Taiwan Semiconductor Manufacturing Company - Earnings Call - Q3 2011

Transcript

Speaker 8

Welcome to TSMC's 3-2-11 results webcast conference call. This conference call is being webcast live via the TSMC website at www.tsmc.com and only in audio mode. Your dial-in lines are also in listen-only mode. I would now like to turn the conference over to Dr. Elizabeth Sun, TSMC's Head of Investor Relations.

Speaker 0

Thank you, Andrea. Good morning and good evening, everyone. Welcome to TSMC's third quarter 2011 conference call. Joining us on the call are Dr. Morris Chang, our Chairman and Chief Executive Officer, and Ms. Lora Ho, our Senior Vice President and Chief Financial Officer. The format for today's conference call will be as follows. First, Lora will summarize our operations in the third quarter and give you our guidance for the next quarter. Afterwards, TSMC's Chairman Dr. Chang will provide his general remark on the business outlook and a couple of key messages. We will open the floor to questions. For those participants who do not yet have a copy of the press release, you may download it from TSMC's website at www.tsmc.com. Please also download the summary slides in relation to today's quarterly review presentation.

I would like to remind all listeners that the following discussions may contain forward-looking statements that are subject to significant risks and uncertainty, which could cause actual results to differ materially from those contained in the forward-looking statements. Information as to those factors that could cause actual results to differ materially from TSMC's forward-looking statements may be found in TSMC's annual report on Form 20-F filed with the United States Securities and Exchange Commission on April 15, 2011, and such other documents as TSMC may file with or submit to the SEC from time to time. Except as required by law, we undertake no obligation to update any forward-looking statements, whether as a result of new information, future events, or otherwise. I would like to turn the call over to Lora.

Speaker 5

Thank you, Elizabeth. Good morning and good evening to everyone. Welcome to our 2011 third quarter earnings conference call. I will start with the financial highlights in the third quarter, then we will move on to the outlook for the fourth quarter. All dollars presented are in $NT unless otherwise stated. In the third quarter, our revenue was higher than the guidance provided on July 28, thanks to some rush orders in August and more favorable exchange rates. Third quarter revenue decreased 3.6% sequentially to $106.5 billion, and the wafer shipments decreased 3% to 3.2 million 8-inch equivalent wafers. Our wafer demand was affected by weakened global economic conditions and customer inventory adjustments. Computer and consumer-related revenue was more affected and decreased by 16% and 18% respectively, whereas communication and industrial increased by 3% and 9% from the second quarter.

Overall, communication accounted for 48% of our total wafer revenue in this quarter, while computer, consumer, and industrial accounted for 21%, 10%, and 21% of our wafer sales respectively. By technology, demand for 40-nanometer and below technologies held up relatively well and reached 27% of wafer revenue in the third quarter, including 0.5% from 28-nanometer, and we expect our 28-nanometer ramp to accelerate in the next few quarters. Combined contributions from 65-nanometer and below represented 64% of our total wafer sales in the third quarter, which is 1 percentage point lower than the second quarter as revenue from the 65-nanometer decreased by 2 percentage points in the third quarter. Gross margin was 42% and operating margin was 29.7%. The sequential decline in margins was mainly due to lower utilization. Operating expense increased 1.4% from the previous quarter, mainly due to higher R&D spending on 20-nanometer and higher patent filing fees.

In the future, we will continue to increase R&D spending to expand our lead in technology and keep SG&A tight at the same time. Overall, our third quarter net margin arrived at 28.5%. EPS was $1.17. In line with what we said in the last conference, we have successfully reduced the inventory level during this quarter. Our inventory turnover days significantly reduced by eight days to the company's normal level of 45 days, mainly reflecting lower inventory levels in finished goods and work in process inventories. Cash flow from operations totaled $65 billion. Free cash flow was $17 billion during the quarter, up from an outflow of $1.8 billion in the second quarter due to less capital expenditure. Combined with a $78 billion cash dividend payment and $18 billion proceeds from a corporate bond issuance, our cash and short-term investment decreased $39 billion to $120 billion.

