Conference call. This is Elizabeth Sun, TSMC's Director of Corporate Communications and your host for today. Today's event is webcast live via tsmc.com. If you are joining us through the conference call, your dial-in 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 CFO, Ms. Lora Ho, will summarize our operations in the third quarter, followed by our guidance for the current quarter. Afterwards, Lora and TSMC's two co-CEOs, Dr. Mark Liu and Dr. C.C. Wei, will jointly provide our key messages. We will open both the floor and the line for the Q&A.
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 tsmc.com. Please also download the summary slides in relation to today's earnings conference presentation. As usual, I would like to remind everybody that today's discussions may contain forward-looking statements that are subject to significant risks and uncertainties, which could cause the 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. Now I would like to turn the podium to TSMC CFO, Ms. Lora Ho, for the summary of operations and current quarter guidance.
Thank you, Elizabeth. Good afternoon, everyone. Thank you for joining us today. I will start my presentation with financial highlights for the third quarter, followed by the guidance for the fourth quarter. We had a good quarter. This third quarter, we have set a new record of revenue and profitability thanks to strong demand and our successful ramp of 20 nanometer. Our revenue increased 14% sequentially, 29% on year-over-year basis, to reach TWD 209 billion. Our gross margin exceeded 50% to reach 50.5%, which is also a record since the second half of 2006. Compared with second quarter, gross margin improved 0.7 percentage point. The higher margin was contributed by consistent cost improvement, variable inventory valuation adjustment, partially offset by 20 nanometer dilution, as we are still in the early stage of the production.
Operating margin was 40.4%, up 1.8 percentage point from the second quarter, reflecting an improving operating efficiency for the company. After a big jump in tax rate in the second quarter due to an accrual of 10% return earning tax, in the third quarter, the effective tax rate fell back to normal level of about 11% of profit before tax versus the 20% in the second quarter. Overall, the third quarter EPS increased 47% sequentially to TWD 2.94. The single quarter ROE was 33.3%. Let's take a look at revenue by application. Compared to the second quarter, communication showed the strongest growth. Revenue increased by 26%. Industrial-related revenue also increased 9%, while computer and consumer declined 6% and 3% during the third quarter. By technology, after two years of meticulous preparation, we began volume shipments of 20 nanometer wafers.
The revenue contribution went up from 0% to 9% of the third quarter wafer revenue. This is the fastest and the most successful ramp for a new technology in TSMC history. In addition, customer demand for 28 nanometer wafers continued to be strong. Our 28 nanometer wafer revenue continued to grow sequentially in the third quarter, representing 34% of total wafer revenue. Accordingly, the two advanced technologies, 20 nanometer plus 28 nanometer, represented 43% of our third quarter total wafer revenue, increased from 37% a quarter ago. Now let's move on to the balance sheet. We ended the third quarter with cash and marketable securities of TWD 290 billion. Current liability decreased by TWD 74 billion, mainly due to we paid out the TWD 78 billion of cash dividend in August. On the financial ratios, accounts receivable turnover days is 44 days, which is the normal level of our average days of receivable.
Days of inventory increased five days to 56 days, mainly due to higher working process inventories associated with the fast ramp and the longer cycle time for 20 nanometer. Now let me make a few comments on cash flow and CapEx. During the third quarter, we generated TWD 91 billion cash from operations, invested TWD 48 billion in capital expenditure, and paid out TWD 78 billion in cash dividend. At the end of the third quarter, our cash balance decreased TWD 29 billion to TWD 255 billion. Free cash flow for the third quarter was an inflow of TWD 43 billion, a big improvement versus previous quarters. In US dollar, our second quarter CapEx was $1.6 billion. This adds to the total of $7.8 billion for the first three quarters. Regarding our capacity, we expect to increase our capacity by 12% from last year.
