TSMC Q3'14: No 16nm in 2015, no EUV at 10nm

[witeken]

TSMC Q3'14: No 16nm in 2015

TSMC will start HVM of it's 20nm "16nm" FinFET node in the middle of next year. Their 10nm node won't use EUV and is now expected to reach trial production in 2016. Both of those dates are later than previously disclosed (Q1'15 and 2015).

TSMC now expects to “pull in” the production schedule for 16 nm FinFET chips to the second quarter or early third quarter of 2015, Wei said.

TSMC’s 16 nm program “looks very smooth,” said Rick Hsu, an analyst with Nomura Securities, speaking after the earnings event. “It’s going to be about a quarter ahead of schedule.”

With 16 nm “the real volume will be in 2017 and onward,” said co-CEO Mark Liu.

By the fourth quarter of 2015, 16 nm FinFET chips will account for a “high-single-digit” percentage of the company’s sales revenue, Wei said.

10 nm expectations
[...]

“The goal is to enable production in 2016,” he said. “Certification is complete using an ARM core. Our 10nm will achieve industry-leading speed, power, and gate density.”

TSMC will not use EUV lithography for its first 10 nm products, according to Liu, but the company is working with ASML to develop tools for the process technology.

“There is still some way to go to catch 10 nm. We’re looking for an entry point after 10 nm.”

TSMC's financial documents: http://www.tsmc.com/tsmcdotcom/ExtI...rAndQuarter&year=2014&theQuarter=3&language=E

This doesn't look look "very smooth" at all. 20nm HVM started right at the start of January. Back then the promise was to get a quick move from 20nm to 16nm within 1 year. Now 16nm production will start about a full 6 months later, which is a node that doesn't even reduce density. One think that should also be noted is that the first production most likely will be mobile chips. Who knows when we'll see 16nm GPUs, maybe even after Intel's 10nm launched, so those would be 2-3 nodes behind for a short amount of time.

The same goes for 10nm. I think the best way to approximate its most optimistic time to market is 2 years after 16nm. My prediction of earlier this year that we'll see 10nm (remember, this is equivalent to Intel's 14nm, certainly compared to performance) towards the end of 2018, seems to be true.

With those things in mind, one should be careful with interpreting TSMC's statements, because they seem to be quite optimistic.

We should have a little bit more information about Intel's struggles (or who knows, lack thereof) at Moore's Law after their Investor Meeting somewhere next month, for comparison.

 

[AtenRa]

Thread title is misleading, you say no 16nm in 2015 when your own quote from the source says high-single digit revenue by the end of 2015.

By the fourth quarter of 2015, 16 nm FinFET chips will account for a “high-single-digit” percentage of the company’s sales revenue, Wei said.

 

[Nothingness]

So net revenue is up, gross margin is up, operating margin is up, net profit margin is up. What's not to like?

 

[Qwertilot]

High single digit sounds rather like you know who and no one else
(Like this year I guess.).

 

[liahos1]

So net revenue is up, gross margin is up, operating margin is up, net profit margin is up. What's not to like?

actually i think there is a lot not to like going forward

First on Apple:
http://www.zdnet.com/samsung-expects-sales-boost-from-apple-chip-deal-7000034269/

Apple is like 20% of revs in q4. They will still benefit from q1-q3 2015 but if the Samsung LSI pres is being truthful that business goes away in q4 and by 2016 a 10% customer is gone

Second on QCOM, a ~25% customer
http://www.digitimes.com/news/a20141014PD202.html - 28nm at UMC

https://www.qualcomm.com/news/relea...lcomm-collaborate-28nm-wafer-production-china -

http://www.digitimes.com/news/a20140701PD201.html

Looks like they shifting business to Samsung, UMC and SMIC. probably as payback for the price gouging they received at end of 2011 at 28nm

28nm pricing:

34% of their revenue last quarter was 28nm. 80% of it was HKMG and they expect that to shift to 70% HKMG and 30% Poly-Si.
I think alternate foundries are charging 10-15% less per wafer at 28nm for Poly-Si. So you have to believe pricing gets impacted there

Finally Intel a threat as they take more mobile share in tablets (in house mfg) and smartphones (built at tsmc but eventually moved inhouse)

Add the fact that capital intensity is rising and they have the smallest financial scale to pay for it relative to intel and samsung and you have some serious issues.

