Optics remains the core sector driving AI infrastructure, but the days of simply chasing keywords are over. A deep understanding of the specific technological 'chokepoints' each company controls within the value chain is now absolutely essential. Damnang2 ’s latest piece

While the market fixates on flashy GPU specs and front-end node shrinks, the real battleground for the next AI hardware cycle has shifted to the very end of the line: Assembly. In the era of 16-Hi HBM4, chiplets, and CPO, ecosystem dominance now depends entirely on how precisely

Honored to be mentioned by Vikram Sekar — a regular read for me. I’ve also been working through the board-level VRM side in my own series, so it could be interesting to compare how the picture looks from different layers of the stack. Always happy to hear where I should sharpen

This whole series is great! We need to rethink power in the age of AI, and there are a number of ambitious founders tackling the worlds hardest power problems American Terawatt is building the HVDC grid we desperately need.

Photon Capital, Nutty, and I often meet to discuss the next major theme. We strongly agree that after optics, the next key theme is the AI data center power stack. It is a very interesting area, and I will cover it in depth soon.

x.com/i/article/2051…

This post was published in collaboration with Damnang2 and Photon Capital, two writers who consistently share exceptional insights. Even if the latest GPU arrives at a data center, compute does not begin unless power and cooling are ready. In the 1MW rack era, the next bottleneck

The best performing thematic basket in our universe since the recent market lows is our “AI Power Plumbing” basket - comprised of analog semis names making capacitors, PSUs, inductors, magnetics, MLCCs and discrete power semis. We were a bit early to this - at the end of 2024 ,

Optical test stocks all look good. That's the trap. They get bucketed under the same "CPO test" theme, but some names need CPO to arrive fast to make money, while others get carried by the pluggable cycle even if CPO is delayed. Companies sitting far apart on the matrix end up in

x.com/i/article/2051…

AI power story usually starts with 800V, rack power, and data center electricity. But one bottleneck sits much closer to the GPU. This free essay explains why tiny MLCCs became a critical part of AI power delivery, and why the story is not simply “more AI servers need more

x.com/i/article/2051…

Capital can fund a turbine. Capital can pull permits. Capital cannot easily compress the physics of how a 500-megawatt transformer gets built. That is the layer this piece is about.

What Damnang2 shows especially well in this piece is how a broad theme like AI power can be turned from a simple list of beneficiaries into a map investors can actually compare across companies. Companies grouped under the name “AI power” may sound similar, but in reality they

Great read from Photon Capital. Glass substrates are often discussed as a material transition, but this piece frames them more usefully as a packaging architecture shift — from EIC to CPO and eventually EPIC. That distinction matters.

This is such a witty response to my article that I almost wanted to put it directly into the piece 😆

[Investment Map] 15 Companies in the Glass Substrate Cycle: From Material to Mass Production Written with Nutty and Damnang2. One-year returns: +850%. +675%. +554%. +341%. +328%. All sitting on the glass substrate supply chain. Which stocks? That unfolds inside. The GVM

x.com/i/article/2052…

This is not just a country-level introduction to a few companies. It is a story about a nation — and the key companies within it — sitting at one of the most critical bottlenecks of the AI era: semiconductors.