totallyMAYHEM

[HenryMorgan @HenryMorgan_X]

这段关系看起来像"双性恋纠葛"，本质上更像权力与孤独制造出来的"错位共生"——大家都在用别人填补自己的洞，却不肯直视自己到底想要谁、想要什么。

李宪宏：对沈晓雯是长期婚外伴侣式的利益＋习惯依附；对林莉莲是中年危机式的肉体刺激与浪漫幻象。他不处理真实欲望，只把不同人当"功能模块"轮流用。

陈汉荣：十年单恋老板，爱意被压进"完美助理"的身份里；转去和老板挚友 Patrick 交往，更像把对李宪宏求不到的爱投射/替代到Patrick身上——身体在同性关系里，心还在别处。

于是畸形感就在这里：没有一条线是干净利落的"我爱你、你也爱我、我们面对现实"，全是绕路——绕开阶级权力、绕开社会眼光、绕开自己的怯懦，用三角/多角把真相一直拖下去。

[Art of Speculation @ArtofSpecuycky]

英特尔CEO陈立武最新访谈，提到的几个被市场低估的方向
英特尔CEO陈立武最近的一场访谈，提到了几个值得关注的新材料和资源方向，整理一下，结合我自己的判断说说怎么看。

陈立武这次访谈让我印象最深的是他反复提到的几个底层方向：SiC、GaN、Glass Core，以及很多人忽略的人造钻石。这几个方向看起来都不性感，却指向同一个问题：AI正在从拼算力，转向拼连接、拼散热、拼材料、拼电力。未来市场会越来越关注谁控制瓶颈，谁掌握关键材料。

1. InP（磷化铟）
作用：光模块，光通信，800G/1.6T，AI数据中心互连

投资逻辑: AI集群越大，光互连越重要。高速激光器和下一代光引擎，越来越依赖InP材料。

首选股票: 纯InP（AXTI），InP + 光器件龙头（COHR、LITE）

2. SiC（碳化硅）
作用：电动车，电源管理，AI数据中心电源

投资逻辑: AI最大的瓶颈之一是电力。SiC提升电源转换效率，是下一代高功率系统的重要材料。

首选股票：ON、STM
高风险高弹性：WOLF

3. Glass Core（玻璃基板）

作用：下一代封装基板，超大尺寸GPU封装，HBM高层堆叠，高频高速信号传输

投资逻辑：随着GPU越来越大、HBM堆叠越来越高，传统ABF有机基板正在接近物理极限。Glass Core拥有更好的平整度、更低的热膨胀系数以及更优异的高频信号性能，被认为是下一代先进封装的重要基础材料。

首选股票：GLW（康宁）

4. Gallium Nitride（GaN，氮化镓）

作用：AI数据中心电源，800V DC，电源管理，射频通信

投资逻辑：
GaN属于第三代半导体材料，拥有更宽的禁带宽度、更高的击穿电场以及更高的开关频率。相比传统硅器件，GaN能够实现更高的电源转换效率、更小的体积和更低的能耗，因此正在逐渐进入AI服务器、电源模块和下一代数据中心供电架构。

首选股票：NVTS、POWI、IFNNY（ADR）

5. 人造钻石
这个是很多人忽略的。

作用：GPU散热，ASIC封装，高频光器件

投资逻辑：AI芯片越来越热，散热正在成为新的瓶颈。
人造钻石拥有远超铜的导热性能，未来有望应用于GPU、ASIC等高功率芯片散热，甚至部分高端光器件热管理。但目前成本和工艺仍是限制，更像是2030年以后值得重点关注的新材料方向。

首选股票：大部分都是A股公司，严格说没有特别纯的美股标的。

如果按我自己的投资优先级排序：
Glass Core（玻璃基板）≈ InP（磷化铟） > SiC（电力） > GaN（氮化镓） > 人造钻石

[Lin @Speculator_io]