In light of weakened demand, we decided to further revise down 2011 capital expenditure to $7.3 billion. Total capital expenditure in the first three quarters reached $6.3 billion. As a result of the adjustment, our total capacity would be flat in the fourth quarter. Full-year total capacity will increase 17% to 13.2 million 8-inch equivalent wafers, while 12-inch wafer capacity will increase by 29%. Now, let's turn to the outlook for the fourth quarter. Based on current business expectations and a forecast exchange rate of 30.3, we expect our consolidated revenue in the fourth quarter to come in between $103 billion and $105 billion. In terms of margins, we expect our fourth quarter gross margin to be between 43.5% and 45.5%. Operating margin to be between 30% and 32%. The reason for the lower revenue guidance but higher margin guidance is due to the following reasons.

Number one, in the fourth quarter, we have better cost control, so the cost savings contribute to margin improvement. Also, the fourth quarter utilization will be slightly higher than the third quarter. In addition to that, the exchange rate in the fourth quarter is higher than the third quarter, which is favorable to the company. The three favorable will be offset slightly by an unfavorable product mix due to the ramping of 28-nanometer. This concludes my remarks today. Now, I would like to turn the call over to Dr. Morris Chang, our Chairman and CEO for TSMC.

Speaker 1

Hi, everyone. I'm Morris Chang, CEO of TSMC. I want first to talk about the world semiconductor market and foundry market this year. The weakening world economic outlook has impacted the demand for semiconductors. We now forecast this year's world semiconductor growth at 1%. We had forecast 5% in January and 2% in April, and we now forecast only 1% growth for the world semiconductor market this year. We have also trimmed our forecast for semiconductor foundry revenue growth to 4% this year. We had forecast 7% as late as July. All these are in U.S. dollars. Now, TSMC's growth this year in U.S. dollars will be 9%. Now, I want to talk about the supply chain inventory.

In our last quarterly conference, we said that the excess inventory in the semiconductor supply chain will be mostly digested in 3Q and that the BOI will be at the seasonal level by the end of 3Q. We also said that TSMC's utilization rate will be higher in 4Q compared to 3Q. These statements hold true today. The supply chain inventory is estimated to be three days below seasonal at the end of 3Q. TSMC has reduced our own BOI by 8 days to 45 days in 3Q, which is about our normal level. However, the increasing uncertainty of future demand has continued to prompt IC vendors, namely our customers, to de-stock. We believe this effective supply chain BOI will be further reduced to several days below seasonal by the end of 4Q. We do expect our fourth quarter utilization rate to be slightly higher than 3Q.

Now, I'd like to talk about our competitive strengths. In these economically uncertain times, TSMC's competitive position is stronger than ever. Our technology development is progressing well, and I'll talk about a few of them here. First, 28-nanometer. Our 28-nanometer process is in volume production. More than 40,000 12-inch wafers will be shipped to customers in the next few months, and we anticipate the ramp next year will be fast. Three out of the four 28-nanometer process offerings are in volume production. They are 28HP, 28HPL, and 28LP. The remaining version, 28HPM, will be ready for production by the end of this year. Our 28-nanometer process has surpassed the previous generation's production ramps at the same point in time, and product yield continues to improve and is on track.

We expect 28-nanometer to be an important engine of our growth next year since we are the first and only foundry to volume produce 28-nanometer. We not only demonstrate our leadership in technology but also provide arsenal for our customers for their design wins with competitive products. Now, a few words on 20-nanometer. Our 20-nanometer development is on track and demonstrates fully functional SRAMs. Multiple customers have already engaged for more than a year with TSMC on the development of 20-nanometer. We plan two offerings for the 20-nanometer process: 20G for high-performance products such as CPU, GPU, and server, and 20SoC for high-performance and low-leakage products such as mobile computing devices. Risk production for 20-nanometer G is scheduled for 3Q 2012 next year, and risk production for 20-nanometer SoC will start in January of 2013.

