Semiwiki on LinkedIn Semiwiki on Facebook Semiwiki on Twitter Register / Log in 
Array
(
    [content] => 
    [params] => Array
        (
            [0] => /forum/threads/analyst-says-sell-tsm.5319/
        )

    [addOns] => Array
        (
            [DL6/MLTP] => 13
            [Hampel/TimeZoneDebug] => 1000070
            [SV/ChangePostDate] => 2010200
            [SemiWiki/Newsletter] => 1000010
            [SemiWiki/WPMenu] => 1000010
            [SemiWiki/XPressExtend] => 1000010
            [ThemeHouse/XLink] => 1000970
            [ThemeHouse/XPress] => 1010570
            [XF] => 2021770
            [XFI] => 1050270
        )

    [wordpress] => /var/www/html
)

Analyst says sell TSM?

E

EDAGuy

New member
Here you go Dan I know you like Barrons:

By Shuli Ren
Taiwan-based semiconductor foundry TSMC (2330.TW/TSM) is a divisive stock.

TSMC has two big problems. First, two key customers – Apple (AAPL) and Qualcomm (QCOM) – account for 30-40% of TSMC’s business. So if one of them ditches TSMC, the semiconductor foundry’s top-line will take a hit. Second, competitor Samsung Electronics (005930.KS/SSNLF) is ramping up its semiconductor business to make up for its losses in the smartphone segment. Samsung is now bullish that its 14nm node, a more advanced production than what TSMC has, will win Apple’s A9 business this year.

TSMC: Key Customer Risk, Maybank Says Sell - Asia Stocks to Watch - Barrons.com

<script src="//platform.linkedin.com/in.js" type="text/javascript">
lang: en_US
</script><script type="IN/Share+init" data-counter="right"></script>
 
Last edited by a moderator:
On Jan 11, A Barron’s story reports the following:

1. Samsung is producing its own Exynos on 14nm node. (This contradicts the pro-Samsung PR of on-going A9 production.)

2. The yield of the above production reached 60-70%.

3. Samsung will produce iPhone 7 SoC; TSMC, iPad. Similar to Nenni’s view. That is, Samsung will produce most of A9s

4. The article is bearish on TSMC.

This article cites an analyst, and is better than the unnamed “insiders” PR nonsense.

TSMC: Key Customer Risk, Maybank Says Sell - Asia Stocks to Watch - Barrons.com

As I stated before, I disagree with such analysis; TSMC will produce most of A9s.

Anyway, TSMC’s 4Q conference call is this coming Thursday. As usual, Apple won’t be mentioned directly. But, the revenue forecast will give strong hint. If the revenue is to grow double digits this year, then my view is probably correct. If the forecast is flat or lower, then Samsung has won most A9s.
 
Last edited:
Some of the reasoning of that article is suspect:
While TSMC has narrowed the gap by pulling forward its 16nm development, we think Samsung is still six months ahead in terms of commercial production. This is evident by its production ramp up of in-house Exynos7420 AP in 4Q14 which has yielded 60-70% and is expected to reach 80%+ in 1Q15.
Click to expand...
A quick search shows the Exynos 7420 is just a rebranded Exynos 5, which is a 20nm chip, not 14nm. (Samsung rebrands its first 64-bit chip to Exynos 7 Octa | Greenbot)
Production of a 20nm chip does not prove that Samsung is ahead with 14nm.
 
This one is from Morgan Stanley:

TSMC delays 16nm equipment move-in pushing production schedule from 1H to 2H15, per UDN. The article says Samsung's 14nm process may beat TSMC for most orders of AAPL's next-generation A9 processor. TSMC has guided for 16nm to start volume production in end-2Q15 or early 3Q15 and Bill Lu believes this is on track as 16nm yield has been improving rapidly. He thinks the delay is likely a result of AAPL postponing the decision on order split between Samsung and TSMC.

Why these folks think moving a design from one foundry to another is like flipping a switch I do not know... Who is UDN? The TSMC conference call on Thursday should be interesting!
 
UDN...maybe United Daily News?

That said...I'd be inclined to believe that Samsung won the Apple A9 orders given Morris Chang's previous statements regarding TSMC falling behind Samsung in 16/14nm marketshare. His comments sounds like to me TSMC is concentrating more of its resources to 10 nm.
 
