mcboom

Betting big on uranium & nuclear. Stacking ETH on the side.

[mcboom @mcboom1]

Trump’s recent strategic investment in 🇺🇸 $MP Materials sent the stock up 100% in days.

If you’re able to spot the next company on Trump’s radar, the upside could be massive.
Top sectors? Rare Earths, Nuclear, and Lithium. Here’s the breakdown👇

[まさたけ@AI社会に全振り投資家 @casa1214185131]

■MLCCの次はPCB！？

Morgan Stanley推計で気になったのが、NVIDIA次世代AIラックの部材コスト内訳。
GB300 → VR200で

PCB：+233%
MLCC：+182%

PCBが、MLCC以上に伸びている。

個人的に注目してる企業が「メック」。
PCB向け銅表面処理剤で世界シェア1位。
PCBが高多層・高密度化するほど重要性
が増す“物理ボトルネック”候補。

■そもそもPCBとは？

Printed Circuit Board、つまりプリント基板。
いわゆる「緑の板」。
ただし今回見るべきは、車や家電に使われる普通の基板ではなく、AIサーバー内で使われる高性能PCB。

メインボード、スイッチボード、ネットワークボード、メモリーモジュール基板など、
AIサーバーの中で部品同士をつなぐ“神経網”。

なぜAIサーバーでPCB需要が伸びるのか。
AIサーバーがGPU単体の勝負ではなく、ラック全体の通信勝負になっているから。

高性能PCBに求められるのは、ざっくり言えば3つ。

・高密度に詰め込む
・高速信号を劣化させない
・熱と電力に耐える

AIサーバーの進化で、PCBにも「安い汎用品」ではなく「高付加価値部材」としての性能が求められている。

[dylan ツ @demian_ai]

The best-performing thing in the AI trade this year is not the GPU, the model, or the memory
-> it is a wafer most investors have never heard of, can't pronounce, and would struggle to explain. It ran over 1000% in one of my books. Here is what it actually is.

It is the substrate.
The crystal wafer a laser gets grown on, sitting four layers upstream of the chip everyone watches. And the reason I trust the signal is that it did not turn up in one book. It turned up in two.

I run a US bottleneck book and a separate EU one. No shared holdings, different construction, different markets. This year the strongest layer in both was the same one almost nobody can name. Not the accelerator. Not even the laser. The wafer the laser is grown on.

Look at the leaderboards side by side.

In the US book, InP and Substrates is the second-strongest basket of the year, behind only memory. In the EU book, InP and Substrates is also second, behind only photonics. Same layer, same rank, two markets that do not coordinate.

The names tell it cleaner:

- US -> $AXTI +475% YTD
- EU -> $IQE.L +1018%, $SOI.PA (Soitec) +438%
and $AIXA.DE (AIXTRON) +247% selling the reactors that grow the wafers

The mechanism is the photonics build. Co-packaged optics needs lasers. Lasers are III-V material, indium phosphide and gallium arsenide, not plain silicon. III-V starts as an epitaxial wafer grown on a specialty substrate, and that supply is small, slow, and hard to add. When optical demand turned from slideware into orders, the squeeze hit the substrate first. Both books found the same constrained object without talking to each other.

Worth the honesty: these are thin floats, so the moves are violent and part momentum. Strip the two biggest and the picture cools. But the substrate is the real upstream gate on the optics bottleneck everyone keeps drawing one layer too high.

The feed watches the photonic switch. The squeeze is one floor below it, in the wafer that switch can't exist without.

Dig deeper: https://t.co/93XT041lcS

[Jukan @jukan05]

"We’ve also heard anecdotal checks/chatter of Vera Rubin rental pricing being better than expected which is allowing neoclouds to take more pricing." via @TMTBreakout

Bullish Neocloud? 👀

[Michael Thomas @curious_founder]

OpenAI and its partners are betting $100 billion+ that the fastest way to build a data center is to build a power plant alongside it.

But their Stargate data centers are taking longer and costing more than their competitors'.

OpenAI's build-your-own-power-plant strategy is as radical as it is novel.

For decades data centers had been powered just like every other industrial operation in America. They called up the local utility, submitted a grid connection request, and soon after received electrons from the power grid. But the developers of Stargate believed this path to be too slow.

Rather than wait for a grid connection, OpenAI and its partners have bought everything from jet engines to turbines strapped to semi-trucks. They even tapped a cruise ship engine manufacturer for equipment.

At first, OpenAI’s strategy appeared to be working as expected. Its development partner Crusoe built and powered its first two buildings in 12 months—a faster timeline than any we’ve measured at Cleanview.

