Elizabeth Sun - Director of Corporate Communication Division Lora Ho - Chief Financial Officer and Senior Vice President of Finance Morris Chang - Chairman and Chief Executive Officer

Randy Abrams - Crédit Suisse AG, Research Division Michael Chou - Deutsche Bank AG, Research Division J.J. Park - JP Morgan Chase & Co, Research Division Steven C. Pelayo - HSBC, Research Division Mehdi Hosseini - Susquehanna Financial Group, LLLP, Research Division William Lu - Morgan Stanley, Research Division Donald Lu - Goldman Sachs Group Inc., Research Division Andrew Lu - Barclays Capital, Research Division Brett Simpson - Arete Research Services LLP Roland Shu - Citigroup Inc, Research Division Eric Chen - Daiwa Securities Capital Markets Co. Ltd., Research Division

[Chinese] Welcome to TSMC Second Quarter 2013 Earnings Conference and conference call. This is Elizabeth Sun, TSMC's Director of Corporate Communications and your host for today. The event is webcast live via TSMC's website at www.tsmc.com. If you are joining us through the conference call, your dialing lines are in listen-only mode. As this conference is being viewed by investors around the world, we will conduct this event in English only. The format for today's event will be as follows: First, TSMC's Senior Vice President and Chief Financial Officer, Ms. Lora Ho, will summarize our operations in the second quarter, followed by our guidance for the current quarter. Afterwards, TSMC's Chairman and Chief Executive Officer, Dr. Morris Chang will provide his key messages. Then, we will open the floor to questions. For those participants on the call, if you 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 earnings conference presentation. Before we begin, I would like to remind everybody that today's discussion may contain forward-looking statements, and they are subject to significant risks and uncertainties, which could cause actual results to differ materially from those contained in the forward-looking statements. Please refer to the Safe Harbor notice that appears on our press release. And now I would like to turn the podium to Ms. -- to TSMC's CFO, Ms. Lora Ho.

Thank you, Elizabeth. Good afternoon, everyone. Thank you for joining us today. I will start my presentation with the financial highlights for the second quarter, and I will follow that by providing the guidance for the third quarter. I'm very pleased to announce another record quarter for TSMC in both revenue and net income in the second quarter this year. The record results once again demonstrate that our investment in R&D and advanced capacities has started to pay off. Looking at the numbers, our revenue increased 17.4% to TWD 156 billion mainly due to strong demand for 28-nanometer technology, while customers need in building their inventory for the launch of mobile product in the second half also helped. On the margin side, second quarter gross margin was 49%, up 3.2 percentage points from the first quarter, thanks to high utilization rate, cost improvements and more favorable exchange rate. Total operating expenses increased 14.8% to TWD 18.8 billion, mainly due to higher R&D and operating expense. Despite that, second quarter operating margin expanded 3.5 percentage points to 37% on larger revenue base and higher gross margin. Non-operating item was a gain of TWD 2.4 billion in the second quarter, up from TWD 1.3 billion in the previous quarter. The increase is mainly due to the legal settlement income, dividend from invested companies and one-time gain from the deconsolidation revaluation of assets. None of these are recurring, but together contributed about TWD 0.04 of EPS to the second quarter. Overall, we made TWD 2 EPS in the second quarter and the single quarter ROE was 24.7%. Let's have a look at the revenue by application. Mobile computing devices continue to drive our business growth in the second quarter. The revenue contribution from the communication segment further increased from 55% in the first quarter to 57% of total wafer revenue in the second quarter. On a quarter-over-quarter basis, revenue increased across the board of all major segments. Communication again show the strongest growth of 22%, followed by: computers, 18%; industrials, 11%; and the consumer, 9%. In terms of revenue by technology, the 28-nanometer continue to ramp successfully. The revenue contribution has significantly increased to 29% in the second quarter from the 24% in the first quarter. Riding on the success of 28-nanometer, revenue from advanced technologies, defined as 40-nanometer and below, now already represents 50% of our total wafer revenue. Moving to the balance sheet. Our cash and marketable securities increased TWD 38 billion to TWD 228 billion at the end of the second quarter, mainly due to proceeds from issuance of U.S. dollar corporate bonds. Correspondingly, our long-term interest-bearing debt increased TWD 43 billion to TWD 171 billion. Current liability increased by TWD 85 billion mainly because we accrued the cash dividend of TWD 78 billion which will be paid by the end of July. Looking at the financial ratios, our accounts receivable turnover days remained flat at 43 days. Days of inventory decreased by 4 days to 47 days, mainly due to shipment out of finished goods and a lower working process inventory days. On the cash flow side, in the second quarter, we generated TWD 75 billion from operations, invested TWD 78 billion in capital expenditure and have raised TWD 45 billion through corporate bonds. In U.S. dollar terms, we spent $2.6 billion in capital expenditures in the second quarter, together with the $2.7 billion spend in the first quarter, we have already spent about 55% of our guided full year CapEx, which is USD 9.5 billion to USD 10 billion dollars. Overall, our cash balance increased TWD 40 billion to TWD 226 billion. Free cash flow ended slightly negative at 3 billion. Lastly, I would like to talk about our capacity plan. Our total capacity increased 3% to around 4 million 8-inch equivalent wafers in the second quarter and a wafer increased by another 6.5% to 4.3 million wafers in the third quarter. For the full year, our 12-inch capacity is expected to increase 17% year-over-year, and our total annual capacity will increase 11% to 16.4 million 8-inch equivalent wafers. I have finished my financial summary. Before we talk about third quarter outlook, I would like to brief you one changes that impact our consolidated revenue. We used to consolidate Syntek, one of our invested company because TSMC support representation in Syntek was greater than half. Recently, Syntek expanded its core structure by adding 2 independent directors in June. As a result, TSMC's board representation is now less than half, therefore, we will no longer consolidate Syntek starting June 30 this year. The consolidation impact would be close to 1% reduction of our total revenue in the third quarter. Taking this into consideration, our third quarter guidance will be as follows. Based on current business expectation and a forecast exchange rate of TWD 29.83, we expect our revenue to be between TWD 161 billion and TWD 164 billion. In terms of margins, we expect the third quarter gross margin to be between 47% and 49% and operating margin to be between 35% and 37%. This concludes my remarks. Let me turn the podium to our Chairman and CEO, Dr. Morris Chang.

