Semiwiki on LinkedIn Semiwiki on Facebook Semiwiki on Twitter Register / Log in 
Array
(
    [content] => 
    [params] => Array
        (
            [0] => /forum/threads/tsmc-q1-2015.5882/
        )

    [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
)

TSMC Q1 2015

P

Pawan Fangaria

New member
TSMC reported Q1 results of 2015 today - TSMC Reports First Quarter EPS of NT$3.05

At NT$ 222.03 billion, the year-over-year revenue is impressive with 49.8% increase with EPS increase of 65%. However, the quarter-over-quarter revenue is almost flat or declined by 0.2% with 1.3% decline in EPS. In US dollar terms the revenue is lesser because of stronger dollar.

View attachment 13994

It's interesting to see that 28-nm process accounted for 30% of total wafer revenue and 20-nm accounted for 16% of total wafer.

TSMC expects lower revenue between NT$ 204 billion and NT$ 207 billion in the second quarter. The reasons - customers' business loss, inventory adjustment, and less favorable exchange rate. So, who are the customers other than we know?
 
Last edited:
From CFO Mrs. Lora Ho's report, among several causes, TSMC 2Q2015 revenue will get negative impact due to a customer lost business to a captive IDM. So I believe she was talking about Qualcomm didn't sell S810 to Samsung because Samsung (an IDM) started using it's own Exynos processors for Galaxy S6.
 
I think the most interesting news is that TSMC cut CAPEX by $1B after announcing it at the symposium last week. That is a pretty big change in one week? i wonder if Intel's $1B CAPEX cut this week had something to do with it? Best of luck to the equipment manufacturers this year.
 
It's interesting to see that 28-nm process accounted for 30% of total wafer revenue and 20-nm accounted for 16% of total wafer.
Click to expand...

By looking at the presentation materials, there are more interesting things related to the 28nm and 20nm contribution to the overall revenue.

Wafer Revenue by Technology
[table] style="width: 331px"
|-
| style="height: 20px; width: 139px" |
| style="width: 65px" | 1Q15
| style="width: 65px" | 4Q14
| style="width: 65px" | 1Q14
|-
| style="height: 20px" | 20nm
| align="right" | 16%
| align="right" | 21%
| align="right" | 0%
|-
| style="height: 20px" | 28nm
| align="right" | 30%
| align="right" | 30%
| align="right" | 34%
|-
| style="height: 20px" | 40/45nm
| align="right" | 15%
| align="right" | 13%
| align="right" | 21%
|-
| style="height: 20px" | 65nm
| align="right" | 12%
| align="right" | 11%
| align="right" | 16%
|-
| style="height: 20px" | 90nm
| align="right" | 7%
| align="right" | 6%
| align="right" | 7%
|-
| style="height: 20px" | 0.11/0.13um
| align="right" | 2%
| align="right" | 2%
| align="right" | 3%
|-
| style="height: 20px" | 0.15/0.18um
| align="right" | 13%
| align="right" | 12%
| align="right" | 14%
|-
| style="height: 20px" | 0.25/0.35um
| align="right" | 4%
| align="right" | 4%
| align="right" | 4%
|-
| style="height: 20px" | 0.50um and above
| align="right" | 1%
| align="right" | 1%
| align="right" | 1%
|-
[/table]


Source: http://www.tsmc.com/uploadfile/ir/quarterly/2015/16cBq/E/1Q15ManagementReport.pdf


While maintaining a flat revenue from 4Q2014 to 1Q2015, TSMC Advanced technologies (28nm and below) accounted for 46% of total wafer revenue compared to 51% in 4Q2014.

So how does TSMC make up the 5% drop from the 20nm revenue? The answer is they just sold more in non-advanced 40/45nm(+2%), 65nm(+1%), 90nm(+1%), and even the 0.15/0.18um(+1%). Because these are not leading edge technology so TSMC probably offered some very encouraging pricing to gain more orders. It also indicated that while those high profile nodes (20nm, 16/14nm, 10nm) are important but not everyone has to use them.

This is also an interesting situation for those new or potential new comers in the foundries business. Where is the market for them to begin with? Leading edge nodes will be too expensive and risky and no guarantee to implement successfully. If they turn to non-advanced nodes then they will face companies like TSMC who can afford to cut price to compete. I suspect some of those TSMC non-advanced node manufacturing equipment has been fully or almost fully depreciated on their accounting books. TSMC can afford to be very aggressive on those non-advanced nodes.
 
Last edited:
Daniel Nenni said:
I think the most interesting news is that TSMC cut CAPEX by $1B after announcing it at the symposium last week. That is a pretty big change in one week? i wonder if Intel's $1B CAPEX cut this week had something to do with it? Best of luck to the equipment manufacturers this year.
Click to expand...