The Five Layers of Memory

1. Near Memory:
Sk hynix, Samsung $MU

2. Main Memory:
Sk hynix, Samsung, CMXT $MU

3. Expansion Memory:
SK hynix, Samsung $MU $ALAB $MRVL $MCHP $RMBS

4. Contexted Memory:
SK hynix, Samsung, Kioxia $MU $WDC $SNDK $SIMO

5. Data Lakes:
$STX $WDC $DELL $NTAP $P $HPE $IBM

[Shay Boloor @StockSavvyShay]

2026 WILL BE THE YEAR OF AI MEMORY

• $MU supplies the DRAM & HBM next to GPUs that let AI think longer & hold more context.
• $SNDK, $STX, $WDC store massive AI datasets & model files so training and inference can scale.
• $PSTG, $NTAP turn raw memory & disks into reliable storage systems enterprises actually run AI on.
• $LRCX, $AMAT build the equipment that makes advanced memory like DRAM, HBM & NAND possible.
• $MRVL, $RMBS, $SIMO design the controllers that control how fast data moves between GPUs & memory.

[Michael | Hypermarkets @itsmichaelluu]

Why $RMBS Is Critical to $MU:

Think of it this way:

$MU builds the memory.
$RMBS owns many of the rules, designs, and traffic systems that make memory work at AI speeds.

Why it matters

• Patents & Licensing
Micron licenses Rambus IP to build modern DDR5, DDR6, and future memory products.

• Server Memory Chips
Rambus supplies critical RCD chips used on enterprise server DRAM modules sold into hyperscale AI data centers.

• AI Data Movement
Rambus develops the interconnect technology that helps move data efficiently between CPUs, GPUs, and memory.

• HBM & AI Infrastructure
As Micron scales HBM memory for AI, Rambus technology helps enable signal integrity at extreme speeds.

The AI Supply Chain Connection

NVDA → MU → RMBS

NVIDIA needs massive amounts of HBM memory.
Micron Technology supplies that memory.
Rambus provides critical IP, chips, and interconnect technology helping that memory function at AI scale.

One-Line Investment Thesis

If Micron is selling the AI memory, Rambus is helping build the roads that allow that memory to operate at full speed. Without advanced memory interfaces, controllers, and IP, AI servers become bottlenecked long before they reach their full performance potential.

[tan @tansalpha]

$SIMO is an under-the-radar picks-and-shovels play on AI-driven NAND demand.

Everyone looks at the NAND suppliers first: $SNDK, Samsung, SK hynix, Kioxia, $MU. Makes sense.

But if AI workloads start using flash as a cheaper warm memory tier between HBM/DRAM and cold storage, then the SSD controller becomes a key piece of the stack.

That’s why $SIMO is interesting. They sell the controller brains that make NAND usable: speed, reliability, endurance, power efficiency, QoS, and enterprise-grade performance.

And there are signs they’re moving beyond phones/client SSDs.

Their MonTitan enterprise SSD controller platform is aimed at data-center and AI workloads, and they showed AI storage products at NVIDIA GTC around near-GPU storage, KV-cache extension, warm data storage, and AI server boot drives.

That maps directly to the idea that flash is becoming part of the AI memory hierarchy, not just cell phone/PC storage.

The catch: this is still early. $SIMO is not a pure data-center name yet. eMMC/UFS and client SSDs count for a large chunk of revenue, so the enterprise AI storage piece has to become big enough to move the model.

But if NAND becomes more important to AI infrastructure, $SIMO looks like one of the cleaner picks-and-shovels names hiding in plain sight.

Q1 2026 revenue was $342.1M, up 23% sequentially and 105% year over year, and the stock has clearly re-rated since. Not cheap anymore, but if AI-driven NAND demand keeps growing, this could still have room to run.

[Alphatica @alphaticaio]

MARKET MECHANICS LESSON:

Why MRVL Crashed Into the Close

Yesterday MRVL ran from ~$305 to ~$329 by mid-afternoon. Then it fell $19 in 10 minutes and crossed at $310.58. No news. No earnings miss. No downgrade. Pure mechanics. Here's what happened and why it matters.