Good afternoon, ladies and gentlemen. On the screen, you'll see an outline of what I'm going to say in my message today. First, second quarter and third quarter. Second quarter was a record quarter for TSMC, both in revenue and in earnings-per-share. The strength was mainly driven by the strong growth in margin-related -- mobile-related applications and TSMC's strong position in the 28-nanometer technology. We are rather pleased about the results of the second quarter. Third quarter guidance has already been given to -- has already been given by our CFO, and there will be growth in the third quarter, and we are rather pleased about the expected performance of our third quarter as well. Next, let me talk about the industry outlook, the forecast and supply chain inventory. For full year 2013, we are forecasting a global GDP of 2.6%, which is unchanged from our forecast 3 months ago. For semiconductors, however, we are lowering our earlier forecast of 4% growth to 3% in 2013 for semiconductor growth. For fabless, we continue to forecast 9% growth. This is unchanged from our last forecast. For foundry, we are forecasting now 11% growth. This is the foundry industry. We are forecasting 11% growth, which is up from 10%. For TSMC, we again forecast a revenue growth, which is much higher than the foundry industry growth, and that's unchanged. Now let's look at the supply-chain inventory. Two things have happened in the last 3 months or 2 things have actually become more obvious in the last 3 months. First, the IC vendors inventory preparation for product launch by several major handset makers, the IC vendors preparing for product launch by major -- by several major handset makers, has caused the supply inventory days, inventory days to increase. Second, the lower than expected sales of PCs and several smartphone models have again caused the supply-chain inventory to become higher. Now in April, we have forecast the fabless supply-chain inventory, fabless supply-chain inventory to be 73 days -- I'm sorry, to be 70 days, 70 days. In April, we have forecasted fabless supply-chain inventory to be 70 days at the end of the second quarter. Now we are forecasting -- saying, we are saying that it will be -- or it was 73 days at the end of the second quarter. Now 3 months ago, we're forecasting the supply chain, the fabless supply-chain inventory to be 68 days at the end of the third quarter, and now we are forecasting it to be 71 days at the end of third quarter. About the fourth quarter, 3 months ago, we were forecasting the inventory again for the fabless to reach 66, and now we are still forecasting 66. So our inventory days forecast for the Fabless supply-chain has increased for the second and third quarter, but remains unchanged in the fourth quarter. This is an early indication that the fourth quarter may be a down quarter because we expect the supply-chain to take serious action to manage their inventory in the second half. And the overall inventory, however, will approach the seasonal level by fourth quarter. I'm talking about the 66 days that we are forecasting. That's very close to the seasonal level. Now a few words on our structural profitability. Since '09 -- I picked '09 because that was the year when I resumed my CEO responsibilities. Since '09, our structural profitability at constant exchange rate has improved by 7.7 points -- 770 basis points. On the other hand, the '09 exchange rate was 32.87 and the year-to-date 2013 exchange rate is 29.66, a change of 9.8%, almost 10% and we have said several times that each point of change, being exchange rate, causes 0.4% change in our margin. So the almost 10% change in exchange rate has caused a 4% change unfavorable in margin. And therefore, at prevailing exchange rates, our structure profitability has improved by only 3.8%. Still, I think it's very encouraging, and there are signs that the exchange rate may stabilize or perhaps the NT may even depreciate a little bit, and if that happens, it will of course help our progress in structural profitability even more. Next, I want to talk about -- anyway, the message in the last section is that, in the last 4 years, our structural profitability has improved by 3.8 points at prevailing exchange rates. Next, I want to talk about high-end, mid-end and low-end mobile product world and TSMC's position. Mobile products have been important in driving the demand in recent years. And we'll continue to enjoy robust growth in this year, as well as in coming years. High-end year-on-year growth this year was 18% -- or is 18%, we estimate, from 361 million last year to 428 million this year. Mid-end grows from 167 million units to 227 million units, a 36% growth. Low-end rose from 202 million to 341 million, 69% growth. So this year, we will see high-end units to grow 18%; mid-end, 36%; low-end, 69%, for a total smartphone year-to-year growth of 36%. We are uniquely -- as a foundry, we are uniquely positioned. Because of our comprehensive technology portfolio, we have solutions to address each tier, high, low, middle, each tier of the smartphone market, and of course, would benefit from the overall strength of the smartphone demand. We, in particular, I want to point out, I think everyone knows that our 28-nanometer technology has allowed us to be very well-positioned in the high-end, and probably, mid-end. But I want to particularly point out that we are very well-positioned in the low-end. Because of the comprehensive technology portfolio, we can -- through silicon area shrink and layer reductions, help our customers to streamline features and to integrate functionalities for their overall lower-cost designs. China, of course, is a fast-growing area for mid- to low-end smartphones. We are seeing our Chinese customers taking a more important role in providing chip solutions to the market. Many of them have accelerated their cadence in adapting advanced technology. As a result, our business from Chinese customers has doubled in the first half this year from 1 year ago, and for the whole year, our Chinese -- China region is expected to account for 6% of our total business this year. Next, a few words on 28-nanometers. Our 28-nanometers is on track to triple in wafer sales this year and our 28-nanometer high-K metal gate is ramping fast, and will exceed the Oxynitride solution starting this quarter. For the Oxynitride solution in which we do have competitors, we believe that we have a substantial lead in yield. For the high-K metal gate solution, we do not have any serious competitors yet. We believe we have a substantial lead in performance. If you recall, ours is a gate-last version and our competitors are mainly in the gate-first version. So in high-K metal gate, we have a lead in power. Power is lower. In performance, our performance is better and should the competitors who are in gate-first now switch to gate-last, they will be considerably behind us in yield learning in yields. Our continued lead in yield and performance will keep our 28-nanometer market segment share strong for both this year and in future years. Now a few comments on 20, 16, and I want to say a few words on our grand alliance, and also I want to show you a few photographs of our new GIGAFAB, which is dedicated to 20-nanometer and 16-nanometer. 