Nah, it's because 20nm is winding down faster than they expected and they are just converting more of the 20nm to 16 instead of building new 16nm
 
By looking the TSMC report:

Wafer Revenue by Customer Type
[table] style="width: 331px"
|-
| style="width: 139px; height: 20px" |
| style="width: 65px; height: 20px" | 1Q15
| style="width: 65px; height: 20px" | 4Q14
| style="width: 65px; height: 20px" | 1Q14
|-
| style="width: 139px; height: 20px" | Fabless/System
| style="width: 65px; height: 20px" | 83%
| style="width: 65px; height: 20px" | 85%
| style="width: 65px; height: 20px" | 87%
|-
| style="width: 139px; height: 20px" | IDM
| style="width: 65px; height: 20px" | 17%
| style="width: 65px; height: 20px" | 15%
| style="width: 65px; height: 20px" | 13%
|-
[/table]

Source: http://www.tsmc.com/uploadfile/ir/quarterly/2015/16cBq/E/1Q15ManagementReport.pdf


The IDM contract share is growing. I assume a big portion is Intel's SoFIA. Will it continue?

In the long run, TSMC may turn out to be Intel and some other IDMs' (not Samsung) important partner. Both time to market demand and heavy foundry buildup cost can further encourage this kind of collaboration. Apple and Qualcmm have good R&D and make a lot money while owning no factory. Intel has good R&D capability and they don’t have to limit their own potential just because everything has to be made in-house.
 
Last edited:
Pawan Fangaria said:
The 5% drop in 20-nm would also mean loss of Qualcomm s810 (20-nm build) business to Samsung's own Exynos 7420 (14-nm build)
Click to expand...
Apple A8 uses TSMC 20nm. I'm not sure if Apple scaled down A8 order in the 1Q2015 due to the inventory, market demand, and the upcoming A9? Actually Qualcomm probably has already reduced their S810 order a while ago.
 
hist78 said:
By looking the TSMC report:

Wafer Revenue by Customer Type
[table] style="width: 331px"
|-
| style="width: 139px; height: 20px" |
| style="width: 65px; height: 20px" | 1Q15
| style="width: 65px; height: 20px" | 4Q14
| style="width: 65px; height: 20px" | 1Q14
|-
| style="width: 139px; height: 20px" | Fabless/System
| style="width: 65px; height: 20px" | 83%
| style="width: 65px; height: 20px" | 85%
| style="width: 65px; height: 20px" | 87%
|-
| style="width: 139px; height: 20px" | IDM
| style="width: 65px; height: 20px" | 17%
| style="width: 65px; height: 20px" | 15%
| style="width: 65px; height: 20px" | 13%
|-
[/table]


Source: http://www.tsmc.com/uploadfile/ir/quarterly/2015/16cBq/E/1Q15ManagementReport.pdf


The IDM contract share is growing. I assume a big portion is Intel's SoFIA. Will it continue?

In the long run, TSMC may turn out to be Intel and some other IDMs' (not Samsung) important partner. Both time to market demand and heavy foundry buildup cost can further encourage this kind of collaboration. Apple and Qualcmm have good R&D and make a lot money while owning no factory. Intel has good R&D capability and they don’t have to limit their own potential just because everything has to be made in-house.
Click to expand...

IDM probably refers to TI, Infineon, Freescale, NXP, ST Micro, Renesas and yes, finally, Intel. It's not just Intel or even mainly Intel. There are a whole boatload of IDMs who have gone fab-lite and work with TSMC.
 
One interesting thing I noticed: they clarified the "16 FinFET+ is 10% faster than competition" statement.
They mean 10% faster than Samsung 14nm LPP (that's the fastest version)
 
lefty said:
One interesting thing I noticed: they clarified the "16 FinFET+ is 10% faster than competition" statement.
They mean 10% faster than Samsung 14nm LPP (that's the fastest version)
Click to expand...
Are there independent comparative figures somewhere between 16FF+, 14 FinFET and 14 Tri-gate?
 
lefty said:
One interesting thing I noticed: they clarified the "16 FinFET+ is 10% faster than competition" statement.
They mean 10% faster than Samsung 14nm LPP (that's the fastest version)
Click to expand...

This was verified by a large fabless customer based on simulation models and test chip reports. I think TSMC referenced a paper but I know where the number came from. Comparing 16FF+ against 14nmLP is not really fair. It is like comparing the next iPhone SoC against the iPad SoC. In fact, it is exactly like it.

Next year Samsung will have a 14nm HP and TSMC will have 16FFC so lets wait and compare those for accuracy. I have not seen numbers on Intel 14nm. None of the fabless companies I know use it except Altera and their marketing is by far outpacing engineering.
 
Daniel Nenni said:
Here is the full transcript:

Taiwan Semiconductor Manufacturing Company's (TSM) CEO Mark Liu on Q1 2015 Results - Earnings Call Transcript | Seeking Alpha

It's a pretty good ready. Other comments would be appreciated.

It seems the CAPEX cut is due to a faster transition from 20nm to 16nm. Meaning they will convert 20nm to 16nm faster than expected. My guess is QCOM moved from TSMC 20nm to Samsung 14nm much quicker than anticipated. Maybe due to the rumored 20nm Snapdragon power issues.
Click to expand...

Now the TSMC official 1Q2015 Earnings Conference Transcript can be viewed at:

http://www.tsmc.com/uploadfile/ir/quarterly/2015/16cBq/E/TSMC 1Q15 transcript.pdf
 
You must log in or register to reply here.
Back
Top