We are the world's first and only foundry to complete the tape-out of 20-nanometer ARM Cortex-A15 MP core processor, which is high performance for future mobile computing products. We have also started the development of 14-nanometer FinFET. The device pathfinding has been ongoing for more than one year, and we plan to risk start our 14-nanometer process in 2014. A few words on CoWoS. CoWoS stands for Chip on Wafer on Substrate. Last quarter, I showed you a framework for subsystem integration through silicon interposer. This time, I'm able to show you the picture of the actual test chips. Let's view with one. The logic chip and the DRAM chips, or cubes, are placed on top of the silicon interposer. This is CoW, chip on wafer, and then packaged on a substrate. This is CoWoS, chip on wafer on substrate.

We have achieved very good use and obtained encouraging reliability on these test vehicles. The process flow for CoWoS is shown on the next view graph. The advantages of CoWoS indicate the resolution for warpage and the thin wafer issues. A clear ownership of the long process flow simplifies supply chain and shorter cycle time. The improvement on yield, cost reduction, and reliability can be continuous, efficient, and effective. With this CoWoS technology, our business model is such that we will provide service in, one, top-die logic wafer process, wafer sort, and microbumping, and, two, integrated back-end assembly solution for interposer wafer, final assembly, and final test. We do not plan to sell just the interposer. A few words on embedded flash. Our business in embedded flash has grown by 30% to a level close to $800 million this year, thanks to the strong growth in smartphones and tablets.

Battery management IC, as well as touch sense controllers, enjoy the highest growth within the embedded flash segment. TSMC's 0.18 micron embedded flash with a complete platform solution saw a record level of new product tape-outs this year. Our 90-nanometer embedded flash has started production ramp this year. 65-nanometer joint development programs and 90-nanometer phased-in in MCU have been established, and we expect to launch 40-nanometer embedded flash program in 4Q this year. Now, a few words about next year's semiconductor and TSMC outlook. We expect next year's semiconductor growth, semiconductor market to grow 3% to 5% over this year. TSMC, with 28 high ramp-up constituting our important engine for growth, will grow several points higher than the semiconductor market. All these, of course, in U.S. dollars. Now, a few words on capital expenditure, CapEx. We have two opposite forces in action here.

We have a relatively soft growth outlook in the semiconductor market, and yet, we are going to ramp up our 28-nanometer very fast next year. Now, this year, we have already trimmed our 2011 capital spending to $7.3 billion, as Lora just told you. It is $100 million less than we announced last quarter. For next year, we plan to reduce our 2012 capital spending. It will be at a lower level than this year. Those are all the prepared comments I have.

Speaker 0

This concludes our prepared statements. Operator, please open the floor to questions.

Speaker 8

At this time, we will open the floor for questions. If you would like to ask a question, please press star then one on your touch-tone phone now. Questions will be taken in the order in which they are received. If at any time you would like to remove yourself from the questioning queue, please press pound or the hash key. Please limit your questions to two at a time with one follow-up to allow all participants an opportunity to ask questions to the management members. Your first question comes from the line of Randy Abrams of Credit Suisse.

Speaker 7

Yes, hi, good evening. I wanted to ask a couple of follow-ups on the CoWoS and 3D packaging. The first one is if you think it makes more sense to vertically integrate more and offer more of the traditional back-end assembly and test, even beyond some of the 3D packaging.

Speaker 1

Will we consider even a further integration?

Speaker 7

Yes, further. Doing more of the back end, you're starting to move into wafer-level packaging and 2.5D and 3D packaging. More broadly, do you think back-end assembly and test is an area that you could vertically integrate, even more of a more full suite of steps there?

Speaker 1

At this point, from the business point of view, we're not considering further than what we already have told you. However, we have quite a large 3D group for our research, which are constantly looking at the possibility of further integration.

Speaker 7

Okay, for the CoWoS process, what technology would you see this getting adopted? Maybe if you think of how broad the application would scale across your business.

Speaker 1

I think that it will start in 28-nanometer and that it will be bigger in 20-nanometer.

Speaker 7

To get a sense of how you're planning, coming out of this inventory de-stocking, where 2009, we had a strong rebound, are you planning to build, are you, I guess, for your planning, are you trying to put in place buffer capacity, particularly on 28-nanometer in the event we have a speed-up recovery? Do you want to take a conservative approach given what you see out there?

