🚨Leopolds 13F is out now:

It’s obvious that data centers, AI and energy is still the theme here.

With AI Datacenters as the core theme:

He added to $IREN $KEEL $CLSK $APLD $CRWV

Stay patience. Know what you hold.

-BP

Not financial advice.Wall St Engine: Leopold’s Situational Awareness 13F is out.

New positions include $TE, $HIVE, $SHAZ, $INTC, $AMD, $SMH, $TSM, $MU, $GLW, $ASML and $NVDA

Added to $KEEL, $CLSK, $RIOT, $IREN, $APLD, $BTDR, $CRWV and $SNDK.

Trimmed $SEI, $CORZ and $BE.

Exited $COHR, $LBRT, $TSEM, $HUT, $LITE,

Let’s be real here…

Every AI data center Nvidia ever builds needs optical transceivers to move data between GPUs.

Not because of GPU performance.
Not because of CUDA software.
Not because of rack density.

Because of physics. Electrical signals cannot move data at the bandwidth AI clusters require over distances greater than a few meters. Light can.

$INTC own data center documentation confirms fiber optic interconnects at 800G+ are required for every GPU cluster at scale.

Demand for 800G transceivers is STRONG with volume ramps underway in 2026.

$AAOI is still ONLY approximately $15B compared to peers as $LITE $COHR $CIEN at +50$B.

The GPU market is $500B+. The company supplying the optical layer those GPUs depend on for data movement is at $15B.

Rosenblatt raised their price target to $220 in May 2026, up from $140 in March.

Current price implies limited upside on Rosenblatt’s numbers relative to earlier valuations.

Now execution matters.Either the AI GPU capex cycle delivers the optical component volume the data suggests, or it doesn’t.

At current market cap vs. the $500B+ infrastructure build, valuations reflect much of the outcome.

But when execution follows $AAOI will easily run past $280, $300, $320, $350, $400.

I’m long $AAOI.

-BP

Save this.

Not financial advice. DYOR.

RT Investing visuals
A breakdown of the photonics value chain and best positioned business:

Layer 1: Materials & wafers
• $GLW
• $AXTI
• $IQE
• $AIXA
• $AMS

Layer 2: Core photonic devices
• $IPGP
• $COHR
• $LITE
• $LASR
• $SIVE

• Layer 3: Components & modules
• $AAOI
• $MTSI
• $FN
• $VIAV
• $LPTH

Layer 4: Systems & equipment
• $ASML
• $BESI
• $ASM
• $LPKF
• $MKS

Layer 5: Test, metrology & yield
• $CAMT
• $FORM
• $AEHR
• $ONTO
• $VIAV

Investing visuals: Photonics stocks are going parabolic:

• $LITE +1621%
• $AAOI +1434%
• $COHR +529%
• $CRDO +377%
• $VIAV +353%
• $FN +281%
• $MTSI +189%
• $MRVL +186%
• $IPGP +136%
• $POET +127%

A TON OF THINGS HAPPENED IN THE STOCK MARKET TODAY.

Here’s a full recap:

1. The $SPX S&P 500 and $QQQ Nasdaq closed at fresh all-time highs today, even with oil spiking and U.S.-Iran peace talks stalling. The market is basically saying AI earnings momentum is more important than geopolitical risk right now. The PHLX Semiconductor Index rose 2.6%, with semis now comprising 17% of the S&P 500.

2. The optical networking trade continues to gain momentum. $LITE Lumentum surged after being added to the Nasdaq 100, while peers like $COHR Coherent and $GLW Corning also benefited from the AI data center connectivity theme. This is becoming one of the clearest “picks and shovels” trades in AI infrastructure. $LITE Lumentum’s move matters because the market is realizing AI is not just about GPUs. It is also about optical transceivers, lasers, fiber, switches, power, cooling, copper, glass, and every bottleneck inside the data center stack. The AI trade is expanding from chips into the entire physical infrastructure layer.

3. $TSLA Tesla was in focus after the White House invited Elon Musk, alongside more than a dozen top U.S. executives, to join President Trump’s trip to China this week. The delegation includes Apple’s Tim Cook, and BlackRock’s Larry Fink with other executives from Goldman Sachs, Cisco, Mastercard, Citi, Meta, Micron, and more. The trip is expected to focus on U.S.-China trade and investment, potential Boeing aircraft purchases, agriculture, energy, and possibly extending the rare earths truce, making Musk’s inclusion important given Tesla’s exposure to China manufacturing, demand, and supply chains. Jensen Huang $NVDA and Lisa Su $AMD were not invited as per Reuters.