THE SETUP: INDEX INCLUSION ARBITRAGE

When a stock gets added to a major index, passive funds are required to buy it at the closing price on inclusion day. They have no choice. They must own it by the close. Everyone knows this in advance.

Arbitrage firms exploit this. In the days leading up to inclusion, they accumulate shares before the passive funds are forced to buy. Then they sell into the guaranteed demand at the close. The profit is the spread between their entry price and the closing cross.

THE MRVL TRADE:

June 15-17: Arb desks accumulated roughly $20-30B of MRVL inventory through dark pools. Blended cost: approximately $290-300.

Yesterday morning: MRVL ran from $305 to $329. Momentum traders and short-covering piled in ahead of the inclusion print. The arbs sat on their inventory and watched the price rise. Every dollar higher was additional profit on their position.

3:50 PM: The MOC (Market on Close) imbalance feed published. This is the official number that tells the market how much net buying the passive funds need at the close. The imbalance indicated $5-8B of net buying needed.

THE PROBLEM: The arbs had $20-30B of inventory to sell. The passive funds only needed $5-8B. There were 3-4x more sellers than buyers.

3:50-4:00 PM: The arbs recognized the mismatch instantly. They started selling aggressively on the lit order book to clear inventory before the close. They couldn't wait for the 4:00 PM cross because there wasn't enough passive demand to absorb their full position. MRVL dropped from $327 to $310 in 10 minutes.

4:00 PM cross: $310.58. The passive funds got their full allocation at this price. The arbs unloaded the bulk of their inventory. Still profitable. They bought at $290-300 and sold at $310. But significantly less profitable than $327.

4:01-4:14 PM: The remaining $10B+ of residual inventory cleaned up through dark pool and swap structures at the official close mark.

THE OPTIONS TELL:

At 3:33 PM, with MRVL trading at $327.62, someone bought $310 puts expiring that same day for $0.24. Those puts were 5.4% out of the money with 27 minutes until expiration. On any normal day, that's a lottery ticket that expires worthless.

Those puts went deep in the money. MRVL crossed at $310.58. The $310 strike was at the money at the close. A $0.24 option became valuable in 27 minutes because the buyer understood the inclusion mechanics that were about to unfold.

THE LESSON:

This wasn't manipulation. It wasn't a crash. It was the predictable resolution of an arbitrage supply-demand imbalance.

The passive funds created guaranteed demand. The arbs front-ran that demand. The arbs accumulated more inventory than the demand could absorb. The MOC imbalance revealed the mismatch at 3:50 PM. The arbs sold into each other trying to exit first. The price fell to the level where supply met demand.

Every index inclusion event has this dynamic. The magnitude depends on the gap between arb inventory and passive fund demand. When the arbs overestimate the demand (or over-accumulate inventory), the closing cross is violent. When the sizes match, the cross is orderly.

The flow told the story before the price did. The dark pool accumulation over three days mapped the arb inventory. The MOC imbalance at 3:50 PM revealed the mismatch. The options market at 3:33 PM showed someone positioning for the exact outcome. The mechanics were visible to anyone watching the right data.

This is why we track flow, not headlines. The mechanic says: index arbs over-accumulated and the MOC imbalance was too small to absorb them. Same result. Different understanding. Different edge next time.

$MRVL $SMH $QQQ

[Tigris 会讲课教授是好老师 @tig88411109]

华尔街和量化看到美联储点阵图就认为年底利率3.8%，肯定加息啊，卖出大跌。这完全错误。

要知道点阵图缺一票是Warsh主席故意没投，一方面是他认为没啥用，这种预测都是扯淡。一方面，这种统计偏差欺骗市场。

Warsh要投肯定是低数字，那么这一票缺了，按照中位数18个人的统计取值肯定偏高。

如何正确解读，全网唯一系统拆解美联储历史性机制的变革指南。

具体投资策略和颗粒度更细的仓位分析见订阅者周报

[Terry @RichTerry123]