Total annual capacity will reach 8.2 million 12-inch equivalent wafers this year, slightly higher than our previous estimate of 8.1 million. I have finished my report on the financial part. Now let me turn into the fourth quarter outlook. Based on our current business outlook and the forecast exchange rate of 30.30, we expect our fourth quarter revenue to be between TWD 217 billion and TWD 220 billion. This will translate into around 4%-5% quarter-over-quarter increase. On the margin side, we expect the fourth quarter growth margins to be between 48%-50%, and operating margins to be between 38%-40%. You may ask why we guide slightly lower margin rate despite 4%-5% revenue growth. This is because we will continue to ramp our 20 nanometer to more than 20% of our wafer revenue in the fourth quarter.
We expect to see a mild margin dilution with our aggressive productivity improvement efforts. This concludes my remarks. Thank you very much.
All right. We will deliver our key messages. They will be offered by our CFO as well as by the two presidents and co-CEOs. We will start with CFO, Lora Ho first.
I will talk about the three items, the CapEx, free cash flow, and supply chain inventory. Let me start with CapEx. In January of this year, we guided TSMC's 2014 CapEx budget to be between $9.5 billion-$10 billion. Today, we are able to provide more specific number, which is about $9.6 billion. Majority of this year's CapEx budget is spent for 20 nanometer expansion. In 2015, we will continue our investment in 16 nanometer capacity installation and 10 nanometer engineering capacity. Based on our current plan for 2015, our CapEx budget for the next year is likely to be slightly higher than $10 billion. On free cash flow outlook, you may recall in the past four years, TSMC has increased CapEx substantially to capture the growth opportunity brought by the mobile computing devices.
Our free cash flow dropped to about $1 billion-$2 billion per year, which were below the cash dividend we paid. The cash dividend, $3 cash dividend is about $2.6 billion a year. With a substantial increase in operating cash flow this year and a similar level of capital expenditure compared to last year, we expect to more than double our free cash flow in 2015, and we are confident that the free cash flow level will rise further in the foreseeable future. The sustainably higher free cash flow should enable TSMC to afford paying a higher level of dividend per share going forward. The board of director will consider increase in cash dividend in February of next year.
Regarding the supply chain inventory, we have noted in our last quarterly conference that we estimate fabless DOI will increase and be two days above seasonal at the end of third quarter. Fabric DOI will decrease and be two days below seasonal level at the end of this year. Our data and model for the forecast still warrants the same estimate today. You can see from this chart, we estimate 4Q14 fabric DOI will be two days below seasonal. Compare with 4Q13 last year, which is one year ago, the fabric DOI was 60 below seasonal. Based on our model, we anticipate a much milder inventory correction in fourth quarter this year. With that, I will turn the podium to Mark, who will share with you our view on the near-term demand.
Good afternoon. I will continue cover the near-term demand. The strong demand of our 20SoC customers enable our continued growth in the fourth quarter, overcoming our seasonal demand pattern of a sequentially weaker fourth quarter, and the cautious inventory adjustment actions taken by some of our customers rendering the slower 4Q demand. This fourth quarter demand from our customer does not validate the recent forecast by Microchip. Our recent demand in China still appears normal, little deviation from their seasonal pattern. We see China's 4G smartphone sales and infrastructure buildup remains to be very aggressive. We are expecting another record-breaking quarter with a 4%-5% growth in the fourth quarter. 10 nanometer development. Our 10 nanometer development is progressing according to plan. Currently, we are working on early customer collaboration for product tape-outs in 4Q 2015.
The risk production date remains targeted at the end of 2015. Our goal is to enable our customers' production in 2016. To meet this goal, we are getting our 10 nanometer design ecosystem ready now. We have completed certification of over 35 EDA tools using Arm CPU core as vehicle. In addition, we have started the IP validation process six months earlier than previous nodes with our IP partners. We are working with over 10 customers on their 10 nanometer product design. The product plans show wide range of applications, including application processors, baseband, CPU, server, graphics, network processor, FPGA, and game console. Our 10 nanometer will achieve industry-leading speed, power, and gate density. I'll say a few words on our clarify next growth momentum of TSMC. We think the growth of smartphone and tablet in propelling our revenue growth will continue for at least several years.
In addition, the recent innovations in wearable devices, including smartwatch, in cloud computing, in fog computing, in Internet of Things, including smart car, smart homes, all are very exciting. We are currently closely working with our customers on all these applications to set the stage of the next growth wave to move us forward. That's my comment. Now I turn the microphone to C.C.
