Fake TSMC 7nm News from Seeking Alpha

[Daniel Nenni]

Seeking Alpha used to be a good source of information but not so much anymore. They really have embraced the fake news format to boost readership. The problem with owning a commercial website is the you must increase readership every year to represent growth. Fake news seems like a shortsighted strategy but in a world where anybody can post anything they want with relative impunity, fake news is now Standard Operating Procedure.

SemiWiki traffic experienced hyper growth the first 5 years but has been flat the last 3, which is fine by me. SemiWiki is the number one semiconductor blog site because there are no others. We are working semiconductor professionals that have reputations to protect so we are very careful what we write. Other sites don't seem to care and cut/paste their way into infamy.

In this case the rumor site DigiTimes reports that TSMC 7nm capacity "unlikely" to be fully utilized in 1H 2019. From there SA reports that TSMC "won't" use full 7nm capacity in 1H19. I'm not sure where it will go next, maybe TSMC 7nm will "never" be fully utilized? By the way, with the downturn we are predicting for 2019 no semiconductor processes will be "fully" utilized, my opinion.

I guess the good news is that there will always be a place for sites like SemiWiki. The bad news is that non semiconductor professionals will continue to have a distorted view of our industry, absolutely.

Report: TSMC won't use full 7nm capacity in 1H19
TSMC (NYSE:TSM) is unlikely to see its full 7nm process capacity utilized in 1H19, according to a Chinese media report via DigiTimes
The company has lower chip orders than expected due to order cutbacks from Apple and Qualcomm
TSMC could utilize only 80% to 90% of its capacity during that period
TSMC previously said it planned to tape out more than 50 chip designs with the 7nm tech by the end of this year and over 100 chip designs by the end of 2019
The company expects 7nm chips to account for 20%+ of Q4 wafer revenue and nearly 10% for the full year

 

[mhagmann]

Have any of you followed the sales of the new iPhone Xs and Xs-Max which are based on the TSMC 7-nm chips and were first sold in late October? I saw two tables full of them in my local Apple store.
Would any one make an estimate as to what fraction of the sales of these chips will be for the iPhones? What would the other markets be like in the first year or two?

 

[Daniel Nenni]

Have any of you followed the sales of the new iPhone Xs and Xs-Max which are based on the TSMC 7-nm chips and were first sold in late October? I saw two tables full of them in my local Apple store. Would any one make an estimate as to what fraction of the sales of these chips will be for the iPhones? What would the other markets be like in the first year or two?

I have heard by the supply chain that the new iPhones are not selling well. I have an iPhone X and see no extra value with the new phones. In fact, I will buy my daughters iPhone 8's to avoid the facial recognition feature. The problem is that Apple is not discounting prices for the holidays. I normally buy iPhones on Cyber Monday but there were no big discounts on any iPhone unless you buy a reconditioned one.

Let's see what they do for Xmas but my guess is that Apple will miss iPhone sales targets by a wide margin. We will know from TSMC in Q1 2019. 7nm was supposed to be 20% of revenue and that is Apple.

 

[Portland]

Seeking alpha is a more general site. They don't only follow semiconductor companies but stocks in general like nike and ge. What I think happened is they have sources that gave bad info that could be agenda driven or just wrong.

 

[raghu78]

I have heard by the supply chain that the new iPhones are not selling well. I have an iPhone X and see no extra value with the new phones. In fact, I will buy my daughters iPhone 8's to avoid the facial recognition feature. The problem is that Apple is not discounting prices for the holidays. I normally buy iPhones on Cyber Monday but there were no big discounts on any iPhone unless you buy a reconditioned one.

Let's see what they do for Xmas but my guess is that Apple will miss iPhone sales targets by a wide margin. We will know from TSMC in Q1 2019. 7nm was supposed to be 20% of revenue and that is Apple.