4. Oil jumped again as the U.S.-Iran situation remained unresolved, with Brent crude moving above $104 and the Strait of Hormuz risk still hanging over the market. This is the main bear case investors are watching: if oil keeps rising, inflation expectations can come back and pressure the Fed. CPI will be reported tomorrow with expectations of a 3.7% print, the highest in 2.5 years.

5. Per Jeffries: "47% of semis/hardware stocks are screening overbought on 14-day RSI (high was 70% in Dec '23, 66% on April 24) as of Friday's close. The SOX is now 60% above its 200-day moving average, a level not seen since March 2000 and July 1995."

6. Earnings season continues to come in strong. Reuters reported that 440 S&P 500 companies $SPX have reported, with 83% beating estimates and Q1 earnings growth now projected at 28.6% year-over-year. That is why the market keeps absorbing bad macro headlines.

7. $CBRS Cerebras increased its planned IPO price range as investor demand for AI chip exposure remains extremely strong. The company is now looking to sell 30 million shares at $150 to $160 each, potentially raising up to $4.8 billion. The AI IPO window is officially open again.

8. $CRCL Circle reported a 20% increase in quarterly revenue and reserve income to $694 million, helped by rising demand for USDC during a volatile period. USDC circulation grew 28% year-over-year to $77 billion, showing that stablecoins are becoming a bigger part of the financial infrastructure story.

9. The most traded options today were $TSLA with 5.5M contracts, $NVDA with 4.8M, $MU with 1.1M, $INTC with 1.0M, $AAPL with 1.0M, $NOK with 886K, $META with 637K, $AMZN with 621K, $MSFT with 591K, and $AMD with 514K.

10. $HIMS Hims & Hers reported Q1 revenue of $608M, up 4% YoY, with subscribers growing 9% YoY to nearly 2.6M. The company raised full-year 2026 revenue guidance to $2.8B-$3.0B, but profitability moved lower as gross margin fell to 65% from 73%, net loss was $92M, and adjusted EBITDA declined to $44M from $91M last year. Management said 2026 is a “defining year” as the company expands branded GLP-1 offerings, international markets, diagnostics, and technology infrastructure.

11. $ASTS AST SpaceMobile reported Q1 revenue of $14.7M and said it remains on track for full-year 2026 revenue guidance of $150M-$200M. The bigger story is deployment: BlueBird 8, 9, and 10 are expected to launch in mid-June, BlueBird 11 through 33 are already in advanced production, and the company is targeting roughly 45 satellites in orbit during 2026. ASTS also received FCC authorization for commercial SpaceMobile service in the U.S., hit 98.9 Mbps peak data speeds from an in-orbit satellite to an unmodified smartphone, and ended the quarter with about $3.5B in cash

12. China is also in focus ahead of a Trump-Xi summit, with Chinese stocks $BABA $KWEB $JD hitting an 11-year high and the yuan reaching a three-year peak. Investors are watching whether the U.S. and China could coordinate around trade, Iran, and global supply chains.

I used to do these recaps a year ago...have gotten a bit busy but looking to bring them back...would people like them at the end of the day? I use AI to help summarize the events but the real time (around 30-40 min daily) comes in curating the best headlines and including the relevant details. Happy to start it again if people want it back!

WALL STREET IS THE GREATEST SHOW ON EARTH.

最近光模块板块有点回调，不少小伙伴又开始担心了，纷纷建议我们来讲一讲。正好最近 ChatGPT Images 2出来，测试了一下用它做行研分析图，如👇图1，就是用Images 2出的光模块拆解，效果真的不错。

言归正传，我们依旧长期看好光模块领域的核心企业，AI 数据中心的算力战争，早已经从 GPU 打到光子芯片了。🧐

很多人还在死盯着英伟达，殊不知真正的卡脖子环节，早就转移到光互联和光子学这条产业链上了。

我们看看数据就知道了：AI 服务器的功耗从传统的 5-10kW 直接飙升到 100kW+，这些算力怪兽之间的数据传输，靠传统铜线？根本扛不住！光子芯片，才是 #AI 基础设施的下一个必争之地。

我们最近花了不少时间，把整条光子学产业链从头到尾梳理了一遍。从最上游的材料、设备，到中游的激光器、代工厂，再到下游的光模块、网络连接，每个环节的核心标的都扒拉了个遍。