Thanks, Mark. Good afternoon, ladies and gentlemen. This afternoon, I will update you the 20 nanometer ramp status, followed by 16 nanometers of progress and 28 nanometer status. First, the 20 nanometer ramp status. We shipped 20 nanometer in high volume during third quarter. The yield is meeting our target. Revenue generated from 20 nanometer accounted for 9% of third quarter wafer revenue. Because of the strong demand from the high-end 4G smartphone, which are equipped with 64-bit cores, the LTE Cat6 or Cat7, and more advanced graphic and video performance, our 20 nanometer will continue to grow and is expected to contribute greater than 20% wafer revenue in the fourth quarter. We expect the strength of our 20 nanometer business to continue in 2015. We expect the revenue will account for roughly 20% of next year in wafer revenue.
I'll talk about the 16 nanometer ramp and the competitive status. In 16 nanometer, we have two versions, 16 FinFET and the 16 FinFET Plus. FinFET Plus has better performance and has been adopted by most of our customers. 16 FinFET, we began the risk production in November last year, and since then, have passed all the reliability qual early this year. For the FinFET Plus, we also passed the first stage of the qualification on October 7th, and since then, entered the risk production. The full qualification, including the technology and product qual, is expected to be completed next month. Right now, we have more than 1,000 engineers working on ramp-up for the FinFET Plus. On the yield learning side, the progress is much better than our original plan. This is because the 16 nanometer uses a similar process to 20SoC, except for the transistor.
Since 20SoC has been in mass production with a good yield, our 16 FinFET can leverage the yield learning from 20SoC and enjoy a good and smooth progress. We are happy to say that 16 nanometer has achieved the best technology maturity at the same corresponding stage as compared to all TSMC's previous nodes. In addition to the processing technology, our 16 FinFET design ecosystem is ready also. It supports 43 EDA tools and greater than 700 process design kits with more than 100 IPs. All these are silicon-validated. We believe this is the biggest ecosystem in the industry today. On the performance side, compared with the 20SoC, 16 FinFET is greater than 40% speed faster than the 20SoC at the same total power or consume less than 50% power at the same speed.
Our data shows that in high-speed application, it can run up to 2.3 gigahertz. Or on the other hand, for low power application, it consumes as low as 75 milliwatts per core. This kind of performance will give our customer a lot of flexibility to optimize their design for different market applications. Far, we expect to have close to 60 tape-outs by the end of next year. In summary, because of the excellent progress in yield learning and readiness in manufacturing maturity, and also to meet customers' demand, we plan to pull in 16 nanometer volume production to the end of Q2 next year or early Q3, year 2015. The yield performance and smooth progress of our 16 FinFET Plus further validate our strategy of starting 20SoC first, quickly followed with the 16 FinFET and FinFET Plus.
We choose this sequence to maximize our market share in the 2016 nanometer generation. I would like to repeat what our chairman stated last time. In combined 20 and the 16 nanometer, TSMC will have an overwhelming leading share every year from year 2014. In total foundry market share, TSMC will lift several percentage points in 2014. He also said that he's happy to add that this trend, increasing the market share, will continue in the year 2015. I'll talk about the 28 nanometer status. We have a strong growth in the second quarter on 28 nanometer. The business grew another quarter and account for 34% of TSMC wafer revenue in the third quarter.
On the technology side, we continue our effort to improve yield and tighten the process corners. So that our customer can take advantage of these activities and shrink their die size, and therefore reduce the cost. Let me give you an example. On 28 LP, the polysilicon gate version, we not only offer a variety of enhanced processes to achieve better performance, we also offer a very competitive cost so that our customer can address the mid-to-low-end smartphone market. In addition to the 28 LP, we also provide a cost-effective High-K Metal Gate version, the 28 HPC, for customers to further optimize the performance and the cost. Recently, we added another 28 nanometer offering we call 28 ultra-low power for ultra-low power applications, obviously. We believe this 28 ULP will help TSMC customers to expand their business into the IoT area.