Soon after that, the first problems began to arise.

During its first winter in operation, Stargate Abilene went offline for days at a time due to issues with power and cooling equipment, according to The Information’s reporting.

It’s also not clear that the strategy has achieved its goal of building a gigawatt data center as fast as possible.

While Crusoe built the first two buildings of Stargate Abilene in record time, subsequent buildings have fallen behind schedule.

The company began construction on the third and fourth buildings in March 2025. They originally planned to finish in March 2026, according to building permits. But in June 2026 the data centers still aren’t online, according to a recent press release from Crusoe.

These delays allowed Anthropic and its partners to build a gigawatt-scale data center before OpenAI. The data center Amazon built to power Claude now has 1 GW of capacity, according to Cleanview's data center tracker—5 times as much as Stargate Abilene's.

Ultimately grid power, not jet engines, delivered the first gigawatt data center.

I wrote much more about all this in Cleanview's newsletter today. The in-depth story is adapted from Cleanview's behind-the-meter data center report, which we released earlier this month. A free exec summary of the report is available on our website.

[Rittenhouse Research @RHouseResearch]

Seth Klarman has been buying land adjacent to existing power infrastructure to flip to data center developers.. Did not have that one on my bingo card.

"We also own some raw land that we think could be sites for future data centers. It has adjacent power..

We know there's at the moment, and it could change, insatiable demand for data centers. The key thing is will it get built? Will it be permitted politically? And do you have power, which is the most important thing?

.. We're not sure we're going to build any data centers, but we can make them (the land sites) available to people who are needing more capacity."

A subtle but perhaps telling data point around the idea that the runway for growth amongst AI data center developers might be longer than expected..

$CIFR $CORZ $GLXY $HUT $WULF

[Rittenhouse Research @RHouseResearch]

Baseten CEO @tuhinone tells Altimeter's @apoorv03 that one of Baseten's cloud providers has already indicated their B200 prices ($/GPU hour) are set to double when existing contracts expire and are up for renewal later this year.

"If you go out right now saying you want a thousand GPUs, truly.. people are talking about Q2 of next year. So 12 months out, maybe 15 months out.

We have a cluster.. in one of these clouds.. of B200s.. Our unit price right now is $2.63 an hour.. that's up for renewal in October. They came to us already in May and said $5.10 is the new price.. So double."

[P Equity Research 📰 @pequityresearch]

J.P. Morgan - 57 AI Supply Chain Names

Passive Components

Murata $6981.T
TDK $6762.T
Taiyo Yuden $6976.T
Kyocera $6971.T
Panasonic $6752.T
Yageo $2327.TW
Walsin Technology $2492.TW
TA-I Technology $2478.TW
KOA $6999.T
Rohm $6963.T
Samsung Electro-Mechanics $009150.KS
Vishay Intertechnology $VSH

Test Equipment

Advantest $6857.T
HonPrecision $7769.TW
Chunghwa Precision Test Tech $6510.TW
Chroma ATE $2360.TW
Teradyne $TER
Cohu $COHU
Keysight Technologies $KEYS
CCTECH $300604.SZ
Accotest $688200.SS

Substrates

Ibiden $4062.T
Unimicron $3037.TW
NanYa PCB $8046.TW
Kinsus $3189.TW
Samsung Electro-Mechanics $009150.KS

PCBs/CCL

Meiko $6787.T
CMK $6958.T
Taiwan Union Technology $6274.TW
Gold Circuit Electronics $2368.TW
Dynamic Holding $3715.TW
Elite Materials $2382.TW
Iteq $6213.TW
Tripod Technology $3044.TW
Info-Tek $8183.TW
HannStar Board $5469.TW
Compeq $2313.TW
Unitech Printed Circuit Board $2367.TW
Zhen Ding Technology $4958.TW
Flexium Interconnect $6269.TW
Shengyi Technology $600183.SS
Victory Giant Technology $300476.SZ
Wus Printed Circuit $002463.SZ
Shennan Circuits $002916.SZ
Shengyi Electronics $688183.SS
Shenzhen Kinwong Electronics $603228.SS Guangdong Champion Asia Electronics $603386.SS Suzhou Dongshan Precision Manufacturing $002384.SZ
Doosan $000150.KS
ISU Petasys $007660.KS
TTM Technologies $TTMI

Power, Thermal, OSAT & Display

Chicony Power Technology $6412.TW
Sunonwealth Electric Machine Industry $2421.TW Taiwan Surface Mounting Technology $6278.TW
WIN Semiconductors $3105.TWO
E Ink Holdings $8069.TWO
Radiant Opto-Electronics $6176.TW

[Investing visuals @InvestingVisual]

The photonics market is expected to grow from $3B in 2025 to $91B in 2028.