这不是什么概念炒作，是实打实的硬科技赛道，每一层都有明确的投资机会。咱们掰开揉碎了讲，看看这条产业链到底藏着哪些机会：

🔹 第一层：材料与晶圆 - 半导体之基

这是产业链最顶端，主要提供磷化铟、砷化镓这些化合物半导体材料。说白了，没有这些材料，后面啥都别想搞。

• $AXTI (AXT Inc)
垂直整合的衬底供应商，在 InP（光通信激光器专用）市场占据重要地位。这哥们是从源头卡位的玩家。

• $WOLF (Wolfspeed)
虽然主攻碳化硅功率器件，但在氮化镓射频和光电衬底方面也有深厚积淀。多条腿走路。

• $COHR (Coherent)
全球领先的工程材料和光子器件厂商。不仅卖材料，还向下游垂直整合到激光器和模块，这是真正的巨头玩家。

🔹 第二层：制造设备 - 卡脖子环节

光子芯片的制造需要极高精度的光刻和外延生长设备，这可是真正的卡脖子环节。科林研发(LRCX)与应用材料(AMAT）、阿斯麦(ASML）并列，是全球前三大半导体设备商。

• $ASML(阿斯麦)
绝对的霸主，没啥好说的。随着硅光技术向更小纳米节点迈进，ASML 的光刻机是不可或缺的。垄断级别的存在。

• $AMAT (Applied Materials) & $LRCX (Lam Research)
在薄膜沉积、刻蚀等关键工艺上提供设备。2026 年的逻辑很清楚，随着 AI 驱动的晶圆厂扩产，这些设备厂的订单可见度极高。稳稳的。

🔹 第三层：激光器 - 光源引擎

激光器是把电信号转化为光信号的核心心脏，不可或缺的关键要素，妥妥的卖铲子王者。

• $LITE (Lumentum)
提供用于800G到1.6T高速数据传输的磷化铟激光器、EML激光器及光收发器，是NVIDIA和各大超大规模数据中心的核心供应商，市占率高达80%。在云数据中心和电信网络的高性能激光器领域极具竞争力。大厂背书，靠谱。

• $SMTC (Semtech)
专精于高性能模拟、混合信号集成电路、物联网系统及云端连接服务。尤其在激光驱动芯片和 CDR领域优势明显，是信号调理环节的关键。虽然不是最性感的环节，但不可或缺。

🔹 第四层：代工厂 - 制造产能

有了设计和材料，谁来生产？代工厂就是把图纸变成芯片的关键环节。

• $TSM (台积电)
正在积极布局 CPO（共封装光学）技术。台积电的 COWAS 封装技术与硅光技术的结合，是未来 AI 算力卡的标配。龙头就是龙头。

• $TSEM (Tower Semiconductor)
总部位于以色列，专门从事模拟和混合信号制造，在硅光代工领域有先发优势，是很多初创光子设计公司的首选。小而美。

🔹 第五层：测试、检测与封装 - 质量守门员

光子芯片出厂前，必须经过严格的测试和检测，这一环节直接决定良品率。

• $AEHR (Aehr Test Systems)
目前的明星股！核心逻辑是光子芯片，尤其是激光器在出厂前需要长时间的"老化测试"。随着硅光芯片上量，AEHR 的晶圆级测试设备需求呈指数级增长。这是今年的大黑马。

• $ONTO (Onto Innovation)
专注于微电子器件、逻辑芯片、DRAM 内存以及先进封装领域提供高性能的过程控制量测、缺陷检测、光刻技术和数据分析系统，提供自动光学检测，在异构集成封装检测中不可或缺。稳健型选手。

🔹 第六层：光学模块 - 系统的物理表现

这一层是把光芯片、电芯片、光纤连接器整合在一起的"黑盒子"，也是最接近终端应用的环节。

• $AAOI (Applied Optoelectronics)
随着数据中心向 800G/1.6T 演进，AAOI 作为具有自主激光器产能的模块厂商，成本优势和产能弹性都很明显。高弹性标的。

• $GLW (Corning)
康宁是光纤之王。无论芯片怎么变，最终光信号还是要通过康宁的光纤进行传输。躺赢的存在。

🔹 第七层：网络连接 - 价值兑现终点

这是整条产业链的最后一环，也是价值兑现的终点。所有的技术最终都要在网络设备上体现出来。

• $AVGO (Broadcom)
光子学领域的终极 Boss！博通不仅拥有全球最强的交换芯片，还通过集成硅光技术CPO直接将光接口做进芯片里。同时也是ASIC定制芯片的龙头企业，是谷歌TPU的核心合作伙伴。属于真正的王者。