In summary, we expect our technology spend in 28 nanometer node will enhance TSMC's competitiveness and ensure a good market share. We also expect the strength of the demand for our 28 nanometer will continue for multiple years to come. In response, we are preparing sufficient capacity to meet our customers' future demand. Thank you for your attention.
All right. This concludes our prepared statements. Before we begin the Q&A session, I would like to remind everybody to limit your questions to two at a time to allow all participants an opportunity to ask their questions. Questions will be taken both from the floor and from the call. Should you wish to raise your questions in Chinese, I will translate it to English before our management answers your questions. For those of you on the call, if you would like to ask a question, please press the star, then one on your telephone keypad now. Questions will be taken in the order in which they were received. If at any time you would like to remove yourself from the questioning queue, please press the pound or the hash key. Now let's begin the Q&A session. Our first question comes from the floor of Deutsche Bank, Michael Chou.
Hi. Thank you. Two questions. First question is, you mentioned you're pulling mass production by end of Q2 next year, early Q3. Does that imply you will have an earlier schedule than you planned before?
Yes.
Okay. Second question is, do you have any comment for your market share for next year in 16, 14 nanometer?
Very hard for me to comment on the whole market share. If there's any indication, I would say that we narrow the gap with our competitor.
Okay. It means your view now is more positive than three months ago. Can we say that?
Yes.
Okay. Thank you.
All right. Next question still comes from the floor. It will be from Nomura. Sorry, Daiwa. Losing co-chair. Daiwa's Rick Hsu.
Thank you so much. It's been a while, so hello. Hello, C.C., Mark. Hello. Hi, Elizabeth. Sorry, in English. The first question is about your 20 nanometer ramp-up. I think it's been progressing quite well, and I thought it's going to be around 20% by the end of Q4. I wonder if by that level, 20% at the end of Q4 in terms of revenue, would that reach your corporate average margin?
No, it will not. We have talked about this several times. Any new leading-edge technology, it would take seven to eight quarters to reach corporate average. If you count the 20 nanometer ramp starting from third quarter shipment start, it's by mid-2016 you will get to corporate level profitability.
All right. Thank you so much. That will be the same for 16 nanometer, I assume, in term of progress.
Yeah. We count the 20 and 16 as one node in terms of ramping and schedule capacity, so on and so forth.
This doesn't count as a second question?
All right, you can have a second question. Yes.
Okay. Second question is more about a picture about your China competition. I think recently we had observed quite a lot of moves inside China, including privatization of RDA or Spreadtrum, and OmniVision could be another target to shoot for. Also, your competitor, UMC, last week announced that you're going to set up a Xiamen joint venture with the local government. My feeling is these guys seem to be aiming for the next growth potential market, which is IoT. And by taking the leverage of the huge Demand market in China, number 1, and also the very favorable Chinese government policy in favor of local production. My feeling is, are you worried about this kind of potential competition in the longer term because you only operate in H fab in China?
Do you have any strategy to cope with this potential threat from China in next couple of years? Thank you.
This is a complicated question. Let me put this question into two parts. One is the China government's subsidy effects. Secondly is the IoT opportunities. On China government's subsidy, the recent announcement about CNY 1,200 billion subsidy does cost a lot of activities across the industry. For us, we think currently we have a very strong penetration on our China design houses, many of them over 80%. Most of them are clinging towards leading edge. With this subsidy, they will be more aggressive. We think we will be ready to capture the business, given the existing good penetration. This subsidy may also bring into a merger position because China government wants to small company bring into big to be more competitive. I do think that is a healthy development for the industry with the bigger, stronger design houses to compete with.
There will be downsides because on the back of this subsidy, maybe some of the company will, under the influence of using local foundry and capacities, and that is a threat. May I put it this way. Putting these positive and negative factor together, so long as we have technology leadership, so long as we have a strong manufacturing, so long as we have a good customer service, we think our business opportunity in China will grow, will be bigger with this development. On the IoT, indeed. IoT has been anticipated by many companies, including us. For us, we are currently actively developing IoT-related technologies, okay? Including sensors, including processors, including wireless connectivity, advanced packaging, and power management IC are included. All we are striving for ultra-low power and the power. A technology provide a lower power design.