20-nanometer SoC and 16-nanometer FinFET are both progressing well. On 20-nanometers, we see little competition. The risk production has started in the first quarter, and volume production will start in early 2014, next year and I'll see you -- I will show you the photograph of our brand-new fab, and the equipment are already being installed. The equipment are streaming and are being installed, are going to be tried out and the volume production will start in early 2014. On 16 FinFET, it will start volume production about 1 year after 20 SoC, in other words, early in 2015. Our R&D progress on 16 FinFET is very good. New improvement is better than planned, and is better than 20-nanometer a year ago, and we have been working on -- we have been working with several major customers and many tape-outs. Many product tape-outs are planned for next year, but the joint work is essential for before tape-outs, of course, and that joint work has been progressing for some time. Grand alliance, by that, we mean our alliance with customers, with the design -- electronic design EDA companies, and companies such as ARM and imagination technology and companies like Cadence, Mentor Graphics, as well as our own platform, the Open Innovation Platform. So it's an alliance with the customers, with the EDA companies, with the IP providers, and of course, with the -- our key vendors, critical vendors. Now the reason I want to point that out is that for TSMC, we have entered a new era of competition. We pointed that out almost every time we get together in this meeting, and we've been pointing out for the last 1 or 2 years now. Now in this new era of competition, the competition is not foundry to foundry. It is not foundry to IDM. It is grand alliance to IDM. Have I made it clear? That's the reason I'm pointing it out. Now so we feel on the 16 -- on the 20, as I've already said, we see little competition. On the 16, if we put it on a foundry to foundry or foundry to IBM basis, we are competitive. If you put it on a grand alliance to IBM basis, we are more than competitive. Now I want to show you a few photographs of our new GIGAFAB, which as I said, is dedicated to 20, 16 -- 20 and 16. Now is there no pointer? There's no pointer? Well, that's too bad. But I don't think there'd be any problem with this photograph or with the next 2, because you see, the big building almost in the middle of the photograph, and that's our P5, phase 5, of Fab 14. It's located at -- in Thailand, and it will start -- it's the first one, it's just 1/3 of the GIGAFAB, 1/3. I will show you the other 2/3 in the next 2 photographs. And this 1/3 of this 16, 20, 16-nanometer, 20-nanometer GIGAFAB, will start production, I said earlier, in early 2014, and this picture says February 2014. And this particular phase 5 has a total floor space of 184 square meters. Next, please. Why, if you -- can you switch it back, please? This is finished, the building is finished, as I said. The equipment are streaming in, being installed, being tested out to get ready for the volume ramp-up in early 2014. Next photograph is the second 1/3 of our 20-nanometer, 16-nanometer GIGAFAB. And this one is only half-finished because volume production will start only in May of next year. And it's about the same size, the total floor space is 180,000 square meters, about the same as the last one. Next, please. This is the last 1/3 of this new GIGAFAB. And you can see that you don't see very much yet, but you will see what's going on, it's very busy. I think the foundation and so on has been laid, et cetera, et cetera. It's -- it will start production, volume production, in April of 2015. And its total floor space, it's about the same as each of the other 2. Altogether, these 3 phases of this mega fab, the GIGAFAB, is about 550,000 square meters. Just in comparison, our 28-nanometer facility, which is in Taichung, is about 400,000 square meters. So this one is 550,000 square meters, it's dedicated for 20-nanometer and 16-nanometer, and the 28-nanometer is about 400,000 square meters. And I -- all right, let's show the next photograph. Without a pointer, this is going to be more difficult. But this is our Tainan campus or complex. And well, the -- in -- on the -- in the upper part, you will see the older fabs. Well, old, well, most of them are less than 10 years old, but they are old compared to the new phases I just pointed out to you. Then right below them, you see the new fabs. And there's one on the side, is our back end. That's where our 3D IC is going to be. That's where our CoWoS is also now. And now actually -- yes, thank you very much. I think this will help. Well, I'm not going to point out the old fabs, but this is phase 5. That's the first of the new phases, phase 5. And this is phase 6, and this is phase 7. Now I do want to point out one very, very important aspect of this super fab. Even though you see 3 separate buildings, but they will all be connected, they will all -- all of these 3 will be connected by clean room links, clean room links. All right, so much for the photographs. A few words on EUV. Last time, I think I pointed out that we had -- we just had a breakthrough last time on EUV. Or I should say ASML had a breakthrough on EUV. Now progress has continued. Saima [ph] and ASML have demonstrated an EUV power output of 40 watts on their factory floor. The next step for ASML is to achieve 80 watts of source power. While there's still some distance to reach the source power of 250 watts, 40 watts already achieved. Next milestone, 80 watts. And then the economically desirable threshold is 250 watts. 250 watts will give us a throughput of 125 wafers per hour, which is the economic threshold. While there's still some distance to get there, we are collaborating with ASML very closely to ensure that EUV can become production ready by 10-nanometer, at least for the critical layers. Next, specialty technologies. Specialty technologies, when we say specialty technologies, we mean we include high-voltage or power management ICs, mixed signal for audio codec, embedded flash for MCU, MEMS for motion sensors, image -- I'm sorry, CMOS image sensors for digital cameras. Now in addition to those that we have been working on for a long, long time, in addition to those, the following specialty technologies are expected to grow significantly next year: Image signal processors stacked with CIS, that is processors stacked with the sensors; mixed signal for fingerprinting sensors; high voltage for small panel drivers, small panel drivers; and better flash for Near Field Communication, smart cards and touchscreen controllers. What differentiates TSMC in the specialty technology foundry arena, besides the fact that we have in -- we have the largest capacity in the broadest scope, is our superior ability to integrate specialty devices, flash, CIS, RF, high voltage, power MOSFET, integrate those specialty devices into our strong CMOS baseline while maintaining our CMOS IP compatibility. Those are all the comments I've prepared. I believe we are now open for Q&A.