Speaker 1

At this point, we are taking the conservative approach now, and I really don't expect as strong a surge if there is a surge. I think there will be an increase, there will be a surge, but I don't expect a surge to be as strong as the one we saw in the second and third quarters of 2009.

Speaker 7

Okay. The last question I had was just on, there were comments the press had about one IDM, Panasonic, but a few others have discussed outsourcing more. It's been very gradual over time, but are you starting to see any acceleration in outsourcing from that market?

Speaker 1

Yes. I'm not going to comment on any specific customer, but the answer to your question is yes.

Speaker 0

Your next question comes from the line of Brett Simpson of Arete Research Services LLP.

Speaker 7

I'm curious about your guidance. You're guiding for utilization rates to increase, a little bit more of a tailwind from currency, yet total revenues to decline 1 to 3% or so. If it's me, then that would mean your ASPs have to be falling off quite a bit. Could you help me reconcile that and maybe talk about pricing both in terms of like-for-like and mix?

Speaker 5

Steven, because the fourth quarter, we will ramp 28-nanometer. If you talk about from the company-wide, the average ASP point of view is you will support it. You will go up because of 28-nanometer. Other than that, as I just described, the fourth quarter revenue declined just slightly with the support of FX, which will be 4% appreciation. This is the third quarter. Now, the ASP is actually channeling to about 4% of revenue. Did I answer your question?

Speaker 7

I mean, I guess it still doesn't really add up that much. Without the currency impact, your revenue guidance would have been something lower. How do I reconcile that with rising utilization rates? To me, the one missing component was ASP. I was trying to understand.

Speaker 1

I think that you probably didn't hear my statements on my prediction, actually, three months ago. I actually predicted that the fourth quarter position will be higher than the third quarter, even though the revenue will be flattish or might even be down. The reason I made a prediction, and it has come true, was that we reduced our own inventory in the third quarter. While we're reducing the inventory, naturally, the utilization was held low. Our own inventory reduction has been completed at the end of September. In the fourth quarter, our utilization reflects a normal mode of operation. You see, when you're reducing inventory, you have to make less stuff. Therefore, your utilization is low. After your inventory reduction has been completed, you have to make stuff normally. Therefore, your utilization comes up. Does that make sense?

Speaker 7

Yeah, I understand. It makes a lot of sense. Thank you for restating that again. I'm trying to understand a little bit with this CoWoS process. You said it starts in 28-nanometer. Could you just give us a general idea of the number of customer engagements you have or designs you're working with, or some way that we can try to quantify the potential impact next year and beyond, I guess?

Speaker 1

I think the customers, we are actually not working with a large number. We are working with several customers on the development phase. The impact to us, revenue impact to us, will be minimal, even next year. 28 is going to run for quite a long time. I do think that the CoWoS will start with 28. I think that we are already working with a few customers also on the 20-nanometer co-founders. By the time 20-nanometer is the volume production, I think the impact on our revenue will be reasonably significant.

Speaker 7

Understood.

Speaker 1

I can't quantify it beyond that. Yeah.

Speaker 7

If I could just speak in one final one, Lora. Today you mentioned that the depreciation would grow about 23% this year, but next year would be something lower. Given that we're all a little concerned about, you know, margins impacts from higher depreciation, can you quantify a bit more maybe on what you think depreciation growth would be besides just lower than 23% next year?

Speaker 5

We haven't finished, we haven't made a decision on the CapEx for 2012, depending on how much 28-nanometer we want to build. With the last CapEx for next year, we believe the depreciation year-over-year increase will be less than 20%.

Speaker 1

She said less than 20%. Let me say this. Of course, the ideal situation financially is for the increase of depreciation to be below the increase of revenue. You have to keep in mind that depreciation is a counter in NT, even though the expenditure is largely in U.S. dollars. Once we get the equipment, once we install the equipment, the depreciation base is in NT, so the depreciation cost is in NT. On that basis, this year, our revenue increase in NT is not very great. What is it? It's almost flat, isn't it? It's almost flat. Yet, we had a depreciation increase of 23%. That is certainly far from ideal. Next year, I said earlier that I expect our U.S. dollar revenue growth will be several points more than 3 to 5%.

