🐦 推特SerenityV2 · 实盘不做把握 5/10
北京时间·🕒 发布 11:11:40时区推断·⟳ 抓取 11:14:52·⚡ 决策 06-25 15:04:20 +3h49m·124.2 天前
KOL(Serenity)光电/存储产业链择时随笔——点名LITE/COHR/AAOI/MU/POET谈'8个月提前定价'框架,非硬催化
这是一条KOL个人择时随笔(原文'Just from my personal experiences''I'm learning along the way'),没有订单/财报/政策/监管等任何硬催化——帖子本身不会驱动这些大盘股价格,本质属于'事件太小不足以move股价'+'无信息edge'。
📄 原文Serenity↗ 原文
Just from my personal experiences: My new heuristic is markets tend to price things in ~8 months in advance unless your…
不做 LITE.US
KOL择时随笔非催化;1.6T可插拔已充分price-in(一年~12x,PE148,距高-24.5%),'right around now'窗口已走完,无新增edge
不做 COHR.US
同LITE:已~5x、PE187、距高-15%,1.6T主题已兑现;无新信息edge
不做 MU.US
基本面最强但$1.18T巨头、近20日+17%逼近高点;一条推文无信息edge且非事件可归因,真催化是其财报而非本帖
不做 AAOI.US
已因CPO延迟报告深跌(距高-37%,20d-17%),本帖无新催化;KOL框架也暗示该主题需等待,当下无可交易触发
不做 POET.US
KOL自评CPO Scale Out要late H2 2026'太早';距高-49%、20d-20%,无时间敏感催化,纯beaten-down早期票
展开详情(成交明细 / 研究档案 / 原文全文 / 决策记录链)
研究档案
关键理由
- 这是一条KOL个人择时随笔(原文'Just from my personal experiences''I'm learning along the way'),没有订单/财报/政策/监管等任何硬催化——帖子本身不会驱动这些大盘股价格,本质属于'事件太小不足以move股价'+'无信息edge'。
- 1.6T可插拔标的已充分price-in:LITE一年$88→$1085(~12x)、COHR$80→$440(~5x),现PE分别148/187,距高点-24.5%/-15.3%、近20日-10%/-2%——KOL说的'right around now'定价窗口实际已走完,无新增edge。
- KOL真正看多的前瞻主题要么太远(CPO Scale Up H2 2027、玻璃基板2027-2028,他自己都说'不该12M+重仓'),要么是已深跌的早期票(POET距高-49%、AAOI距高-37%,CPO Scale Out要等late H2 2026),当下均无时间敏感的可交易触发点。
- 唯一基本面最扎实的是MU(PE仅23.7、memory bottleneck仍在re-rating),但它是$1.18T巨头、近20日已+17%逼近高点,一条推文对它毫无信息edge,且非事件驱动可归因的成交。
- 新鲜度核查:帖子6-24首发、扩散low、底层CPO时间轴机构已广覆——既无独家新信息,也无'刚出未扩散'的抢跑窗口,medium price-in。
主要风险
- 若memory/photonics板块继续动量上行(MU临近财报、HBM4量产兑现),纯动量交易者本可吃到一段beta,AVOID会错过——但这是beta不是本帖的alpha。
- POET/AAOI已深度回调,若CPO延迟担忧反转或出现新订单,反弹弹性大(历史AAOI单日+24%/+57%先例在),但需真催化,本帖不提供。
- KOL帖若被大账号转载放大,短线可能有情绪性脉冲;但当前扩散low,概率不高。
- 现为美股盘前不可成交,即便想博也只能挂次日开盘单,跨夜gap风险叠加无催化,性价比差。
原文全文
展开原文 · 2,000 字
Just from my personal experiences:
My new heuristic is markets tend to price things in ~8 months in advance unless your theme is Space, Robotics/Quantum.
For themes I'm tracking:
For photonics:
- Next-gen pluggables (1.6T) are right around now (ex. $LITE, $COHR, Innolight, etc. then $AAOI, $JBL 1.6T LRO and other later bloomers H1 2027).