台积电亲自下场，玻璃基板这把火算是彻底点着了。

联合Ibiden和群创光电，三方共同验证玻璃基板适配CoWoS先进封装。过去这东西只在实验室里躺着，台积电这次官宣直接把它从PPT里拉进了现实。半导体封装材料的代际切换，发令枪响了。

为什么非换不可？因为现有的硅基板和ABF载板已经被下一代超大尺寸AI芯片逼到了物理极限——翘曲变形、散热差、高频信号损耗大，三个硬伤摆在那里没法绕开。玻璃基板热膨胀系数和硅芯片天然匹配，三个问题一次性消掉。不是想换，是必须换。

节奏上保持清醒：现在只是验证立项，量产还要等2-3年，超薄玻璃加工和超大面板级封装布局两大工程难题还没解决。英特尔、三星早就开始布局了，台积电这次加速本质上是不想把下一代AI封装的定价权拱手相让。

产业链从上到下，谁卡在这条路上：

上游玻璃原片被康宁、AGC、NEG三家锁死，国内彩虹股份、凯盛科技、旗滨集团靠显示玻璃技术在硬挤。

中游TGV精密加工是最硬的壁垒。海外群创、Ibiden、三星电机占着，国内只有沃格光电做到了全流程量产，稀缺到离谱。

ABF载板这边国内更惨，只有兴森科技和深南电路两家实现了高阶量产，其他的还在追。

设备端激光打孔看帝尔激光和大族激光，电镀设备看东威科技。

封测下游长电科技和通富微电接着。

玻璃基板不是新故事，而是一条早就确定的技术路线。台积电这次下场的意义在于：时间表明显提前，全产业链重新定价和国产渗透的窗口正式打开。目前市场对这条链的估值和重视度仍处于早期，真正算账的阶段才刚开始。技术路线确定，执行节奏、良率爬坡和成本控制才是未来胜负手。值得持续跟踪上下游验证进展。

[Terry @RichTerry123]

昨晚美光、英伟达等一众科技普跌，SpaceX逆势收涨，上市三天涨49%，市值2.65万亿超越亚马逊成全球第五。盘中一度超过微软。给大家拆解一下短期内撑住股价的三个机制。

一、流通盘被焊死了

市场上可以自由交易的SpaceX股票只占总股本3-5%。马斯克长期锁仓，机构基石投资者锁定，散户平台还加了史上最严的防倒卖规则——上市后15到30天内卖出，轻则限制未来所有IPO打新资格，重则永久封号，部分平台还收高额违约手续费。

SpaceX史无前例把20%份额分给散户，按理说会有大量首日抛售。但这套惩罚机制直接把散户的筹码焊死在账户里，卖盘几乎消失。

二、被动基金被迫强制买入

各大交易所修改规则，超大体量公司纳入指数的等待期从3个月压缩到15天。预计6月底到7月初，富时罗素、纳斯达克100将相继纳入SpaceX。

被动型指数基金没有选择，必须按当时市价强制买入，哪怕在高位接。为了腾出权重资金，甚至可能要卖出微软、苹果。这部分被动买盘预计规模达300亿美元。

三、期权市场形成正反馈

SpaceX期权周二开始交易，首日超160万张合约易手。机制是这样的：散户买看涨期权→做市商需要买入正股对冲→股价涨→对冲需求增加→做市商继续买→股价继续涨。这个正反馈循环在放大每一次上涨。

客观说一句，这种建立在流通稀缺和机制保护上的高估值是有隐患的。真正的考验在二季度财报后，部分限售解禁条件触发，真实供给开始释放。如果xAI收入掉队、星舰进度不及预期、卫星通信需求见顶，都是风险点。机制保护是有时效的，时效过了才是真正的定价时刻。

[totallyMAYHEM @samcchannn]

@hanking66 on 9

[totallyMAYHEM @samcchannn]

@w6644999 他懂个屁

[True market Leader @TmarketL]

Dr.David Paul trading Psychology

[Market Rebellion @RebellioMarket]

William O’Neil (1933–2023) explains how the general market creates major tops long before most investors realize what’s happening.