We certainly would like to think that the NT revenue growth will be comparable, and that's assuming the exchange rate doesn't vary very much. If the NT revenue growth is in the, oh, let's say, whatever, 6 to 8% range, and the depreciation increase is less than 20%, then that's a heck of a lot better situation than we had this year. Does that make sense?

Speaker 0

Your next question comes from the line of maybe.

Speaker 7

I'm sorry, I was hoping that the gentleman.

Speaker 0

Andrea, I think you cut off Steven's line. Chairman is asking if Steven is satisfied with that answer.

Speaker 7

If you can still hear me, yes, I'm satisfied. Thank you. Yes, I'm satisfied. Thank you for taking the time. I really appreciate it.

Speaker 1

Thank you.

Speaker 0

Your next question comes from the line of Mehdi Hosseini of Susquehanna Financial Group.

Speaker 6

Thanks for taking my question. Dr. Chang, I want to ask you about the trend on the technology side. Your up-and-coming competitors are obviously moving towards 32-nanometer and planning to have a product line for 28-nanometer sometime next year, but they're still on the gate-first where TSMC is gate-last. Is this something that you think people still haven't really understood the differences? Is this something that could really set TSMC apart from these new competitors? Anything else that you can offer would be appreciated. I have a follow-up question.

Speaker 1

The most important people to understand are the customers. I believe that our customers unanimously understand that the gate-last approach is superior.

Speaker 6

When you say customers, obviously designers prefer gate-last. When it comes to cost, gate-first offers some advantages if you can resolve the yield issues. When you say customers, does that refer to the decision-makers or the people who are actually involved with product design?

Speaker 1

It is certainly not obvious to me at all that gate-first offers even cost advantage. I believe that superficially, it seems to be a simpler process, yet there is a big difference between a simpler process and lower cost because the simpler process may be harder to make and the yields may be lower, and therefore, the cost will even be higher. That's my feeling about it.

Speaker 6

When do you think there will be?

Speaker 1

Again, in my feeling, I think it's less important than our customers now. I think our customers unanimously know that gate last. Actually, you know, for that matter, our competitors have decided to change to gate last also.

Speaker 6

Not until 20-nanometer. I'm just wondering if from now till then, there would be enough difference in performance, cost-benefit of the product that would prevent these new competitors from gaining any more traction.

Speaker 1

I don't think so. I don't think so. Our technical people, whom I respect, have unanimously told me that gate-last is the way to go, and it will be competitive.

Speaker 6

My follow-up question, I kind of missed your commentary on inventory correction. I apologize if I'm repeating the question I already asked. When you look at the rolling forecast, given what we know today of the market demand, and when you look into your rolling forecast for beyond December, how do you see the trend playing out, especially when Christmas and Chinese New Year are back to back? Many chip companies have already told us over the past two weeks that chip inventory correction is going to come to an end by Q3. What I'm trying to figure out is, is there a realistic scenario to expect a pent-up demand if the sell-through during the month of January doesn't fall off the cliff? Is this something that you're contemplating?

Speaker 1

I'm sorry. I'm asking Elizabeth to.

Speaker 0

I think your question is some people are saying that the inventory correction will be ending by the end of Q3 instead of by the end of Q4, right?

Speaker 1

Yes, by the end.

Speaker 6

When you look into your rolling forecast for beyond December, is there anything that you can share with us? I'm asking that because inventory correction is coming to an end by Q3. We have two major catalysts for chip consumption: Christmas and then two weeks later, Chinese New Year. If sell-through doesn't fall off a cliff, there could be a scenario where there's a pent-up demand and some of your customers may rush back.

Speaker 1

I think that, you know, actually, last quarter, three months ago, we thought that the supply chain inventory correction would probably end in September. The inputs that we have been getting from our customers, or I should say the implications that we have been getting from our customers, is that the stocking is still going on. Keep in mind that we are not like an IBM such as Intel. We do have our customers as another layer between us and the market and the equipment market. The IBMs can control their inventory very well, and they are in immediate contact with the equipment market. We are one layer more, one layer apart, one layer more apart from the equipment market. Therefore, the way we see this restocking is a bit more complicated than the IBMs with the restocking. I don't know whether I made myself clear.