• $ANET (Arista Networks)
AI 数据中心网络架构的领导者，是光模块的最大采购方和应用场景定义者。需求端的话事人。

• $MRVL (Marvell)
在光互联 DSP 领域与博通双头垄断，其 PAM4 DSP 芯片是 800G 光模块的标配。技术壁垒很高。

总结来说，光子学这条产业链，不是什么遥远的未来技术，而是正在发生的现在进行时。AI 算力的爆发，已经把光互联从"可选项"变成了"刚需"。

从材料到设备，从激光器到封装，从光模块到网络连接，每一层都有明确的赢家和清晰的投资逻辑。我个人策略很简单，一旦光模块板块出现回调，逢低买入是一个好机会！

这份清单里面：
• 有稳健的行业巨头（ASML、台积电、博通）
• 有高弹性的成长标的（AEHR、TSEM、AAOI）

我们不需要全买，但至少要知道这条产业链的关键环节在哪里，钱流向了谁。

AI 的下半场，不只是 GPU 的游戏，光子学才是真正的物理层卡位战。布局正当时，等风来。DYOR🙏

目前上述提及的公司在 #MSX 上面基本都有，炒美股，我选择用 #RWA 美股代币化平台 #MSX，一同投资参与美股市场：http://msx.com/?code=Vu2v44

早期美股投资粉丝和伙伴，可以私信我，填写表单后，可免费进入美股交流和探讨社群(目前每周仅限定10人，助理审核，可能需要一点时间，感恩🙏)！

Rocky: 很多人炒 AI，眼里只有英伟达！

涨了追，跌了慌，最后发现自己就是在坐过山车。其实真正聪明的钱，早就开始布局 AI 的物理层了。

什么叫物理层？就是那些给 #AI 芯片提供电、散热、连接、封装的公司。英伟达再牛，也得靠这些'卖铲子'的公司来支撑整个生态。

我们看看数据就知道了：
• AI

For six months the market has asked the wrong question about $CRDO.

When does CPO kill copper. How fast does the AEC market shrink.

Wrong frame.

I caught Gilad Shainer at GTC in front of NVIDIA's Spectrum-X CPO demo he runs networking at NVIDIA, decides what ships inside Vera Rubin and Kyber Ultra. Held the on-camera footage for weeks. Released it Sunday inside the paid post. Same day Jefferies initiated $CRDO BUY at $175.

Unlocking the framework today.

Shainer to BEP, on camera:

"Everywhere you can use copper, you use copper. Copper consumes zero power. If you can use copper, we use. Period."

On the timeline that breaks the bear case:

"More of a 2028 story." Tied specifically to Feynman NVL1152, when scale-up domains hit 1,152 GPUs and distances physically break copper.

Not 2026. Not 2027. 2028.

The bear case assumed CPO was imminent. The NVIDIA roadmap says otherwise — Vera Rubin NVL72, Kyber Ultra NVL144, even NVL576 all stay copper inside the rack. Feynman in late 2028 is the first platform where copper-to-switch is even in question, and SemiAnalysis base case still has copper as plan of record.

The market was modeling the wrong variable.

Full piece + Shainer video now public:
https://bepresearch.substack.com/p/credo-was-the-wrong-question

$CRDO $NVDA $LITE

Memory and optical are the two main sectors driving performance this year. The reason behind this is the AI bottleneck.

So what’s the next bottleneck?

Shay Boloor: HERE ARE THE TOP 15 PERFORMERS OF 2026

1. $AAOI +357%
2. $AEHR +315%
3. $SNDK +288%
4. $LITE +143%
5. $BE +139%
6. $WDC +116%
7. $PL +95%
8. $TWST +89%
9. $NBIS +88%
10. $COHR +87%
11. $INTC +86%
12. $CRML +81%
13. $DOCN +78%
14. $NVTS +73%
15. $AMKR +71%

RT Mauro
A lot of people seem to think photonics has already had its big run.
They’re missing the next leg.

Stocks like $AAOI and $LITE exploded on the current 800G pluggable wave — a lot of that growth is now priced in.

The real next chapter is CPO - (Co-Packaged Optics).