Daniel Nenni said:
Why these folks think moving a design from one foundry to another is like flipping a switch I do not know... Who is UDN? The TSMC conference call on Thursday should be interesting!
Click to expand...
They simply know nothing. Unless Apple did a double design and tape-out at both Samsung and TSMC (that I still do not 100% exclude), then making a change without massively delay the chip release is impossible. Even assuming this scenario, timeline is already very tight to make big changes to the final supply chain. My bet is that both (or we must better say, all the 3) foundries, took a risk and overbooked/overplanned 14nm capacity trying to compete until the end for the biggest slice of the Apple Pie.
I believe, TSMC was willing to take the risk in this winner takes all race, encouraged by the poor yield performance of the competitor, but eventually had partially to give up, after the solid recovery that happened on the last months in Samsung and GloFo fabs. Considering all the rumors, I say that the Daniel´s guess is still the most reasonable to date. Probably, it is not anymore a guess.
 
TSMC may have some short term issues with Apple but long term is another story. They are the only pure play foundry that has a track record of staying on the leading edge and executing on volume ramps at new nodes.




astilo said:
They simply know nothing. Unless Apple did a double design and tape-out at both Samsung and TSMC (that I still do not 100% exclude), then making a change without massively delay the chip release is impossible. Even assuming this scenario, timeline is already very tight to make big changes to the final supply chain. My bet is that both (or we must better say, all the 3) foundries, took a risk and overbooked/overplanned 14nm capacity trying to compete until the end for the biggest slice of the Apple Pie.
I believe, TSMC was willing to take the risk in this winner takes all race, encouraged by the poor yield performance of the competitor, but eventually had partially to give up, after the solid recovery that happened on the last months in Samsung and GloFo fabs. Considering all the rumors, I say that the Daniel´s guess is still the most reasonable to date. Probably, it is not anymore a guess.
Click to expand...
 
At least, one analyst believes Apple had taped out at both TSMC and Samsung.

Jan 12, 2015:

"our recent industry checks and feedback from the Korean supply chain suggests the A9 will be based on 16/14nm FinFET with both Samsung/GF (primary) and TSMC as suppliers. This comes as a surprise to us, as this is the first time that Apple will split a single chip with more than one supplier, even though the two camps have very different processes and technologies, which require not only redesign but also mask, IPs, etc, expenses on both sides. This is fairly unusual based on all the cases we have seen before."

Samsung And TSMC Will Share The Apple Pie: CLSA - Asia Stocks to Watch - Barrons.com

However, I disagree with the analyst’s view of Samsung gaining 80% of A9. I don’t remember any analyst predicted TSMC to produce all of A8s. They are likely to be wrong again this time.

Anyhow, Thursday is only 2 days away.
 
The following is from a long-time Intel promoter, who is usually not friendly to TSMC and still pumping Intel’s lead in “logical density” that fails to appear in physical density.

"I must admit to a degree of skepticism with respect to Samsung's 14-nanometer foundry efforts. After all, the story over the last several years has been that TSMC seemed to release new technology generations ahead of Samsung, which led to the former raking in most of the profit dollars available to the semiconductor logic foundry industry.

"To put this into perspective, TSMC has generated approximately $8.62 billion in operating income over the last 12 months, while Samsung's foundry division is reportedly on track to lose just under $800 million this year."

"I remain skeptical that the yields on the 14-nanometer Exynos part were that high in that time frame given other data points"

Did Samsung Just Gain an Edge on Taiwan Semiconductor Mfg. Co. Ltd.? (SSNLF, TSM)
 
Let's wait and see (tomorrow is the day).
Anyway, it looks like during your copy and paste job, you missed some portion of the last sentence.

While I remain skeptical that the yields on the 14-nanometer Exynos part were that high in that time frame given other data points, the report seems positive for Samsung.
Click to expand...
Just to be fair
 
Here is a full paragraph quotation from The Motly Fool article mentioned above:

"... The report also asserts that Samsung's Exynos 7420 applications processor, claimed to be built on Samsung's 14-nanometer process, has yielded in the range of 60% to 70% during the fourth quarter and is "expected to reach 80%+ in 1Q15." While I remain skeptical that the yields on the 14-nanometer Exynos part were that high in that time frame given other data points, the report seems positive for Samsung. "

Did Samsung Just Gain an Edge on Taiwan Semiconductor Mfg. Co. Ltd.? (SSNLF, TSM)

Yet per published information, Samsung Exynos 7420 is made with 20nm process and first showed up in the Galaxy Note 4 in 4Q14. Unless the Maybank analyst (Warren Lau) quoted by Barrons knows something that nobody else does publically, i.e. that there is a 14nm variant of the Exynos 7420, the discussion of yield is totally confusing and misleading, is he referring to the yield of 20nm or 14nm Exynos 7420 ??
 