Right. Yes, this concludes our prepared statements. Before we begin the Q&A session, I would like to remind everybody to limit your questions to 2 at a time to allow all participants an opportunity to ask questions. Questions will be taken both from the floor as well as from the call. Should you wish to raise your question in Chinese, I will translate it into English before our CEO or CFO answers your question. [Operator Instructions] Now let's begin the Q&A session. First question comes from the floor, and that would be Crédit Suisse, Randy Abrams. Randy Abrams - Crédit Suisse AG, Research Division: The first question, on the margins, a short term and a medium term. For third quarter guidance, I just want to see why you're guiding a decline for margins on a small increase in sales. And into 2014, as you ramp 20-nanometer with a steep ramp, do you see any margin impact in the early stage of 20-nanometer?

Randy, you're asking about the third quarter guidance on the margins. We are adding capacity in third quarter. I was talking about another 6.5% leading-edge technology. The third quarter, as Chairman just mentioned, we are going through the image corrections period, so we expect the third quarter utilization will be lower than second quarter. That's the main reason that the revenue improved, the margin does improve. You also asked about the, going to 2014, when we ramp 20-nanometer, how will that affect our margin. Every new technology, when it comes to the mass production, it -- in early beginning, it always start with a lower margin, but when in -- the quantities start to go up, maybe after 7 or 8 quarters, the margin will be getting closer to corporate average. So we expect that will happen for 20-nanometer as well. Randy Abrams - Crédit Suisse AG, Research Division: Can I have one follow-up on that question? Do you expect, even with 20-nanometer ramping up, structural profitability will continue to see, at least maintain -- or how do you see structural profitability and depreciation in 2014?

We are confident we can maintained the structural profitability. Randy Abrams - Crédit Suisse AG, Research Division: Okay. The second question, and it probably relates to some of the pictures of the big fabs that are getting built, and I want to ask if you have any concern. You spend the CapEx and some of these large customers' switch foundries so that they switch foundries every node, and so you put a lot of investment. Just curious to the assurance that if you bring in some of these projects, you can maintain that business for several nodes.

Switch foundries, meaning switch... Randy Abrams - Crédit Suisse AG, Research Division: For TSMC. Just if you put in these fab capacity targeting some large customers, just the assurance you can maintain that customer relationship for several nodes.

Well, I -- you realize that we don't always put in the amount of capacity that a customer requests. And we do have -- we do make our own estimates and so on. And so we have not and will not always follow the customer's estimates for their capacity need [ph]. So I don't know if that answers your question or not. I mean, your question, as I understood it, was that, you put in the capacity and the customer switches foundries, et cetera. My answer is that, well, we try to anticipate what foundries they switch. You still look a little puzzled, so... Randy Abrams - Crédit Suisse AG, Research Division: Okay. It was -- great. I -- maybe to be rephrase it: If -- because some of these customers are getting quite large in amount of capacity they demand. So for product generation, you build capacity for that customer, but because it's a huge volume, at the next node, it switches. Like, if you...

They switch to what? Randy Abrams - Crédit Suisse AG, Research Division: Switch to a different -- switch their business to another foundry.

Oh, that does not happen. That does not happen, no. You must be -- your experience must be with commodity products. Well, I'm sorry to say that because, obviously, your experience is not only with commodity products. But I can assure you that -- I mean, I kept emphasizing all through the last 10, 15 years that our strength was in 3 directions: technology, manufacturing and customer relationships, all right? So when you say that a customer works with us and then suddenly -- I mean, works with us and we build the capacity, and then suddenly after the capacity is built, he switches a foundry. That does not happen, for his sake as well as for ours.

Okay, good. Next question would be coming from the floor again from Deutsche Bank, Michael Chou. Michael Chou - Deutsche Bank AG, Research Division: Chairman, you gave the guidance before regarding 20-nanometers house portion in Q2 with a low single digit. Do you have any revision for that, or you maintain the same guidance?

Q2 as -- is it -- what? Michael Chou - Deutsche Bank AG, Research Division: Q2 next year.

Q2 next year. How much 20-nanometer will account for our revenue?

Well, Q2 next year, I don't know about that. For the whole year next year, I expect it will be in the high-single-digit percent of our total revenue. I think that's correct. Isn't it?

Yes.

For the whole year next year, I think 20-nanometers will be in the high-single-digit percent of our total revenue. Michael Chou - Deutsche Bank AG, Research Division: My second question is, what is the outlook by segment in Q3 this year?

The what?

Segment. Segment outlook. I think, Lora...

So let me take that. In third quarter, computer will decline the most; followed by a -- communication, with modest decline. We expect consumer industrial for TSMC will go up in third quarter.

Next question will still be coming from the floor, and it will be from JPMorgan, J.J. Park. J.J. Park - JP Morgan Chase & Co, Research Division: The first question is the, what's your view on the longer life cycle for the 20-nanometer given increasing capital intensity moving down to 20-nanometer, and then FinFET? So I believe the 20-nanometer, the life cycle, could have been longer than the previous technology.

So your question is whether or not 20-nanometer will have a shorter life cycle, as compared to prior nodes.

I think, yes, probably. But well, I said our GIGAFAB is for both 20-nanometer and 16-nanometer. So yes, 20-nanometer will -- I think, will have probably, I'm not sure yet, a shorter life than 28-nanometer. But then we convert it quickly, convert the capacity, to 16-nanometer. I think 20-nanometer and 16-nanometer together will have a longer life than 28-nanometer. Well, actually, everything has a long life in TSMC, anyway. We are still making a 1 life form [ph] or -- 1 life on [indiscernible], yes?

[indiscernible]

1.5? Well, I guess, we are not making 1 life form [ph] anymore 0.5 life [ph]. And so when we say life, short life or long life, the -- you will refer to a customer's use, a specific customer's usage, yes, maybe. But we have second wave, third wave, and we have the specialty technology users who are something outside the second wave and third wave. We have second-wave, third-wave logic users. And then we have -- outside the second and third wave, we have the specialty technology who, in time sequence, are usually in the second, third or fourth or fifth wave, yes. So... J.J. Park - JP Morgan Chase & Co, Research Division: Okay. And my second question is the, you mentioned about the competitive advantage at gate-last. And if I don't know, your competitor is going to use the gate-last from the 20-nanometer. So based under your experience, how difficult and how much time they require to commercialize gate-last at the 20-nanometer from the gate-first?