The deep dive below covers everything you need to know about the industry, key players, risks and much more.

There are 10+ visuals included to help you fully understand it all. Hope you like it!

[Peter Berezin @PeterBerezinBCA]

Hyperscaler free cash flow is expected to fall to nearly zero in 2026 but then surge to $700 billion in 2030 despite $1 trillion in ongoing capital spending. Sure!

[Ivan Fioravanti ᯅ @ivanfioravanti]

Recent Opensource LLMs releases my rankings:

🥇 place: GLM 5.2
🥈 place: MiniMax M3
🥉 place: Kimi K2.7 Coding

But in any case I use them all!

[Jukan @jukan05]

JPMorgan’s Harlan Sur just published a report on AVGO — is this real? Is Broadcom really involved in Anthropic’s ASIC?

[Dan Nystedt @dnystedt]

SpaceX’s new plan to deploy 1GW of AI1 orbital data center satellites by end-2027 (annualized) has drawn big interest from Taiwan AI server suppliers Foxconn, Quanta, Wistron and optical communications suppliers LandMark, Elite Advanced Laser, media report, adding AI1 satellites are seen to be about 1 AI server rack each, and assuming they are based on Nvidia GB300, each rack would consume 135kW, so potentially 7,400 AI server racks annualized – robust orders for supply chain partners. Related optical communications component orders are expected to go to Lumentum and Coherent, benefiting InP wafer supplier LandMark and Coherent’s laser packaging partner Elite Advanced Laser. $SPCX $COHR $LITE #AI1satellite https://t.co/d0iRLlOPok

[Legion Hoops @LegionHoops]

Jose Alvarado:

“I wasn’t even supposed to be in the league now I got a ring now”

(h/t @HoodieBrunson)

[Khairallah AL-Awady @eng_khairallah1]

this is f*cking gold

How to build your first AI agent (Full guide)

if I had this a year ago, I would've shipped my first agent in a day instead of 2 weeks

in the right hands, this changes everything:

[Jérémie @jeremie0117]

Great analysis of SpaceX and Starlink by Oppenheimer Research. A must-read on Space, AI, and Drones.

[Alok Kumar @Alokkumarzz]

INSTEAD OF WATCHING AN HOUR OF NETFLIX TONIGHT.

This 60-minute Cambridge lecture by Demis Hassabis will teach you more about the future of AI than most people will learn in the next 5 years.

Bookmark it and give it an hour, no matter what.

[AilaunchX @Ai_Tech_tool]

🚨 Anthropic just showed a 27-minute workshop on how to actually do prompts for Claude.

Taught by the people who built it.

Free. No registration. No paywall.

I've seen $300 courses that don't cover what they teach in the first 8 minutes.

Watch it and bookmark it now.

[Khairallah AL-Awady @eng_khairallah1]

Anthropic engineer:

"You're not supposed to prompt Claude. You're supposed to build a system that prompts itself."

this is one of the best workflows I've seen in a long time

in this video he breaks down exactly how most people are using Claude:

- the 14% you lose to CLAUDE.md before typing a word
- the plugins that 95% of users have never installed
- the caching setup that keeps it at 95% hit rate and almost free
- why starting every chat from zero is the slowest way to use Claude

if you've been using Claude for more than a month and never left the chat window, you've been using one project when you could be running a team of them

instead of another show tonight, watch this

make sure to bookmark it before it gets lost in your feed

full guide in the article below

[David Orr @orrdavid]

I mean so much of this stuff is really. I just went down the $5715 Furukawa rabbit hole a couple hours yesterday and it's pretty incredible what you're buying there asset wise... the long term history of the company... they are a big arm of the Zaibatsus. And now they are improving the operation finally and it's already actually working.

[Jukan @jukan05]

According to a Chinese sell-side team’s meeting with a Japanese semiconductor equipment company, TEL clearly stated during the discussion that it is pushing for price increases.

TEL also said it has received explicit requests from DRAM and advanced logic customers for earlier deliveries and additional orders. All of its factories are running at full capacity, and the company has fully shifted from single-shift to double-shift production.

Even on the back-end side, based on communications with wafer fabs, some equipment categories are also becoming tight. These include DISCO’s dicing and grinding tools, as well as EVG’s bonding equipment.