CPO solves the massive power and heat problems in AI data centers way better than today’s pluggables. Hyperscalers are already designing their future clusters around it. The big ramp will be 2027–2028.

The companies that actually enable CPO — $SIVE (critical lasers), $SOI (substrates), $TSEM (foundry), $AEHR (testing) — are still trading at valuations that don’t reflect how big this shift can become.

CPO is the layer where the next major move in photonics will come from IMO.

The space economy is getting its own Wall Street sector map now.

Morgan Stanley just published a 60-stock list organized around the space economy; propulsion, alloys, communications, sensors, launch.

This is how institutional money flows.

First the research. Then the allocation.
Then the retail crowd asks why these stocks moved.

The list includes the obvious names; $MP $RKLB $LITE $COHR $FLY $ASTS $PL and the less obvious ones; $RDM $TSAR $AME $APH.

The supply chain that makes both of them work; launch and satellite, are where the next upside tends to live.

-BP

This is not financial advice. Please do your own research.

A material most investors have never heard of is quietly becoming the single most IMPORTANT chokepoint in the entire AI infrastructure stack.

It’s called indium phosphide. And right now, demand for it is outpacing supply by 70%.

Here’s why it matters:

When hyperscalers build massive GPU clusters; thousands of chips spread across racks, rows, and entire buildings, those chips need to communicate with each other at extraordinary speeds.

Continuously. At near-zero latency.

Copper wire cannot do this at 800G and 1.6T speeds without generating so much heat the signal degrades before it arrives.

The ONLY physical solution is light. And to generate that light, you need indium phosphide.

InP is the material inside the laser components of every serious optical transceiver in production today.

Silicon is brilliant at computation. It is terrible at producing laser light.

Indium phosphide is the opposite, it converts electricity directly into coherent infrared light at exactly the wavelengths used in fiber-optic data centers. Remove it from the stack and the entire system stops.

This is not a niche component. This is the nerve ending that makes the AI nervous system function.

And the supply chain was never designed for what AI is demanding from it.

A handful of companies control global InP substrate production. A meaningful share of that production touches China directly or indirectly, with advanced epitaxial processing concentrated in Taiwan.

In February 2025, China introduced export licensing for InP-related materials. Epitaxial fabs hold weeks of inventory, not months.

A few months of licensing friction halts production. Once lasers stop shipping, everything downstream stalls: transceivers, switches, AI clusters.

Lumentum’s CEO said in early 2026 that despite a 12x increase in manufacturing capacity over 24 months, the company is “falling further and further behind demand.”

Their InP laser production is now fully allocated for the next 32 months. Lead times that were measured in weeks two years ago are now measured in years.

$AXTI, which controls 60–70% of global InP substrate production through its Beijing facility, reported an InP order backlog of over $60 million, more than double what it was six months earlier.

They are tracking to double production capacity by end of 2026. It still won’t be enough.

The SUPPLY-DEMAND gap on InP devices is projected at 70% for 2026.

Here is the thing most investors miss:

Silicon photonics does not replace indium phosphide. It depends on it. Every modern silicon photonics system bonds InP laser dies onto silicon chips using advanced packaging.

As silicon photonics adoption grows and systems get denser and faster, total InP demand increases. The more efficient the architecture becomes, the more InP it needs.

The companies MOST exposed to this chokepoint, as direct beneficiaries of scarcity pricing and multi-year backlog visibility, are $LITE, $COHR, $AAOI, and $AXTI. Each sits at a different layer of the same bottleneck.

The market has partially re-rated the transceiver layer. It has not re-rated the substrate suppliers. That gap is where the asymmetry still exists.

AI progress is no longer limited by algorithms or model architecture. It is limited by atoms.

The investors who understand which atoms are the chokepoints, and who controls them, will find the real alpha in this cycle.

The constraint is ALREADY downstream.

-BP

Not financial advice. Do your own due diligence

Black Panther Capital: $AAOI, $COHR, $LITE have already run 150-250% and trade at extreme multiples.

The UPSTREAM play is where the undiscovered leverage sits.

Here is what most people are missing in the photonics chain:

Transceiver makers:
$AAOI, $COHR, $LITE, visible, crowded, already priced for

The Credo bear case is built on the wrong question

everyone arguing "when does cpo kill copper"

meanwhile Credo shipped zeroflap optics at 800g three weeks ago and quietly walked into the transceiver tam that $cohr and $lite have owned for years

https://open.substack.com/pub/bepresearch/p/credo-was-the-wrong-question

$CRDO