quantum MBE), very pre-ramp ~$350M

Then LPK Laser for glass core substrates (about to ramp) at ~$730M, which thematically should have more premiums than memory. (eg. major advanced packaging shift, CPO adjacent, etc).

Lot of obscure chemical monopolies I know of, but those don't get as much attention cause BOM is much lower than equipment.

But for lpk you have:

eg. “80% of customers among major global players have selected LPKF equipment”.

70% of LIDE market share target for TGV in the glass-core ramp, should be very material. (disclosure: own, the listed names above aside from techwing/aixtron, NFI).

Just kinda the path they go if you follow $AEHR at ~$3.5B now:

- <$500m: "Oh they have no customers!"
---> (probably somewhere in the middle here).
- < $1.5B: "Maybe volume comes soon!"
- $3B+ onward: "Looks like they're starting to volume ramp.

Just a lesson for the cycle of the chokepoint machine suppliers from my personal experiences.

[空空 ｜ 跟着Serenity学投资 @chuge857]

有人问 Serenity：$LPK 这只股票，你觉得最终能涨到多少？

她的回答，不只是给了一个数字，而是给出了一套判断"窄门设备供应商"市值天花板的完整框架。

先说结论：她认为 $LPK（LPKF Laser，德国股市，玻璃基板激光钻孔设备龙头）在完全量产爬坡后，市值达到 30 亿到 50 亿美元是合理的。目前市值约 7.3 亿美元，意味着潜在空间约 4 到 7 倍。

但她也说了一句很重要的话：小型设备供应商的市值，通常会被 TAM（总可寻址市场）封顶，很难长到 200 亿美元以上——除非你是 ASML 那种级别，或者像 KLA 一样同时掌握多个"窄门"。

她给出了一张横向对比表，帮助理解 LPK 在全球"窄门设备供应商"里的位置：

Aixtron（德国股市，MOCVD 化合物半导体沉积设备，碳化硅/氮化镓核心设备商）：市值约 80 亿美元，已处于量产爬坡阶段。

Towa（日本股市，压缩成型封装设备，先进封装核心设备商）：市值约 20 亿美元，量产中。

Techwing（韩国股市，存储测试搬运设备，HBM 测试关键设备）：市值约 15 亿美元，量产中。

MSScorps（台湾股市，CPO 光电共封装检测设备）：市值约 8.5 亿美元，量产前期。

Riber（法国股市，量子材料 MBE 分子束外延设备）：市值约 3.5 亿美元，极早期。

LPKF Laser（$LPK，德国股市，玻璃基板 TGV 激光钻孔设备）：市值约 7.3 亿美元，即将量产爬坡。

Serenity 的判断是：LPK 的主题溢价应该高于存储类设备，因为它同时覆盖先进封装（玻璃基板是下一代封装基板）和 CPO 相邻产业链两个方向，而这两个方向的市场预期溢价都高于纯存储。

LPK 的核心竞争壁垒有两个数字值得记住：

全球主要客户中，80% 已选择 LPKF 的设备。
在 TGV（Through Glass Via，玻璃通孔，玻璃基板的核心加工工艺）的 LIDE 激光技术市场，LPKF 的目标市占率是 70%。

这两个数字说明，LPK 不是在和竞争对手抢市场，而是在等待一个它已经基本锁定的市场从零开始放量。

最后，Serenity 用 $AEHR（美股，SiC 碳化硅晶圆级测试设备，目前市值约 35 亿美元）来说明这类"窄门设备供应商"的典型市值成长路径：

市值低于 5 亿美元时：市场说"他们根本没有客户"。 市值 5 亿到 15 亿美元时：市场说"也许量产快来了"。 市值超过 30 亿美元时：市场说"看来量产真的开始了"。

她说 LPK 目前大概处于这条路径的中间阶段。
对中国 A 股投资者，这套框架有非常直接的参考价值。
A 股半导体设备板块里，同样存在大量"窄门设备供应商"——它们掌握某一道关键工艺的核心设备，客户高度集中，替代难度极高，但在量产爬坡之前市值往往被严重低估。

判断这类标的的核心问题只有两个：第一，它的技术壁垒是否真的构成"窄门"（即客户没有其他选择）？第二，量产爬坡的时间节点是否已经清晰可见？

两个问题都能回答"是"的时候，往往就是最佳的布局窗口。
以上内容仅作海外前沿产业观点译介，不构成任何投资建议。

[P Equity Research 📰 @pequityresearch]

Jefferies: Memory Outlook

> Memory pricing is expected to rise sharply in the near term—climbing 40%–50% QoQ in 3Q26 and another 30%–40% QoQ in 4Q26.

> Price hikes are likely to continue into 2027, with a projected 40%–45% YoY increase due to zero wafer capacity growth.

> Cloud Service Providers (CSPs) are locking down 50% of total capacity (potentially rising to 70%) by signing 2-year LTAs that require a 40% prepayment.

> No CE players have signed these LTAs. Because CSPs are swallowing up supply, CE players face severe pressure and a substantial drop in available supply heading into 2027.

> China will not threaten the current memory bull market in 2026 or 2027 due to a widening technology gap.

> CXMT's DRAM technology lags 1.5 to 2 generations behind global leaders. Without EUV lithography, they cannot upgrade to DDR6 or HBM3E.