Speaker 6

Should I assume we just don't know how the Q1 rolling forecasts are going to look like? Is that what you're implying?

Speaker 1

Q1 is a seasonally weak quarter for us. It has always been anyway. Now, with the continuous restocking, if you want me to say now what I expect, I expect that Q1 will be seasonally weak in revenue, but I expect that the season Q1 may be quite good in orders.

Speaker 6

Got it. Okay, now it's clear.

Speaker 1

Thank you. Thank you.

Speaker 0

Your next question comes from the line of Mike McConnell of Pacific Securities.

Speaker 7

Thank you. I just want to follow up on Mehdi's question. I wanted to see, with the capacity flat in Q4, and you said utilizations are going to be up. Does that mean starts will be up sequentially in Q4 to get the utilizations higher? Or is it just you're trying to rebuild back to a normalized level for inventory and we shouldn't extrapolate starts being up from your customers, which will lead to potential growth in Q1? Just wanted to clarify that.

Speaker 1

Oh, I don't think that Q1 will be a potential growth quarter. As I said, Q1 is seasonally weak anyway for us. I don't think so.

Speaker 7

Just the last question on 28-nanometer and how we should look about the ramp next year. You've talked a lot about tape-out activity being exceptionally strong for this node relative to some past leading-edge geometries. Typically, in the first year, generally, if you're leading-edge geometries, you see revenue as an exit rate in that first year, high single digits. Do you think that this time, when you're looking at 28-nanometer and the amount of tape-out activity you have, that you could be in a double-digit range as an exit rate in terms of % of revenue next year?

Speaker 1

We expect 28-nanometer to be ramping up continuously from this point on all through next year. Remember, 28-nanometer still is a relatively small percentage of wafer size.

Speaker 7

For the revenue, though, do you think it could be about 10% as an exit rate for next year?

Speaker 5

We'll have about 2% in the fourth quarter from 28-nanometer, and we'll continue to go up.

Speaker 1

2% this fourth quarter.

Speaker 5

This fourth quarter, 2%. By the second half next year, it will come for more than 10%.

Speaker 7

Okay. Great. Thank you.

Speaker 0

Your next question comes from the line of Donald Liu of Goldman Sachs Group.

Speaker 2

Hi. Good evening. My first question is, we have been talking about 28-nanometer has a much higher CapEx intensity, much more expensive. I would assume the wafer price would be a lot higher too. When should we, will we see, if any, positive impact on TSMC's ASP on a systematic way because of the higher 28-nanometer and 40-nanometer in mix?

Speaker 1

Yes.

Speaker 5

Donald.

Speaker 1

Yes, Donald. Certainly, it will have a positive impact on our blended average price because, I mean, it's a higher price, 28-nanometer. How much impact? Just as Lora said, the revenue impact will be 10% or more, more than 10% the second half of next year. You know, you just figure out arithmetically, you can see that while it has an uplifting effect, our base is pretty big, Donald. The revenue impact is over 10% by the second half of next year. In the first half of next year, it will be smaller than that. It can't have too big an impact.

Speaker 2

Is that ASP change already included in your revenue assumptions for next year?

Speaker 1

Yes, of course.

Speaker 2

When you talk about your revenue will grow a few percentage points higher than the industry?

Speaker 1

Pardon me?

Speaker 2

Sorry. When you earlier said TSMC's revenue growth next year might be a few percentage points higher than the industry, I assume that's already including this potential ASP effect?

Speaker 1

Yes, that's correct. Actually, to tell the truth, Donald, you know I'm more confident of the growth in 28-nanometer than I am about the growth of everything else. We have included the 28-nanometer numbers in my overall estimate. I think that I have taken a reasonably conservative approach on the rest of the portfolio, which is, by the way, 80, 90%, no, 90%, more than 90% of the total. I think I've taken a reasonably conservative approach when I say that the total growth will be several points above 3 to 5%.