- CPO Scale Out related stuff is late H2 2026 (ex. $POET, Shunsin, etc)
- NPO stuff H2 2026/2027 prob as bridge architectures
- CPO Scale Up H2 2027 (ex. Ayar, Celestial, etc).
If I use the 8 month rule like 4 months too early for CPO Scale Up Names.
But in hindsight, I wanted to be early this time around given I felt this was the most compelling architectural shift given GS tam projections + markets seem to disagree, which is why there's a selloff with Foci/MSScorp and others.
Maybe it should start picking up again and start to be priced in December 2026 IG? idk we'll see.
Stuff like lasers, $AXTI (InP substrates) are used across all those themes. FAU/silicon photonics/PCB midplanes, etc. are more concentrated around different architectures.
Memory theme is quite a massive bottleneck as you've been seeing:
- HBM3e around now as you've seen with Micron, Samsung, Sk Hynix etc. and NAND with Sandisk.
- HBM4 mass production right around now too with SK Hynix, Micron/Samsung.
- HBF 2027?
But again these are the same Kioxia, Samsung, Sk Hynix, $MU, Kioxia players across different architectures, and we're seeing re-rating cause we're mid-bottleneck.
So these are likely to keep compounding.
Glass Substrates:
H2 2026 start with SKC Absolics (probably just early stage ramp).
H2 2027 - 2028 is where the fun begins, with 2028 TSM CoPoS, others like Samsung etc. 2027. and Intel later on.
Humanoid Suppliers:
Very early, but as I said with Space markets probably won't wait to price this in with Elon giving bullish $TSLA optimus projections and Chinese humanoid companies like Unitree/AGI Bot starting to ramp up.
Neoclouds:
I said t
My new heuristic is markets tend to price things in ~8 months in advance unless your theme is Space, Robotics/Quantum.
For themes I'm tracking:
For photonics:
- Next-gen pluggables (1.6T) are right around now (ex. $LITE, $COHR, Innolight, etc. then $AAOI, $JBL 1.6T LRO and other later bloomers H1 2027).
- CPO Scale Out related stuff is late H2 2026 (ex. $POET, Shunsin, etc)
- NPO stuff H2 2026/2027 prob as bridge architectures
- CPO Scale Up H2 2027 (ex. Ayar, Celestial, etc).
If I use the 8 month rule like 4 months too early for CPO Scale Up Names.
But in hindsight, I wanted to be early this time around given I felt this was the most compelling architectural shift given GS tam projections + markets seem to disagree, which is why there's a selloff with Foci/MSScorp and others.
Maybe it should start picking up again and start to be priced in December 2026 IG? idk we'll see.
Stuff like lasers, $AXTI (InP substrates) are used across all those themes. FAU/silicon photonics/PCB midplanes, etc. are more concentrated around different architectures.
Memory theme is quite a massive bottleneck as you've been seeing:
- HBM3e around now as you've seen with Micron, Samsung, Sk Hynix etc. and NAND with Sandisk.
- HBM4 mass production right around now too with SK Hynix, Micron/Samsung.
- HBF 2027?
But again these are the same Kioxia, Samsung, Sk Hynix, $MU, Kioxia players across different architectures, and we're seeing re-rating cause we're mid-bottleneck.
So these are likely to keep compounding.
Glass Substrates:
H2 2026 start with SKC Absolics (probably just early stage ramp).
H2 2027 - 2028 is where the fun begins, with 2028 TSM CoPoS, others like Samsung etc. 2027. and Intel later on.
Humanoid Suppliers:
Very early, but as I said with Space markets probably won't wait to price this in with Elon giving bullish $TSLA optimus projections and Chinese humanoid companies like Unitree/AGI Bot starting to ramp up.
Neoclouds:
I said t
决策记录链 · 只追加
每次领取产生一条记录,重新决策追加一条并以 supersedes 指向被替代的记录(ADR 0014)。
这条决策早于决策账本(2026-10-08 起逐条追加),没有追加记录。