So your question is with regard to whether or not there will be competitors switching to gate-last at 28-nanometer. J.J. Park - JP Morgan Chase & Co, Research Division: From the 20-nanometer.

In 20-nanometer, switching at 20-nanometer to gate-last. So how much time, how difficult it will be for such a competitor to switch?

I don't know. It depends on his capability, I think, I don't know. I think it's -- it takes a long time, that's for sure, yes. A year, maybe even longer. And do I see anybody switching, any competitors? I don't even know the answer to that. I do know that we don't see any serious competition at the 20-nanometer node.

Okay. I think it is about time that we should take a question, next question, from the call.

Your first question, from the -- comes from Steven Pelayo of HSBC. Steven C. Pelayo - HSBC, Research Division: I'm curious about, you mentioned third quarter utilization rates declining slightly. You mentioned the possibility of fourth quarter revenues also being down. You also talked about capacity being up a lot in the third quarter. So I guess I would like to hear you comment on utilization rates by node. And in particular, are you keeping the leading-edge nodes full? And if not, the negative leverage affects the margins if the leading-edge is not fully utilized.

Well, first, to answer your question, generally, I do think, I -- actually, about a year ago, I predicted almost the same phenomena for the fourth quarter last year and the first quarter this year, about a year ago. In fact, it was a year ago when I said that the fourth quarter might be down and the first quarter might also be down, fourth quarter last year and first quarter this year might a little bit. By the second quarter, we will see a strong rebound. I said that in July of 2012. And it happened that way, pretty much. And now this year, July 2013 and I see almost the same thing, except I think there is a bit of difference. The fourth quarter down could be a little more severe than the last year -- than the fourth quarter down last year. And however, the first quarter down will not be as severe as the first quarter down this year. And did I make it -- myself clear? I mean, I think that, last year, incidentally, our fourth quarter was down by, it's about, 7%. And I just said now that -- well, anyway, it could be a little more severe than that. And the first quarter, however, could be reasonably flat from the fourth quarter. And however, the second quarter, next year, second quarter, rebound is going to be, I believe, a very strong one, just as this year's second quarter rebound was. So that's answering the question generally. And I think that -- our third quarter, by the way, you said the third quarter is -- fourth [ph] quarter will also be down. Our third quarter is up, it's not down, okay? And -- but you asked whether the margin -- you said...

Yes, Steve's question was, if the loading rate of leading nodes... Steven C. Pelayo - HSBC, Research Division: I was just asking [indiscernible] utilization rates by node, if the leading edge isn't yet fully utilized, is that a more profound impact to your margins?

Oh, all right. 20-nanometer is still fully loaded in the third quarter. And how well it will be loaded in the fourth quarter is, at this point, unknown. But I think that I -- it may be less than 100% loaded in the fourth quarter. Now the other nodes, 45-nanometer and 65-nanometer, et cetera, well, they will not be -- some of them are still fully loaded in the third quarter but may not be fully loaded in the fourth quarter. But we are really not giving a fourth quarter guidance right now, so I want to limit that, sorry. Steven C. Pelayo - HSBC, Research Division: Okay. And then just one final question, which is really just a clarification. I think, last quarter, you spoke about the 20-nanometer in its first year being bigger than what 28-nanometer was. This quarter, you're announcing 20-nanometer is going to be high-single-digits percentages of revenues, so I guess it depends on what we think the total revenue number is going to be. I'm just curious, is there a change there on what you think the total dollar contribution can be next year for 28-nanometer -- for 20-nanometer at TSMC?

Well, actually, the first year of 28-nanometer was 2011, yes. The first year of 28-nanometer was 2011, and I'm quite sure that, in 2011, 28-nanometer did not reach 8% of our total revenue, yes.

Next question, I think we will still keep it on the calls.

Your next question from the phone comes from Mehdi Hosseini from SIG. Mehdi Hosseini - Susquehanna Financial Group, LLLP, Research Division: Doctor Chan, early on in your prepared remarks, you talked about the growth rates of -- for high-end smartphones peaking and the node to mid-end is growing at a faster rate. And also you talked about the very extensive and very expensive fab for 20-nanometer and 16-nanometer that is under construction. So do you think that profitability for the new smartphones, [indiscernible] think that they're going to be selling at a lower price, could help breeding out demand to fill the GIGAFAB?

Mehdi, I think I am able to capture the first part of your question. But the second part, after you said they we're building GIGAFABs in 20 -- for 20-nanometer and 16-nanometer, what was -- can you repeat that part again? Mehdi Hosseini - Susquehanna Financial Group, LLLP, Research Division: Sure. Let me repeat the entire question. I'm just trying to better understand [indiscernible] but the demand -- the trend in the smartphone and the trend towards the low end, and we're getting now profitability for the fab foundries company to migrate to 20-nanometer and 16-nanometer nodes.

So you want to -- your question is, based upon the trend, the smartphone trend is towards faster growth on the lower end, whether or not there's still the need to fast RAM, the leading-edge nodes such as 20-nanometer and 16-nanometer. Is that your question? Mehdi Hosseini - Susquehanna Financial Group, LLLP, Research Division: Yes.

You confirm that, that was his question, but would you repeat the question for me?

Right. His question is, the trend in smartphone is shifting towards higher growth at low end, but we are still building very fast a leading-edge capacity at 20-nanometer and 16-nanometer.

So we are still building...

Leading-edge capacity at 20-nanometer and 16-nanometer.

Yes, yes. I think the question was whether the low end will migrate to 16-nanometer.

Yes.