Speaker 2

Got it. My second question is for Lora. I think in the afternoon, you had a very clear explanation on the margin change in terms of how much is from FX and other things. Can you please just repeat that so we can be sure we get the correct answer?

Speaker 5

Donald, I think you're talking about this third quarter versus our fourth quarter guidance. Is that right?

Speaker 2

Correct.

Speaker 5

The margin difference. Okay. If I take the midpoint of our fourth quarter guidance, and the margin I guided will be 44.5%. Compared with third quarter, 42%, it is 2.5% higher. On this 2.5% higher, we have several factors. Number one, the better cost in fourth quarter. This is 1.2%. Better utilization, this is 0.7%. Better foreign exchange rate, 1.7%. The less favorable part of our mix because we're ramping 28, this is a negative 1%. If you add those four things up, you get roughly a 2.5% point.

Speaker 2

Right. Your US dollar assumption for Q4 is?

Speaker 5

30.3.

Speaker 2

30.3. Great. Thank you very much.

Speaker 5

You're welcome.

Speaker 0

Your next question comes from the line of Dan Hehler of BofA Securities.

Speaker 4

Thank you for taking my question. Two quick questions, one for Lora and one for Dr. Chang. Lora, I wanted to double-check my numbers on the fourth quarter revenue. With your revenue guidance to be down about 1% to 3% sequentially in 4Q and the NT dollar currency forecast to appreciate about 4%. That would mean your U.S. dollar base decline is about 4% if we assume the midpoint of your guidance. Would the shipments be down about that much and thereby helping inventory digestion? Is that how we should think about this?

Speaker 1

Actually, shipments are down a little less than that.

Speaker 4

A little less. Okay. Thank you. You got a little bit of ASP decline a bit, not much. Okay. I guess, Dr. Chang, thanks on the follow-up on today's my question on FinFET because you know, you did mention that TSMC contemplated now deeply when to deploy FinFET, whether it would be '14 or, you know, earlier. I'm wondering why TSMC wouldn't deploy this basically just as quickly as possible, particularly since you guys have done so much work over the years in this process. The benefits to, you know, the power seem to be quite, quite favorable. I'm wondering what the thought process is. Could this be something that you would, you know, try to deploy sooner? If so, why not?

Speaker 1

We have been working on FinFET for quite a few years. I have to say that in FinFET, we are somewhat behind Intel. However, we could try to catch up by accelerating FinFET and putting it on 20. The alternative is what we chose. The alternative is to find a way that our planar 20 can compete with Intel's FinFET 22, and delay the use of FinFET to the 14-nanometer generation. We will be more ready with FinFET when the 14 generation comes. We chose that alternative. We believe our 20-nanometer planar in its ecosystem is competitive with Intel's 22 FinFET in Intel's ecosystem. We believe that.

Speaker 4

That's clear. Thanks so much. Because this has been being also the design core as well, including ARM versus the ARM as well, I presume.

Speaker 1

Thank you.

Speaker 4

Thank you very much.

Speaker 0

Your next question comes from the line of Zhou Ying of BNP.

Speaker 3

Hi. Good evening. Actually, the question is about R&D. R&D has gone up quite fast in the last couple of quarters. I just want to know if it's a trend going forward or should we expect it to level off at some point in time. Yeah.

Speaker 1

That has been a deliberate policy to increase the R&D both in amount and in percentage of revenue. My present thinking is to let R&D go up to about 9% of revenue, which is where it is almost at now. Of course, as revenue grows, the amount will still grow. My present thinking is to stabilize the R&D expenses at 9% of revenue.

Speaker 3

Okay. Our question is also on the back-end assembly space. When do we expect it to drive a bigger part of your business, say, like the 5% of revenue, for example? Yeah.

Speaker 1

I try to answer that question. I really can't quantify it. I expect that CoWoS will have a significant impact on revenue only two, three years away. In the next two or three years, I do not believe, I do not think that CoWoS will have a significant impact on our revenue.

Speaker 0

Operator, in the interest of time, we will only accommodate one more caller's question. You have a question from the line of Robert Lee of Jefferies.