In this classic lesson, O’Neil analyzes more than 10 leading stocks simultaneously, showing how market leaders often begin breaking down before the broader market peaks.

Featured names include:

$NASDAQ
$AAPL
$NFLX
$AMZN
Green Mountain Coffee
Cerner
VistaPrint
https://t.co/QTlCXZnv3l
$BIDU
Express Scripts
Synaptics ($SYNA)

One of the greatest investing lessons you’ll ever study:

Don’t just watch the indexes.
Watch the leaders.

When leading stocks start cracking, the market top may already be forming.

A timeless masterclass from the founder of CAN SLIM and one of the greatest growth investors of all time. 👇📈

[Frank Cappelleri @FrankCappelleri]

$SPX Continues to follow the blueprint.

A near-term pause before challenging new highs would actually strengthen this potential bullish formation.

👉Higher lows just under key resistance are powerful.

[PBA @801010athlete]

One of the biggest downsides of chop not discussed enough is the after math opportunity cost missed.

You are sick of taking losses so you hesitate when you shouldn’t and then find yourself locked out wanting to FOMO bc you over traded the chop/basing.

The cycle repeats over and over.

[Art of Speculation @ArtofSpecuycky]

SpaceX上市了，说说我的真实看法和操作计划
先回应一下评论区说的八月解禁潮问题。说得对，早期LP的套现压力确实不小。这也是我做空SPCX逻辑的核心之一。

先说我对SpaceX估值的理解
现在SpaceX的市销率大概在90倍左右，市值2.1万亿。
市场给这个估值，不是把它当波音那样的航天制造业来算，也不是单纯把星链当SaaS来定价。华尔街现在给它的框架是把SpaceX和xAI、Grok合并之后，当成一家集成了轨道AI算力和太空基建的超级公用事业巨头在炒。

这也是为什么SpaceX的财务看起来如此矛盾。200多亿美元营收，Starlink已经贡献了大量现金流，但Starship依然处于高投入阶段，研发、发射和基础设施建设持续消耗大量资本。当前的估值已经远远超出传统航空航天公司的定价框架，投资者愿意支付高溢价，押注的是未来十年Starlink全球网络的扩张、Starship商业化、国防业务增长，以及更长期的太空经济想象空间。换句话说，今天的SpaceX股价里，已经提前包含了相当长时间维度的增长预期。只要未来的发展速度低于市场最乐观的假设，估值就存在重新锚定的压力。

问题在于，SpaceX估值里很大一部分溢价来自太空AI算力这个方向。但这个业务和星舰、星链比起来，落地时间遥远太多，而且极度烧钱，2029年都不一定能开始量产。这个溢价是否支撑得住，这里要打个问号。

历史告诉我们什么
超级IPO上市第一年的中位数最大回撤高达53%。Meta上市后跌31%，Uber和Robinhood跌了80%。故事越好，短期情绪顶往往越极端。
再举个例子：去年Circle上市，赛道完全不同，但上市逻辑类似顶级热度、情绪溢价极高。涨了两周之后，开始漫长的洗盘，一路跌了将近一年。
SpaceX的估值比Circle当时还要夸张得多。
像这类市场极度瞩目的明星股，第一波FOMO情绪退去之后，主力机构的洗盘几乎是必然的。历史上类似的IPO，从顶部跌去50%是常见的，有些甚至跌了80%以上。
如果你现在把大量资金押进去，等它跌了一半，大概率扛不住，最后在最低点割肉。这不是看空SpaceX的长期逻辑，这是对IPO规律最基本的尊重。