Yes, I think so, very definitely. Well -- but it's only a question of time. And I don't frankly know. 16-nanometer? Well, even the leading edge won't start using 16-nanometer until 2015. And I think that the lower-end manufacturers will probably be, well, I would say, at least 2 years behind, 2017, yes. Are you worried about -- is he worried about our capacity utilization for 16-nanometer? Well, I don't know what I can say to assure you again and again and again. Well, I mean, this is a question that came up almost for every generation. And I -- -- again, I think, well, first, the premise: the premise is that we do not build capacity until we are quite confident of the demand. That's premise number one. We do not build capacity until we are quite confident of the demand. And we look at the demand not by market segment but by customer, yes. And we base the confidence on the work that we have already been doing with the customer. That's premise number one, that we don't build capacity until we are reasonably confident, highly confident, of the demand. The second premise is that we do have waves, succeeding waves, of customers when the first wave migrates to an even more advanced node. We have second wave, third wave, and we have specialty technology customers taking over the capacity that the first wave, the succeeding waves ahead of them, have left behind. Well... Mehdi Hosseini - Susquehanna Financial Group, LLLP, Research Division: [Indiscernible] you are very clear that you are now going to build capacity ahead of demand. So would you comment or would you have also any idea of how next year's CapEx could look like at this point in time?

Next year's CapEx. His question is next year's CapEx.

Next year CapEx is -- will be about the same, about the same as this year.

Okay. Let's come back to the floor. The next question will be coming from the floor, and would be coming from Morgan Stanley's Bill Lu. William Lu - Morgan Stanley, Research Division: Dr. Chang, you just talked about building capacity based on customers. I went back and looked at your annual report, Top 10 was 51% of sales in 2007. Last year was 59%. Top 1 was 11%, 2007; 17%, 2012, so quite a big increase in customer concentration. And I think, probably, you can expect that going forward, that gets bigger, right, with a big customer coming on board. So Top 2 could be 1/3 of your revenues.

That makes it easier for us in terms of building capacity, in terms of estimated capacity, right? William Lu - Morgan Stanley, Research Division: Yes. But I guess, I'm just wondering, does that -- do you have more customer-specific risk going forward? And how do you deal with that?

Well, there are always risks. You've got risks crossing the street, but you'll take it. And I think that frankly, the way I look at it, we want to make a relationship with the major, major customers such that the risk of their -- the way [Chinese] the way he put it, switching boundaries. The way we work with the customers, the major, major customers, makes the risk of their switching boundaries almost as small as crossing the street. William Lu - Morgan Stanley, Research Division: I guess I'm not as much worried about them switching as much as one customer just doesn't do well with a poor product line or something like that.

Has a what, a blowout? William Lu - Morgan Stanley, Research Division: I guess, I think, customer risk could be, one, is switching, but, two, just one particular chip doesn't do well or...

Yes, sure. Sure, that risk exists. And as I said -- remember what I said earlier that we don't always build capacity -- build as much capacity as they would like us to build. Everybody, I think, tends to be a little optimistic about his own new products or whatever, new markets and so on, yes. William Lu - Morgan Stanley, Research Division: My second question is, you broke out smartphones by high-end, mid-end, low-end. I think it was last year or maybe the year before you had given a content per smartphone type of number for TSMC. Can you give us that maybe now versus when you gave it...

Do you remember the number? William Lu - Morgan Stanley, Research Division: I think it was 8-point-something dollars.

Was it that much? No.

I think we said $7 of average.

What's the use of telling you this if you don't even remember what I told you last time? Well, anyway, the reason I try to -- I test you is because I wanted to tell you a new number. But I just, first, wanted to test whether you remember the last one or not. But in any case, for the high-end or the middle end, the number has risen by about $1. The smartphones have become smarter partly because they carry ICs -- more ICs made by us. And that's why the content has -- or value added in each high-end. And middle-end has risen by about $1. The low-end, I don't -- I haven't seen a number. Do we have low-end value? Yes, you go ahead.

On an average smartphone, we have $7 per phone. On the low-end, it's $4; middle-end, $6; and high-end, $9. William Lu - Morgan Stanley, Research Division: Okay. So the average hasn't changed from before?

$7.

Our average has jumped, but low-end has increased. In the low-end, you can't have a $8 or $9 in the low-end. The consumers don't want to pay that. They don't need all those features either.

Our next question still comes from the floor and will be from Goldman Sachs, Donald Lu. Donald Lu - Goldman Sachs Group Inc., Research Division: [Chinese] Sorry, I've got to be careful, I mean, asking questions so I don't get a question back. My first question is on the second sourcing. I noticed a trend. Before 28-nanometers, there's always a substantial amount of second sourcing going on. In 28 and 20, it seems like very little. And in your speech, quoting your word, you said for FinFET, we need substantial joint work between foundry and customer. Does that mean the second source kind of business model is not going to work very well in the future in terms of foundry second source?

It's very difficult. Foundry second source has always been a difficult thing. And I think that -- but when it will work, yes, I think it could work. But the second source usually will be considerably behind the first source. Donald Lu - Goldman Sachs Group Inc., Research Division: Will this gap continue to increase with each node going forward?

I don't believe so. Certainly -- let me put it this way. Where we are the first source, we are going to certainly do our best to prevent a second source, all right. Where we have the opportunity of become a second source, we will often refuse to be one. It's very different. It's -- I think it's difficult. And I do not think it's the way to go for either the customer or the supplier. Donald Lu - Goldman Sachs Group Inc., Research Division: Just a follow-up on that question, if a customer for a particular product, if it -- is that still possible to source it at both TSMC and another foundry at the same time? Or is that the design process is so different. It's going to be...

It's difficult for the same product; for the same technology, yes. I think -- I can see that's possible. It's not only possible, but that's not really second source anymore, yes. That's the 2 first sources in the same technology, yes. Donald Lu - Goldman Sachs Group Inc., Research Division: Okay. So that's still possible?

Yes. Donald Lu - Goldman Sachs Group Inc., Research Division: And the second question is today, when you showed those pictures of fabs, you stressed that all the GIGAFABs can be connected. And I noticed that comparing TSMC with Samsung and Intel, TSMC definitely have huge fabs. They are connected versus -- the other 2 are more scattered around the world for various reasons. For TSMC, number one, that will give you more cost advantage. Is that still the case for FinFET, et cetera, going forward, even you have one large -- very huge customer?