Also smartphone growth is expected to decelerate from 25% this year to 13% next year and all the units are shifting to low end .

edit-

another corroborating data point that they lost apple is TSM is saying high single digit contribution from 16nm next year by q4. That does not jive with builds assuming apple shifts to 16nm next year.

 

[lefty2]

Thread title is misleading, you say no 16nm in 2015 when your own quote from the source says high-single digit revenue by the end of 2015.

Yes, this is just an obvious anti-TSMC troll post.
FinFET plus begins in 2Q / early 3Q 2015 - which clearly means there is 16nm in 2015. Also, the regular FinFET 16nm production has already started this year with production of the HiSilicon network processor.
Also, TSMC 16 nm FinFET is more dense than 20nm, by 15% according to TSMC.
TSMC said that EUV won't be ready for initial 10nm production, but may be used later on to reduce costs in the node.

 

[Ken g6]

Well, this certainly explains why Maxwell wasn't delayed until the 16nm process was available. I wonder if AMD is delaying their cards until then?

 

[III-V]

Well, this certainly explains why Maxwell wasn't delayed until the 16nm process was available. I wonder if AMD is delaying their cards until then?

They'll get eaten alive, honestly. Maxwell clan ain't nothin to... er... mess with.

 

[witeken]

Thread title is misleading, you say no 16nm in 2015 when your own quote from the source says high-single digit revenue by the end of 2015.

You can do the math yourself:

20nm HVM: Janury 2014
Products (iPhone/iPad): September/October 2014
Delta: 8-9 months

16nm supposed delta: 8-9 months
HVM: ~June 2015
Products: February 2016

Also note that this is only the iPad and iPhone. When will we see the first 20nm Qualcomm, who has a market share of high double digits, SoCs? A year after HVM started.

But because OEM products have to be aligned, the first 16nm product will probably be the Galaxy S6, unless Samsung can do better and Qualcomm moves to them. Weren't they supposed to start HVM in this quarter?

 

[lefty2]

That does not jive with builds assuming apple shifts to 16nm next year.

In my estimation Apple will not switch from TSMC next and neither will it move to 16nm/14nm next year.
Why?
1) Samsung's 14nm starts next year, provided their yield is good. But they said in latest earnings call that the initial production is reserved for internal (i.e. Exnos). Only in 2nd half of year do they start external production. That's too late for the A9, which will be release in September. Secondly, rumours have it that Samsung's 14nm yield is poor. (http://www.eetimes.com/document.asp?doc_id=1324211)
2) TSMC 16nm FinFET plus starts in 2Q/early 3Q. That's also too late for September launch of A9.
Therefore Apple A9 is produced on TSMC 20nm same as A8.

 

[witeken]

So net revenue is up, gross margin is up, operating margin is up, net profit margin is up. What's not to like?

The fact that even a half-node as 16nm now takes 18 months, and a full node more than 3 years (at least for TSMC).

 

[III-V]

You can do the math yourself:

20nm HVM: Janury 2014
Products (iPhone/iPad): September/October 2014
Delta: 8-9 months

16nm supposed delta: 8-9 months
HVM: ~June 2015
Products: February 2016

Also note that this is only the iPad and iPhone. When will we see the first 20nm Qualcomm, who has a market share of high double digits, SoCs? A year after HVM started.

But because OEM products have to be aligned, the first 16nm product will probably be the Galaxy S6, unless Samsung can do better and Qualcomm moves to them. Weren't they supposed to start HVM in this quarter?

Qualcomm was actually the first out the gate on 20nm.

 

[soccerballtux]

ah ha aw yea. could have called that one

 

[liahos1]

In my estimation Apple will not switch from TSMC next and neither will it move to 16nm/14nm next year.
Why?
1) Samsung's 14nm starts next year, provided their yield is good. But they said in latest earnings call that the initial production is reserved for internal (i.e. Exnos). Only in 2nd half of year do they start external production. That's too late for the A9, which will be release in September. Secondly, rumours have it that Samsung's 14nm yield is poor. (http://www.eetimes.com/document.asp?doc_id=1324211)
2) TSMC 16nm FinFET plus starts in 2Q/early 3Q. That's also too late for September launch of A9.
Therefore Apple A9 is produced on TSMC 20nm same as A8.

so the samsung lsi president is lying?

