18 Comments
User's avatar
WritedownCap's avatar

On AIXA, curious how you build up to your InP tool demand? Is that or is SiC and GaN tools the main opportunity?

Jason's Chips's avatar

InP tool build is bottoms up based on epi intensity for EML vs CW vs UHP. I’ll release my full model once I do an article on it. That is probably (?) the main opportunity but GaN and SiC could def surprise me. Haven’t done bottoms up on the latter yet but top down generally looks big.

WritedownCap's avatar

Cool looking fwd to it, I did a bottoms up build on InP but the assumptions stacking makes it hard to underwrite… could either be massive or very meh, excited to see what you come up with

yc's avatar
Apr 3Edited

Great stuff as always, and thank you. Fantastic year-to-date.

Would love to see an in-depth response on the SOI short thesis.

@boymanrobshit's avatar

Dumped TSEM? Seems misguided

Jason's Chips's avatar

revised down my numbers after a SemiAnalysis core research note. they seem to have a great hybrid bonding offering though. but i don't think they can capture majority of CPO share against COUPE. and you have EML taking share at 400G. So picture is too mixed after the massive rally to rank high vs LITE/AIXA/SOI. I still like them but the optics section of my portfolio is too crowded to take a #5 name.

@boymanrobshit's avatar

I kid, I’m gonna look for the research note that you mentioned

@boymanrobshit's avatar

I think you’re gonna regret that one for the sake of the theoretical diversification lol.

I’m not sure CPO even has to be part of the equation for the bull case case.

Jason's Chips's avatar

I appreciate the pushback. I do think I sold too early fwiw.

CPO does have to be part of the equation for the bull case. Or else you run into serious terminal value issues. Right now the share vs TSMC is very unsettled. They are a name I liked when they are cheap, not a growth-at-any-price call like LITE. I've said so in my thesis. But I'm open to hearing why you like them and your own ests if you have them. Happy to be proved wrong by the market over time and learn something. Just made the decision based on my own r/r assessment.

@boymanrobshit's avatar

Splitting hairs base case vs bull i guess. tsm always overwhelming favorite to win cpo. it’s not really new information. It was an option

70% prepayments locked on 5X capacity expansion through 2028 in hypergrowth market that doesnt have a clear top in TAM yet. Robotics etc.

Anyway, who knows. I just think you sold early 😀

@boymanrobshit's avatar

Following up on this - read a super insightful interview with a Meta infrastructure engineer. They’re architecting towards two separate stacks simultaneously — NVDA designed CPO AND inference stack with pluggable transceivers so they can control their own supply chain. Imo at least A Super bullish $TSEM because the natural divergence of these two stacks creates hyper growth in both scenarios. CPO is just a double plus on the bull case. 🤷‍♂️

@boymanrobshit's avatar

Point is - may dispute your terminal value issues because the inference stack grows in tandem and may eventually outpace spending 🤷‍♂️

Mathew Reinhart's avatar

Hey Jason, do you have much opinion on Sivers or IQE?

Jason's Chips's avatar

Sivers doesn’t have the power level of lasers needed for CPO. Which means I’m not really interested at these multiples. As for IQE I don’t know enough to give you a solid opinion.

Brian's avatar

What probability do you assign to VCSEL potentially winning CPO share? From what I can gather, if that were to happen, the bull cases for LITE, AIXA and SOI would all be impacted.

Jason's Chips's avatar

Not technically 0 because there can be some short reach super wide/parallel applications, but you'll need InP for any normal reach/datarate combo

Jason's Chips's avatar

0%. See my intro to sipho