Cost advantage, yes. I believe, there is some cost advantage in connecting all the fab into one GIGAFAB. But I think the main advantage is probably in time to expand or time to market. That is -- we qualify only once because -- in one big GIGAFAB, we qualify only once. Whereas if they are disconnected, if they are separate fabs, there, usually, we have to qualify each fab on the same product for this, yes -- well, on the same technology, let's say, yes. Donald Lu - Goldman Sachs Group Inc., Research Division: And just to follow up on that, does that -- this also mean TSMC will probably not build a 12-inch fab, let's say, in the U.S. or China? Is that strategy still true?

Oh, yes. You are predicting things for me, Donald. No, I -- we always consider doing things. But I think every time in the past, we just run up against a stop, which is very costly to do it in a separate location -- in another very different location, yes. Well, we have expanded our China fab considerably in the last 2 years. It is not what you knew 2 years ago. It's twice as large as the size 2 years ago, yes. Donald Lu - Goldman Sachs Group Inc., Research Division: Is that profitable?

Oh, yes. Oh, yes. Oh, yes. Why will we have -- why will we do it if it wasn't, Donald?

Okay. Next question also comes from the floor, and it will be from...

It is very profitable, actually.

Hopefully, next question is also related to a very profitable question from Barclays, Andrew Lu. Andrew Lu - Barclays Capital, Research Division: Dr. Chan and Lora, I have 2 questions. First one, I remember last year almost at the same time -- I think it's July 20-something. You mentioned that Q4 will be down. In that time, you said what the ballpark of the double-digit probably close to 10% decline, but -- and there was a rush order and Q4 rather declined only 7%, the same for Q1. Earlier, you guide a similar decline like Q4. And that was 1% decline. Based on your current visibility -- earlier, you mentioned it will be similar or worse, worse than last year, are you saying worse than 7% decline or worse than what your original guidance last year, where you give about 10%?

No. I meant worse than the 7% -- the actual 7%, yes, could be worse. Andrew Lu - Barclays Capital, Research Division: Could be double-digit?

Well, actually, I kind of regret that I even went into this fourth quarter thing now. No, I -- no, I don't expect it to be double. I don't -- look, I think it would be a decline and -- but I normally don't even guide or forecast the quarter after this. But when I see something unusual happening, I do try to tell you in advance. And that was the case last time, and it's a year ago. And it's the same case this time -- same situation this time. I see a very -- a finite possibility of the fourth quarter being down from the third quarter. As to its magnitude, I really -- even when I said that maybe a little more serious than the minus 7% last year, I was taking a risk on -- and that risk is greater than causing [indiscernible] okay. Andrew Lu - Barclays Capital, Research Division: My second question is regarding the FinFET competition. Intel -- based on all industry check, Intel will run a 14 FinFET by second half next year, probably will start to do the foundry for year 2015. And Samsung claim -- and they are going to jump from 28 to 14 FinFET, similar -- like the year 2015. Our FinFET also are mass-produced from early '15. So my question is, based on these industry competitors and also some customers' comment, our 16 FinFET, that size is larger than our competitors' 14 FinFET, and the performance are a little bit worse than compared to 14 FinFET. Do you have any words to defend this statement?

No. I'm telling you that our 16 -- well, first of all, you have to remember, nothing is out yet. Everybody is just talking, talking, talking, okay. And then I stand on what I said. I guess, it was 30 minutes ago now. On a foundry to foundry competition, I believe we are competitive on the 16. On a grand alliance versus IDM competition, I believe we are more than competitive.

Okay. I believe this is about time for us to take a question from the call. Operator, could you please proceed to the next caller?

Your next question from the phone comes from Brett Simpson from Arete Research. Brett Simpson - Arete Research Services LLP: Dr. Chan, you mentioned that the competition are not ramping 28 high-K metal yet. You're not seeing any competitors there. And your loading will start to come down towards the end of this year and then rebound in the second quarter. So during this time, we're going to see TSMC adding new capacity. You're ramping your new GIGAFABs. But also this time, you're going to have GLOBALFOUNDRIES clients who have a lot of 300-millimeter capacity. They eventually get into 28-nanometer high-K metal gate. So my question, are you concerned at all about overall foundry supply that's coming onstream over the next 12...

All right. Brett, your question is with respect to the potential oversupply condition for 28-nanometer, starting fourth quarter of this year. You are concerned whether or not TSMC, as well as competitions, are building -- continue to build 28-nanometer capacity. It will create a clot. Is that your question? Brett Simpson - Arete Research Services LLP: Yes, particularly GLOBALFOUNDRIES. You seem to be adding a lot of 300-millimeter capacity. And at some point, they will start to ramp up 28-nanometer high-K metal gate. So I'm trying to understand whether Dr. Chang sees any oversupply conditions over the next 12 months, yes.

28-nanometer oversupply for next year, I don't think it will happen. Well, you want to know the reasons why it won't happen? Well, look, I already said it. I think we have a substantial lead on yield and on performance in 28-nanometer. So while other competitors or at least some of the other competitors are talking about building capacity and even actually building capacity, but -- I think that we will have a much higher utilization in our capacity than those competitors, much higher. And this has happened before. When -- this impact has been happening all along in the last 15 years or so. While we always build capacity when we knew who our customers would be and at least -- we knew at least approximately, what they're demand would be, real demand, while we build our capacity on that kind of knowledge, our competitors often, often build capacity on speculation. So you may say, "Well, that's certainly not very advantageous to you either." Well, now, it was not very advantageous to us, but we still managed to hold our profitability. We still -- over all of these years, in almost all nodes, 0.13, 90-nanometer, 60-nanometer, 65-nanometer, 45-nanometer, I mean, it has always happened. Our competitors build capacity on speculation. Well, the result in every generation was that we still got our profitability. In fact, our structure profitability, as I pointed out, has improved. And we also had very much high utilization in our capacity at each node than our competitors. And if you look at the history, you will find out this is -- this was why our profit -- I think even now -- of course, GLOBALFOUNDRIES is not a public company. And Samsung, Intel, I don't think exactly disclose their foundry revenue or income. But back when we had competitors -- foundry competitors that disclose their revenue and earning, we had only 50% of the revenue of the total industry. But we had 100% of the profit. So that means that the rest of the competitors are -- I mean, they even had -- if they had negative profit -- well, if they had positive profit, it was canceled out by some others' negative profit. We had a 50% revenue and 100% on the profit. And I think that it will continue to be that way, frankly. So I -- yes? Brett Simpson - Arete Research Services LLP: Can I perhaps just ask a follow-on question, Dr. Chang, about a wafer ASP? Your wafer pricing has been rising quite nicely over the last 12, 18 months as you've ramped 28-nanometer much faster than prior nodes. Can you talk a little bit about how you see wafer ASPs trending over the next few years as you start to ramp up 20-nanometer and 16 FinFET. Should we continue to see wafer ASPs rise their levels?