Speaker 6

Thanks for taking my question, and it's been a very long day for you. I just wanted to ask, just looking at the costs, Lora referred to the fact that costs down in Q4 will help the margin. I just wondered if you can elaborate on that, whether it's anything particular in terms of the de-bottlenecking, or is it general reduction in operating costs or manufacturing costs? Also, if we do a quick analysis of the non-depreciation cost per wafer in the third quarter, it did rise by about 6% or 7% Q over Q. I'm just wondering specifically what drove that increase. Thanks very much.

Speaker 5

Okay. The first part of the question, talking about the cost improvement, I was talking, I was referring to the manufacturing cost when I said cost improvement. There are two things that favor our for.

Speaker 6

Manufacturing costs improve.

Speaker 5

Improve.

Speaker 6

Yeah, no, you can go ahead.

Speaker 5

Okay. The cost improvement actually is what I'm talking about is mainly the material wafer cost. That has some improvement. Also, in the fourth quarter, utility cost will be less in the third quarter compared to, will be less than in utility cost. That's the two major items for the cost improvement.

Speaker 6

Got it. What exactly did drive the small increase in costs or non-depreciation costs that we saw in the third quarter within cost of goods sold?

Speaker 0

Okay, Robert. I think in the third quarter compared to the second quarter, we actually have higher utility costs because it's summer season. Secondly, because our utilization rate in the third quarter is lower than the second quarter, those non-depreciation fixed costs on a per-wafer basis would be higher. That's why the manufacturing, the so-called non-depreciation manufacturing cost was higher in the third quarter compared to the second quarter.

Speaker 6

Got it. Thanks very much. A very quick follow-up question. If you draw the parallels with the challenging experience the industry went through in 2008, 2009, what parallels are there at the moment in terms of what your customers are saying sentiment-wise? I mean, just your own gut feel. I guess you're looking for the industry to essentially recover next year. To what extent, how far are we into the bottoming process for the industry based on where we are today? When does that inflection point come? I know this has been referred to earlier on in the call, but essentially, are we looking for the business and the industry to rebound into the second quarter?

Speaker 1

I think it's problematic to try to predict the future on the basis of what happened in the past. You know, in the 2008, 2009 situation, the depth of the plunge was much greater. The plunge was just two quarters: fourth quarter of 2008 and first quarter of 2009. In the second quarter of 2009, there was a surge. A surge began and the surge was very strong. It continued through the third quarter and then sort of stabilized a little bit from the fourth quarter on. It continued to grow, actually, but the slope of the surge obviously became gentler from the fourth quarter of 2009 on. That was the experience.

Now, if you try to draw an exact analog here, I would think that in the first quarter of next year, we will start to get orders, strong orders, and revenue will probably begin to surge in the second quarter of next year. Again, I said earlier that it's not problematic to draw conclusions, to draw predictions about the future from experience in the past.

Speaker 6

I guess the overall read is for both yourselves and many other companies in the sector, although things are maybe getting incrementally worse. It's nowhere near a comparison with the 2008, 2009 situation. In some respects, things are holding up remarkably well despite all the frightening headlines in the newspaper along CNN.

Speaker 1

I think the frightening headlines in CNN newspapers, you're talking about American newspapers, I assume. I think they are about the general economy. You know, semiconductor is a little different, and semiconductor is a little different because I think we do have a killer app, just the merging. That's the mobile part of it.

Speaker 8

That's our new killer app, and I think TSMC is even a little more different than the semiconductor, so yeah. The threatening news, I think, is probably true. The world economy, I think, doesn't look very good.

Speaker 0

Okay. Well.

Speaker 5

Okay.

Speaker 0

That at least seems your business is holding up well. Thank you very much.

Speaker 8

Thank you.

Speaker 5

Thank you, Robert. I think this concludes our Q&A session, and thank you for joining us today. We hope you will join us again next quarter. Bye-bye.

Speaker 8

Thank you. Good night.

Speaker 1

Before we conclude TSMC's 3Q11 results webcast conference call today, please be advised that the replay of this conference call will only be accessible through TSMC's website at www.tsmc.com. Thank you all.