Here is the latest from the analysts:

TSMC may cut 20% of 20nm production as iPhone orders dropped following a seasonal peak, while TSMC may have also halted 16nm FinFET trial production as QCOM shifted orders to Samsung with improved yields on 14nm FinFET, per Liberty Times. Chatter suggests that TSMC may cut 20nm production by 20% because of (1) order loss to Samsung given improving Samsung 14nm FinFET yield, and (2) overheating problem in QCOM's Snapdragon 810 series resulted in delay of time to market. The article concludes that TSMC 1Q revenue may miss estimates so all eyes on their print tomorrow AM.

TSMC's Fourth Quarter Earnings Conference and Conference Call will be held in Taipei on Thursday, January 15, 2015, 2pm Taiwan time/ 1am ET. The Event will be held in English. Due to space limits, it is open to institutional investors only. The Quiet Period is January 5 to 14, 2015. During this period, TSMC refrains itself from making contact with the investment communities.

[table] cellpadding="4" style="width: 100%"
|-
| align="left" |
iconMic.gif
(Broadband) Earnings Conference Webcast
|-
| align="left" |
iconMic.gif
(Narrowband) Earnings Conference Webcast
|-
| align="left" |
iconMic.gif
TSMC 4Q14 Conference Call Dial-in Information
|-
[/table]

Hopefully the Q&A will be informative. I do know one of the institutional investors (he is on SemiWiki) and have sent him a list of questions. Add yours here and I will get them to him.
 
astilo,

The article is linked. Reader can read the original, and see clearly that, despite that the Maybank report is positive on Samsung, this writer, who does not have a pro-TSMC bias, is skeptical of the pro-Samsung arguments.

Since there is no definitive evidences either way, that’s the best can be say – skeptical.

Is it possible that suddenly Samsung surpasses TSMC in logic chip technology? It’s possible, but not likely. I am in “wait and see” mode, so are most knowledgeable people.

Readers will see who is the one with the heavy pro-Samsung bias.
 
Last edited:
Since I started following TSMC 10 years ago, there are always plenty of skeptical or negative reports on TSMC. Majority of time, the analyst ratings are neutral or sell. But, TSMC continues to grow, out of a lot of doubts, to the point of exceeding Intel, the supposed leader, in some areas.

At beginning, even Taiwan government is skeptical of the fab-less model. The government’s bet was on UMC, founded in 1980, modeled after the prevalent IDM at the time, long before TSMC of 1987.

Today, TSMC is over 5 times the size of UMC, using the same pool of local engineers, technicians, and research facilities (ITRI and universities).
 
Another interesting read...seems to support (Dani's popular theory of Applie splitting Samsung/TSMC orders):
Apple may shun Intel for custom A-series chips in new Macs within 1-2 years

KEY TAKEAWAY
Samsung is expected to split orders for the iPhone's next-generation A9 with Global Foundries, while TSMC would handle the A10 in 2016. The A9X — which would power both the iPad and a new low-end Mac — would be fabbed on TSMC's 16nm line, with the A10X moving to Samsung's 10nm plant.
 
There could be many reasons for overheating. To simplify, it could be caused by process issues or design issues (or a combination, or a result of miscommunication between the foundry and a chip designer etc.). If this leads to losses for TSMC, I wonder if the customer is also liable for some damages? How do these issues get resolved in the industry in general?
 
The Snapdragon 810 problem will be fixed, by it will take some time. I heard about a possible change in some of the metal layers already ongoing. But time is very pricey, and a single quarter delay could cost several missing design wins (moreover high end smartphones). Since Samsung is one of the top smartphones seller (meaning one of the biggest QTI customers), Qualcomm could divert some products from TSMC to Samsung trying to get a better overall deal. About a possible liability issue, I do not expect any problem there: it is part of this risky business. Moreover, both the customer and the foundry, work very close together to meet their goals.
I do not think, anyway, that during the conference call, customers (either QTI or Apple) will be mentioned directly.
 
Thanks for all the juicy rumor links. If Apple can build a 4-to-8-discrete-chip laptop, with A9X or A10X, that sounds very cool. What a motherboard that would be though.

The Digitimes article helps substantiate what I've been saying on and off, regarding scaling at 20nm and below: It's getting harder and harder. Maybe this 9-12 month smartphone/tablet cycle is biting off more than many supply chain players can chew. Intel's Core M was a dud, it looks like Qualcomm's Snapdragon 810 is going to have issues too. I'm not a fan of half-baked engineering; better to take 18 months and get it right, in my opinion.
 
You must log in or register to reply here.
Back
Top