我的操作计划。
下周二SpaceX期权上线。
我打算在7月7日纳入纳指之前开始做空，DCA分批建仓，总仓位大约5%，拿到年底。
具体来说，现在160美元对应2.1万亿市值。如果它继续涨到190美元对应2.5万亿，甚至更高到228美元对应3万亿，我都会在那个区域分批买入put，行权价考虑200到240附近。
逻辑很简单，有几个时间节点会带来持续的抛压：
8月Q2财报后，早期投资人和员工开始解禁首批20%的股份。像Founders Fund这样的翻了几十倍机构不可能不卖。
此后每隔一段时间继续释放7%。11月Q3财报后额外释放28%。12月中旬满180天基本全额解禁，马斯克本人锁定一年。
这么大规模的解禁压力，叠加目前的极高估值，是这个空头逻辑的基础。这5%的做空仓位也是我今年整体多头仓位最自然的对冲。

但长期来说，SpaceX必须配置
说清楚一件事，我做空SpaceX不是因为不看好这家公司。
10年维度来看，星链的全球渗透、星舰的轨道经济、以及未来太空AI算力如果真的落地，SpaceX是这个时代最重要的基础设施公司之一，没有之一。
但现在买和一两年后买，是完全不同的两件事。
中短期的布局逻辑是：等股权解禁的压力充分体现，等市场对太空AI算力的过度溢价回归理性，到那个时候，才是真正值得长线建仓的买点。
现在二级市场160美元追进去，你买的是别人的情绪溢价，上方空间非常有限。

总结
SpaceX是这个时代最重要的公司之一，但最重要的公司不代表现在是最好的买入时机。1到2年后再看，到时候洗盘充分，筹码换手，溢价消化，才是真正的价值投资入场点。现在最聪明的做法是拿一点点仓位做对冲，然后等待。

[Art of Speculation @ArtofSpecuycky]

6月份跌完后的看多逻辑，现在又多了一层验证。

AAII最新投资者情绪调查显示，截至6月10日今天刚更新，看空比例已经升至47.7%，接近今年3月18日52%的极端看空水平，远高于历史平均31%。与此同时，看多比例仅30.4%，明显低于历史平均37.5%。

这说明经历6月份这轮调整之后，市场情绪已经从一两周前的乐观重新回到偏恐慌区域。历史上AAII并不是精准择时指标，但当看空情绪接近极端水平时，这就说明大量坏消息已经被市场提前消化，等今天说的流动性不足，美伊局势升级，日元套利交易风险和Kevin Warsh的不确定性这四大不确定性事件利空出尽后，反转就要来了。

当然，情绪悲观本身并不能成为市场上涨的理由，情绪面是我判断的其中标准之一，其他的基本面，技术面，消息面和期权在今日总结里有分析过，真正决定方向的还是盈利和基本面。当AI资本开支仍在增长、企业盈利预期没有明显恶化，而市场情绪却接近年内最悲观水平时，至少说明当前环境是在为未来的反转积累条件，不是处于一轮牛市乐观泡沫的末期。

[XTrader猫姐美股交易 @maojietrading]

检查了一下财报，资产负债表应该是盘后下跌的原因。甲骨文AI云订单虽然暴增，但由于疯狂烧钱建数据中心，导致其2026财年自由现金流（FCF）直接暴跌成-237亿美元的巨额缺口，且长期负债疯狂飙升至1247亿美元。

[Tigris 会讲课教授是好老师 @tig88411109]

这种情况不是考虑加息应该是降息，股市肯定高大幅反弹的。

但为何现在不会这么简单粗暴？

一般来说核心CPI同比2.9%，环比0.2，住房和服务都偏软，通胀风险不大而是集中在少数因AI投资拉动的细项

看看之前的工资收入也是偏软，当然非农就业上周五大超（我觉得主要是世界杯/中期选举拉动暂时性）

看看我周报里面详细解析了这个背后原因，有地缘政治风险，有拥挤交易风险，🈶近期的夏季事件性冲击，有IPO抽水，所以不是单边买入的操作

也就是我说的不能无脑buy the dip，而是要看量价关键点位支撑以及结合我的估值区间。

具体策略 见订阅者专享的6月这两周的周报 以及案例分析。