Daniel i think TSMC 7nm will do fine. TSMC will probably ship a bit lower for Q4 2018 and Q1 2019. But from Q2 2019 AMD 7nm Navi, Rome should launch and start to drive volume at TSMC 7nm. Apple's loss is AMD's gain. There is tremendous anticipation for AMD 7nm server and client CPU and 7nm Navi GPUs in 2019. 7nm EPYC Rome promises to be a disruptive upgrade in terms of generational performance with 2x perf per socket over 14nm EPYC Naples. AMD will take every 7nm wafer that TSMC can ship in 2019 and 2020. I do not see Intel being able to stop AMD Rome in 2019 and Milan in 2020. AMD's server share should go past their all time high of 26% by late 2020.

 

[name99]

Have any of you followed the sales of the new iPhone Xs and Xs-Max which are based on the TSMC 7-nm chips and were first sold in late October? I saw two tables full of them in my local Apple store.
Would any one make an estimate as to what fraction of the sales of these chips will be for the iPhones? What would the other markets be like in the first year or two?

Speaking of fake news ^^^
Every year we get some variant of this, every year it's nonsense.

If you want to tackle the issue of fake news, ask yourself why you're so comfortable believing this. What track record do the people pushing this story have? What's their methodology based on, and what track record does that methodology have?

 

[Daniel Nenni]

Daniel i think TSMC 7nm will do fine. TSMC will probably ship a bit lower for Q4 2018 and Q1 2019. But from Q2 2019 AMD 7nm Navi, Rome should launch and start to drive volume at TSMC 7nm. Apple's loss is AMD's gain. There is tremendous anticipation for AMD 7nm server and client CPU and 7nm Navi GPUs in 2019. 7nm EPYC Rome promises to be a disruptive upgrade in terms of generational performance with 2x perf per socket over 14nm EPYC Naples. AMD will take every 7nm wafer that TSMC can ship in 2019 and 2020. I do not see Intel being able to stop AMD Rome in 2019 and Milan in 2020. AMD's server share should go past their all time high of 26% by late 2020.

I agree, other TSMC 7nm customers are probably already fighting over the 7nm wafers Apple may not consume. Nvidia, QCOM, Apple, MediaTek, etc... I will be in Hsinchu next week. It will be interesting to see what the mood is.

 

[ChrisGar]

Speaking of fake news ^^^
Every year we get some variant of this, every year it's nonsense.
If you want to tackle the issue of fake news, ask yourself why you're so comfortable believing this. What track record do the people pushing this story have? What's their methodology based on, and what track record does that methodology have?

Did you have the right quote in your response ? Mainly I just the person asking questions -- not reporting any "news." (unless you don't believe what the person saw)

 

[benb]

This feels like late 2010 or early 2011, when TSMC epic ramp of 28nm was underway. Why was 28nm so popular (and why I think 7nm will be the same):
- Genuine technology improvements, rather than incremental. In 2010 it was HKMG (RMG), today it's the resumption of Moore's law scaling, with cost reduction and performance improvement. The potential flaw is an egregious cycle time, which TSMC (and Samsung) will address by gradually removing quad patterning with EUV.

There is some doubt and fear in the air, like 2010. I don't see high-end smartphone SOCs being the driver like they were for the last decade going forward. In 2010, the big 28nm application for TSMC was graphics cards and FPGAs; SOCs were still small. In 2019 a mature smartphone market may have to consolidate considerably. The fallout from the crypto bubble is underway, graphics chips will suffer too. And semiconductors, which are integral to economies and grow at GDP pace now, could lead us into a recession.

Even if we have a recession or are in one now, 7nm has given new life to many existing applications, which will gain AI functions, better graphics, faster speeds, and lower power operation. The slowdown, recession or downturn doesn't affect the leading edge (16nm and below). There could be a big drop in 28nm though, affecting TSMC, the biggest 28nm producer. So I would say, TSMC could experience an overall drop in revenue as their cash cow, 28nm, declines but their 7nm will fill the gap, eventually, perhaps not right away.

If I had to make a prediction, it would be this: 2019 will see many, many products migrate to the teens (16/14/12/11nm), 7nm will be off to a good start (although it really needs EUV to be fully productive), and 28nm will disappear quickly.