> While China's current expansion only impacts low-end segments, its NAND technology is expected to become more globally competitive and could catch up by 2028.

> On HBM4, Samsung is expected to erase its previous tech disadvantage as the industry transitions to hybrid bonding. Additionally, checks suggest Samsung’s 4nm base die for HBM4 offers the best industry performance, giving them a competitive edge.

[潮向研究 @chaoxiangooo]

【研报解读】明年 HBM 价格必翻 2 倍以上，继续看涨存储

存储行业正在经历一场前所未有的结构性撕裂。

同一家晶圆厂，造普通DRAM的营收，是造HBM的2倍。 毛利，接近3倍。AI时代最性感的存储芯片HBM，赚钱能力反而不如"老土"的普通内存。

怎么回事？

伯恩斯坦6月22日发了一份全球存储研报，把这个矛盾拆得很透。

从2025年Q3到2026年Q2，普通DRAM价格涨了约4.5倍。而HBM因为绑定年度长协，价格几乎没动。

三星和美光在今年Q1财报电话会上已经明说了：非HBM的DRAM利润率已经超过了HBM，而且差距还在扩大。

这意味着什么？

存储厂没有动力把产能分给HBM。

要让HBM的每片晶圆收入追上普通DRAM，HBM价格需要涨3倍。

但存储厂也不傻，定价过猛会损害整个AI生态。所以伯恩斯坦给出的判断是：2027年HBM价格全年平均涨2到2.5倍。

即便涨了这么多，HBM的盈利水平仍然低于普通DRAM。只是差距会大幅收窄。

但真正的冲击藏在下一环。

HBM不像普通DRAM，云厂商不能直接找存储厂买——它是焊死在英伟达GPU里的，是英伟达销售成本的一部分。

假设英伟达在VR200机柜上维持75%毛利率，那HBM涨价的部分，英伟达需要放大4倍去定价。

伯恩斯坦的测算：HBM原本占机柜售价约5%，涨价后变成6%。但如果英伟达要守住75%毛利率，机柜售价需要涨24%。

对于一个部署VR200的AI数据中心，仅HBM成本传递一项，总资本开支就要上升4%到15%。

叠加普通DRAM和NAND的涨价（约14%），全部加起来，云厂商的AI资本开支需要比原来高出约30%。

伯恩斯坦把这叫"重新校准"。他们认为云厂商不会因此放缓AI投资，但会在供应链各环节分摊成本，甚至可能对不同客户收取不同的token价格。

谁受益？

三大存储厂全部跑赢大盘：

三星：目标价从22.5万上调到44万韩元。2027年EPS比市场共识高26%。HBM4技术领先，5月出口数据暗示已开始出货。

SK海力士：目标价从115万上调到330万韩元。2027年EPS比共识高32%。ROE将达到108%。

美光：目标价从510美元上调到1300美元。2027年EPS比共识高38%。ROE达85%。

联发科也值得关注。如果云厂商为了规避英伟达加价而要求直接采购HBM，ASIC服务商的商业模式正好承接这个需求。联发科在TPU项目执行稳健，过去两个月已涨约130%，伯恩斯坦仍维持跑赢评级。

唯一的输家：铠侠（KIOXIA）。只有NAND没有HBM，无法受益于这轮涨价。维持跑输大盘，目标价4万日元不变。

为什么传统P/B估值失效了？

这轮周期存储厂的ROE到了前所未有的水平，三星55%，海力士108%，美光85%。2027年现金可占账面价值的70%-80%。

伯恩斯坦改用1年远期P/E估值，倍数设在历史周期低谷：三星和海力士6.2倍，美光7.7倍。

即便用了最保守的倍数，三家目标价仍有15%到26%的上行空间。

最后一个重要判断：2028年随着更多洁净室投产，存储价格会走软，营收同比下滑。但即使在下行周期中，DRAM行业毛利率仍有70%，高于除2018年之外所有历史上行周期的峰值。

这轮周期，和以往任何一轮都不一样。

⚠️ 以上为伯恩斯坦公开研报内容整理，所有评级和目标价均为机构观点，不构成投资建议。

[Yiannis Zourmpanos @yianisz]

This chart is exactly why I've been saying the market is underestimating semis.

Look at 2026. Hyperscaler earnings are expected to grow roughly 8-22%, while semis are growing 80-107%. The AI spending is showing up first in the supply chain, not in the companies writing the checks.

Amazon, Microsoft, Meta and Google are spending hundreds of billions on compute. Meanwhile, the suppliers are capturing the profit pool today.

this is the biggest misconception in the market right now. People keep asking whether hyperscaler capex is sustainable. The better question is who gets paid first.

Right now it's $NVDA, $AVGO, $MRVL, $AMD, $MU, $SNDK, $TSM, $ASML and the rest of the AI supply chain.

The hyperscalers are funding the boom.

The semiconductor industry is monetizing it.

[investing @DollarCostAvg]

🚨 BULLISH ON $MU 🚨

$MU is the only U.S.-based memory chip manufacturer with leading-edge DRAM and NAND production at scale.

$SPCX Elon Musk highlighted the strategic importance of domestic semiconductor manufacturing as AI demand continues to explode. 👇. He mentioned and confirmed $MU is only Memory maker in US till 2029-2030, no other company exist till next few years 🔥🤯