Serenity复盘AI光通信、MLCC、InP与CPO激光供应链主题，未披露新的可核验公司催化

off the top of my head, just a recap of the themes I'm tracking:

- MLCC (actively price hiking AI server MLCCs, tightening for consumer + general purpose).

Early NAND cycle similarities, but not sure if we'll see the same type of price hikes since some LTAs were already signed we don't have a US player as aggressive as Sandisk?

Eg. Holy stone, Taiyo Yuden, Samsung electro-mechanics, Murata etc.

- InP Substrates - Extreme severity depending on who you ask:

JX/AXT/Sumitomo Electric. Ask downstream laser suppliers like $LITE, $AAOI - fine for 2027 (which is the funny part about channel checks)

Ask epiwafer suppliers like Landmark/VPEC - world is gonna end

- CW Laser (extreme severity) - many CSPs + semi leaders approaching all suppliers like $AAOI, $MTSI, to secure supply cause $LITE / $COHR have too much allocation to $NVDA.

And the latter two might have some agreements with other companies with capacity like Furukawa/Sumitomo.

-> Since $COHR said they were just using laser capacity internally and $LITE was buying EML off Coherent before if i remember correctly... So Lumentum not gonna have enough space

So my guess in $SIVE earnings is a lot of companies are probably approaching them now to do qualifications + secure allocation.

I don't know how much Sivers secured with Win and their other fabs (they started they were working with 3 back in like 2023-2024?), but we'll find out more in 3 days.

I still think their $70m fundraise might be related to securing allocations.

- CW Laser Foundries - This is Win Semi / Taiwan players.

I was surprised how little pricing these guys get compared to actual laser die output. So sticking with $SIVE and more downstream was right call for me.

Since gross margins for CPO lasers output was like 55-60% or something high (likely higher over time).

- Hybrid Bonding (rip, sorry) - pushed back to like 2029-2030 eg. Auros

- Glass Core Substrates like $LPK (ehh) - pushed back like half a year for early ramp from Samsung Electro-mEchanics + SKC Absolics. not the end of the world, but likely full ramp is 2029

- Epiwafers: Capacity constrained already eg. $IQE / Landmark

- Quantum Dot (the fun one):

QD actually surprised me most by how much focus is going into it after I looked at Sk Hynix CPO nature paper.

Probably US national security to work lessen the inP bottleneck with $ALMU and others.

My guess is $ALRIB + Veco gonna be like $AIXA for this QD equipment chokepoint in like 1 year. Most people just dont understand what Riber does is my guess.

But for actually investing in quantum epitaxial process, TAM IS MEGA MEGA SMALL. Cause it's a subset of epiwafers, and $IQE is actually a leader in this space.

You're better off looking at lasers specs these QD players produce? But it the material is super important.

For robotics (eg. battery, actuators) - my opinion is that mass production # is still too small. So I'm not really allocating much funds into upstream suppliers.

It takes a downstream leader like $CCXI, Unitree (like OpenAI/Anthropic) to hard carry the sector first, so I'm sticking with the humanoid players first.