I mean collectively GLOFO/Samsung are standing up 4 fabs with ~114kwpm capacity. Assuming 80% yield on 6 months of production thats enough capacity for ~370mln a9s. Why would they be putting up all that capacity w/o a big anchor? seems dubious.

 

[III-V]

so the samsung lsi president is lying?

I mean collectively GLOFO/Samsung are standing up 4 fabs with ~114kwpm capacity. Assuming 80% yield on 6 months of production thats enough capacity for ~370mln a9s. Why would they be putting up all that capacity w/o a big anchor? seems dubious.

Apple's supposedly switching back, partially.

 

[liahos1]

Apple's supposedly switching back, partially.

i know. Hence why I think TSM loses apple biz.

Look at their guide for expected q4 contribution of 16nm next year
look at the comments from Samsung LSI pres 2 weeks ago
Look at the amount of capacity glofo/samsung are standing up.

 

[witeken]

Yes, this is just an obvious anti-TSMC troll post.
FinFET plus begins in 2Q / early 3Q 2015 - which clearly means there is 16nm in 2015. Also, the regular FinFET 16nm production has already started this year with production of the HiSilicon network processor.
Also, TSMC 16 nm FinFET is more dense than 20nm, by 15% according to TSMC.
TSMC said that EUV won't be ready for initial 10nm production, but may be used later on to reduce costs in the node.

TSMC 16nm is not more dense. TSMC's 16nm Plus is 15% more dense, which BTW is far from the 50% of a regular node or even half-node, but there was no word about that. It may very well be used first in 2017 or H2'16, like HKMG which was also half a year later than 28nm LP.

 

[lefty2]

so the samsung lsi president is lying?

But what exactly did he say?
He said once Apple becomes customer again LSI profits would improve. He did not say Samsung would produce A9.

 

[liahos1]

I'd also add the smartphone situation in China ex Apple doesnt appear great. i think a ton of oems were building inventory ahead of a big 4G handset ramp in the 2H. Then SASAC came and told the carriers to cut subsidies. Doesnt appear to be a great setup.

http://www.taipeitimes.com/News/biz/archives/2014/07/10/2003594732
http://www.digitimes.com/news/a20141001VL200.html
http://www.scmp.com/business/compan...ress-coolpad-cuts-china-mobile-naked-strategy
http://www.mobilemarketingwatch.com/smartphone-shipments-in-decline-across-china-42815/

 

[liahos1]

But what exactly did he say?
He said once Apple becomes customer again LSI profits would improve. He did not say Samsung would produce A9.

then explain all the capacity they are putting up at 14nm. Somebody is going to have to fill all that...

 

[lefty2]

TSMC 16nm is not more dense. TSMC's 16nm Plus is 15% more dense, which BTW is far from the 50% of a regular node or even half-node, but there was no word about that. It may very well be used first in 2017 or H2'16, like HKMG which was also half a year later than 28nm LP.

Yes it is. Both 16nm FinFET and FinFET plus are 15% more dense and 20nm already gave great density improvement. I imagine first products to hit market will be sometime in 2016.

 

[lefty2]

then explain all the capacity they are putting up at 14nm. Somebody is going to have to fill all that...

Yes, eventually somebody will fill it. But I'm betting it won't be ramped up in time for A9 production.

 

[III-V]

Yes it is. Both 16nm FinFET and FinFET plus are 15% more dense and 20nm already gave great density improvement.

This contradicts what TSMC has stated -- which is that only FF+ is 15% denser.

 

[lefty2]

This contradicts what TSMC has stated -- which is that only FF+ is 15% denser.

Please provide link then. I'm pretty sure they never said that.

 

[AtenRa]

You can do the math yourself:

20nm HVM: Janury 2014
Products (iPhone/iPad): September/October 2014
Delta: 8-9 months

16nm supposed delta: 8-9 months
HVM: ~June 2015
Products: February 2016

Also note that this is only the iPad and iPhone. When will we see the first 20nm Qualcomm, who has a market share of high double digits, SoCs? A year after HVM started.

But because OEM products have to be aligned, the first 16nm product will probably be the Galaxy S6, unless Samsung can do better and Qualcomm moves to them. Weren't they supposed to start HVM in this quarter?

Again, Thread title = TSMC Q3'14: No 16nm in 2015

This thread is about TSMCs 16nm production, not 16nm products in retail. If you want to talk about products in retail, tell a admin to fix the thread title.