Yes. The blend of the average price will continue to rise.

All right. Let's come back to the floor. The next question will come from Citigroup, Roland Shu. Roland Shu - Citigroup Inc, Research Division: Chairman, I think the first question for me is for the 16-nanometer adoption. Will a customer skipping 20-nanometer go to 16 directly or a customer will take -- more likely to take -- sequentially migrate to 20-nanometers first and then to 16-nanometer? That is my first question.

Will they skip 20-nanometer? Is that the question? Yes. I think some customers might. Some, some, but I think a larger part of the customer -- a larger percentage of the customers will go to 20 first and then 16. Roland Shu - Citigroup Inc, Research Division: Okay. So a follow-on question will be for those 20-nanometer customer, will it be enough incentive for them to move to -- from 20 to 16? So this is -- of course, we know there are a lot of improvement from 28 to 16. How about the comparison to -- from 20 to 16? Will they offer more incentive for customer to migrate to...

Now you're talking about performance and the power and that sort of things. Well, I think, they get a bit too detailed for me to talk to you here.

But I think we have said, our advantage of 16-nanometer over 20 is for speed, performance. It will be 20% faster at the same total power. And it will be 35% better efficiency in power given the total speed -- given the same speed. Roland Shu - Citigroup Inc, Research Division: And my second question is on the gross margin side. I think in 3Q, definitely, we are going to have more 20-nanometer wafer contribution. I think that, definitely, for the first quarter, 24%; second quarter, 29%; then 3Q [indiscernible] will be up to -- above 30%. And also, we'll have more high-K metal gate. So the blended ASPs should be up. But still, we still have the lower gross margins. So can Lora comment, maybe give some idea? What's the margin impact from this increasing 28-nanometer and also from this decreasing utilization in 3Q?

28-nanometer continue to increase actually helped the overall margin. Even in the third quarter, I was guiding a lower profitability but not because of 28. 28 is still doing very nicely, and it will be fully loaded if other nodes will be less utilized. So to answer your questions, that's basically other nodes has lower utilizations.

Roland, are you done with all your questions? Roland Shu - Citigroup Inc, Research Division: Yes.

Okay, great. All right. Next, he's the first. I only take 1 question, okay. Next one comes from -- I think, you're Daiwa, right? Daiwa's Eric Chen. Eric Chen - Daiwa Securities Capital Markets Co. Ltd., Research Division: Very quickly, one question regarding to your China strategy. We talked about a lot of the high-end, the mid-end and the low-end of smartphone. And then from my understanding, your market share and for the China [indiscernible] in the smartphone IC lower than your market share or the global smartphone IC maker.

I don't know whether that's true or not. Eric Chen - Daiwa Securities Capital Markets Co. Ltd., Research Division: Okay, my guess. Let me say this way. So...

You're shaking your head to deny that is true or what?

I don't think it is lower, yes. Eric Chen - Daiwa Securities Capital Markets Co. Ltd., Research Division: Okay. That's as soon as the higher. And my question is the china smartphone IC maker, they care about the wafer price.

They what? Eric Chen - Daiwa Securities Capital Markets Co. Ltd., Research Division: They care about the wafer price, price sensitive.

Yes. Eric Chen - Daiwa Securities Capital Markets Co. Ltd., Research Division: And they probably park in the wafer price and even more severe, and then again, they are more sensitive. So any strategy you have for loss of the China, the type of PC IC maker, all the smartphone IC maker. Probably, they [indiscernible] one guy just moved to the GLOBALFOUNDRIES. So are you seeing this trend?

Again, anybody -- somebody moves to a competitor, I feel very sad. This particular one, the last time, when I said that I was very regretful when another one of our customers moved to Intel. Now this particular one, I know -- I think I know what you're talking about. I know who you are talking about. I mean, I will use the word that's considerably less than regret, okay. So yes, we know that they are very price conscious. And we try to work them. Well, actually, we -- I should say we try to work with our customers who have to work with their customers because their customers are the price conscious ones, yes. So yes -- but as I reported in my message, we've been quite successful so far, yes, so... Eric Chen - Daiwa Securities Capital Markets Co. Ltd., Research Division: My second question probably go to Lora, first. Where -- I mean, for the CapEx for next year is spread from this year, what kind of -- how many percent the competitive gross and the depreciation expenses gross in the foundry this year?

I have no idea at this moment. Chairman just gave you a ballpark of the total CapEx. So we have not calculated based on that.

Well, I gave him actually a little more than ballpark. I gave him the size of the diamond anyway. Well, are you a baseball fan? You know what I'm talking about? Eric Chen - Daiwa Securities Capital Markets Co. Ltd., Research Division: Yes, I'm a baseball fan.

A ballpark is maybe 10x the size of the diamond. Eric Chen - Daiwa Securities Capital Markets Co. Ltd., Research Division: Okay. And actually, the follow-up with a question regarding to the CapEx, I just really want to know exactly why are you so confident for the demand. Yesterday, Intel announced they are going to cut their CapEx by...

Why am I so confident of the -- of what demand? Eric Chen - Daiwa Securities Capital Markets Co. Ltd., Research Division: The market demand and then your CapEx.

I'm confident because I have usually been right.

All right. I think in the interest of time, and this is already a little over -- it's about 100 minutes now. So I think we'll just end our investors conference for this quarter right now, and thank you for coming. We'll see you next quarter.