[3/4/25 Market Memo] Panther Lake Delay, $TSM $INTC $QCOM $CEVA $SITM
More evidence that 18A parametric yield is dogshit.
Irrational Analysis is heavily invested in the semiconductor industry.
Please check the ‘about’ page for a list of active positions.
Positions will change over time and are regularly updated.
Opinions are authors own and do not represent past, present, and/or future employers.
All content published on this newsletter is based on public information and independent research conducted since 2011.
This newsletter is not financial advice, and readers should always do their own research before investing in any security.
Feel free to contact me via email at: irrational_analysis@proton.me
Lots of news today. Let’s get started.
Trump chooses TSMC, throws 18A into the dumpster.
Need I say more?
IFS is omega fucked. TSMC has announced multiple facilities, including a R&D Fab and an advanced packaging facility.
There is no timeline on this “$100B” investment into the United States. Also, no clarity on if this $100B is in addition to the already announced $65B (only $35B added on) or not.
Regardless, TSMC execs played this well. Being forced to take over Fabs configured for someone else’s nodes would have been a financial and operational disaster.
Clearwater Forest (18A) has already been massively delayed.
Now Panther Lake (18A) is also delayed… missing the all-important holiday shopping season.
I have been screaming for over a year that 18A **PARAMETRIC YIELD** is a disaster because the PDK is bad.
Shills who keep clinging to the D0 excuse… WAKE UP.
WATCH WHAT THEY DO NOT WHAT THEY SAY.
Let’s see how the Frank Yeary idiot handles this. Try to unlock shareholder value now you clown.
Fire the entire board, re-hire Gelsinger.
I was at a dinner with Dylan and he liked my meme analogy so much he tweeted it lmao.
Once the dust settles, I have a much more inappropriate analogy based on an obscure film.
It’s gona be so fun. Very excited for the creative writing and artistic opportunities.
SiTime iPhone Content Update:
Well someone else (FomalHaut) figured out what I could not.
In my defense, I found the Quartz crystal and that TCXO looks like a capacitor,
Should have spotted the clockgen TBH. It literally says ‘XO’ on the markings.
SiTime’s official LinkedIn account liked the post. The CEO also confirmed it today at the Morgan Stanley TMT conference. Multiple buy-side contacts were basically live texting me what he said lmao.
SiTime friends… those investors asking you very specific questions on PPM synchronization vs jitter are all coming from me.
If you would like to meet me in-person sometime after March 14th, I would be happy to clarify the technical bits.
UCIe (chiplet interconnect) is a clock-forwarded SerDes.
https://people.engr.tamu.edu/spalermo/ecen689/lecture25_ee689_clocking_arch.pdf
Clock-forwarding has many advantages. The most important one is guaranteed 0 PPM.
Imagine if you could get some of the benefits of clock-forwarding in other SerDes like Ethernet, InfiniBand, PCIe, UALink, …
Your products enable this in the datacenter market.
It is a unique advantage.
A unique performance advantage.
If you want to spend time at the meeting discussing the “correct” way to measure jitter, we can do that. I happen to know how to measure jitter.
There are much more interesting ways to spend a possible meeting.
Let me know.
Apple Modem Aggressive Investment Idea
This is for hedge funds and institutional investors. Retail traders, please do not do this unless you are experienced and understand the risks.
I have a strong gut feeling that Apple is going to ramp their modem MUCH more aggressively in the September iPhone 17 launches.
The way to play this in a macro neutral manner is…
LONG: CEVA 0.00%↑ SITM 0.00%↑ QRVO 0.00%↑ SWKS 0.00%↑
SHORT: QCOM 0.00%↑
Qualcomm is in omega cope mode. They have nothing to offset any revenue losses in excess of the ~20% Apple share loss telegraphed to sell-side.
Tim Apple would be delighted to punch Cristiano Amon in the face. Without a doubt, this is the worst relationship in all of tech.
Apple could conceivably go 80% internal modem in 2025. It’s actually possible from an engineering perspective.
99.9% of people who buy smartphones have no clue what any of these terms mean.
Who is the target audience for this marketing? Other Qualcomm employees?
Let’s go one by one.
Sub-6 Bandwidth
How much sub-6GHz (useful 5G) spectrum a modem can process.
Carriers rarely allocate so much spectrum to a single user.
160 MHz is perfectly fine.
98% of iPhone 16e users will not notice this.
Downlink Carrier Aggregation
Intuitively, how many cell towers your phone can connect to at one time.
This is needed for handover (user moving out of range of a tower) and coverage.
3x DL CA is good enough.
Again, wireless service providers (real-world networks) will not allocate tons of carriers to a single user who does not need it.
98% of iPhone 16e users will not notice this.
Uplink MIMO
How many data layers (via antenna diversity) can be used in uplink.
This actually does matter for people who like to upload videos and photos to social media over cellular data.
Switched Uplink
Using uplink slots on a TDD and FDD band at the same time.
Unclear if the Apple C1 modem is so shit, it does not support 3GPP release 16 switched uplink… lol.
Not supporting release 17 switched uplink enhancements is expected given how most of the worlds infra is still on release 15/16.
6 Rx
Nobody cares.
It blows up BOM cost due to more RFFE.
Most Android OEMs skip this.
Cope harder buddy.
mmWave
Nobody cares.
mmWave was a commercial failure in 2019 and it is still a commercial failure.
COPE. HARDER.
AI Features
Apple has total ecosystem control and is collecting massive amounts of data.
I’m sure they will get some AI modem features next year given how many engineers have migrated.
I wrote in my last Substack post that nobody needs any higher speeds. Most mobile features, even mid range, is sufficient for 98% of consumers. supporting all these features that nobody will notice only drives up cost.
Apple won’t reduce the phone cost. They will increase it, build mediocre modems that work well enough and pocket the extra cash.
CC stated at the presentation with Trump that the 100b was in addition to the 65b