On the capacity, yes, we are increasing our mature technology capacities today. We will continue to expand the mature technology capacity, deviating from our past strategy, increasing those capacity to capture the potential growth of demand. In terms of further new fab, we don't exclude any possibility, including the fab in China. Okay.
Thank you so much.
Right. Next question will also be coming from the floor. It will be from Bank of America, Merrill Lynch, Dan Heyler.
Thanks, Elizabeth. Thanks. Good afternoon. A couple questions. As you're expanding your capacity in mature technologies for the IoT and MEMS and NFC and the whole range of things, you're expanding your advanced technology. TSMC's held a dominant position in 28 nanometer for almost four years, 20 nanometer accounts for 34% of revenue now. As your key customers are moving to 28 and 16, competition is obviously heating up as well. I'm wondering what steps you guys are taking to keep your 28 nanometer fabs fully utilized. You talked about HPC as a cost-down version. Will this be a potential headwind to keep those fabs full in 2015? If so, how are you able to address the competition as well? Thanks.
We believe the demand on 28 nanometer will continue, as I stated in my statement.
From where?
From where? Okay. All the mobile devices and IoT and a lot of applications. Actually, the 28 nanometer today we find out is a very cost-effective technology. This is the last node that you enter into the
Double patterning. Its application is being widely adapted. We expect that the demand for the next few years, or actually for a long time, they will continue to increase. A lot of our companies will take the advantage of the cost-effective and the performance also. As we said, future from where is from the application side. Application side is also all the mobile devices and all the consumer. Even on the industrial part, we can find some applications.
Okay, great. I presume you intend to hold market share. Do you need to hold the current market share in order to prevent a falling utilization?
You bet.
The second question is with regard to 20 nanometer margins, I guess, for Lora. It looks as though 20 will be over 20% of revenue in the fourth quarter, and your guidance is for 20 to be about 20% contribution for all of next year. That implies pretty much flat revenue. How do you achieve margin expansion on flat revenue?
Dan, fourth quarter 20% is not equal to whole year 100%. It's increased price significantly, is number one. The whole year 2015, 20% is a big number. How do we hold the profitability? I would say today, 2016 profitability is acceptable. I won't say it's great, but it is acceptable, and we continue to work on it as we have done for other nodes in the past. The margin will gradually improve as we have more volume coming on the line. I'm still saying what I said, and we believe we can achieve corporate level margin in a quarter time frame, 2016.
Yeah. My question was just specific on '28 margins. Yeah.
Will additional customers coming into '28, because you previously said you felt this was going to be a major node. Is part of this margin expansion or margin improvement a function of more different products coming in that will help margin, or is it more internal efficiencies and getting smarter, better about how you're doing things?
Both. Our 28 nanometer, currently 80% of them are High-K Metal Gate, and 20% of them roughly about 28 LP. Next year, we'll compete on both fronts. We will compete on 28 LP, and we are planning to increase the capacity on 28 LP also. That will render the next year's 28 nanometer will be bigger than this year. Competition will always be there, of course. This is already from 2011. This is the fifth year of our 28 nanometer production. The learning curve bring us to a very mature state that all the cost reduction and many of the yield improvement will be our competitive advantage, and also increase our margins.
All right. I think we'll now take our next question from the call. Operator, please proceed with the first caller on the line. Thank you. Your next question comes from the line of Donald Lu from Goldman Sachs. Please ask your question.
Good afternoon. My first question is still on smartphone. Can you let us know what % of revenue today in Q4, for example, is from smartphone? Also, what would be the TSMC's addressable market per smartphone, i.e. on average, how much revenue TSMC can generate on average per smartphone this year and next year?
Okay. Donald Lu, let me repeat your question. Your question is how much % of revenue TSMC derives from smartphone-related applications. Your second question is, in terms of revenue per smartphone box, what is TSMC's revenue per smartphone on average?
Yeah.
Donald Lu, we actually don't count on quarterly basis % of smartphone. I can tell you, I can see for the whole year 2014, there will be a little bit more than 50% of our revenue coming from a smartphone.
Revenue per box?
Revenue per box, the average $8 this year, which is the improvement from last year's $7.
Okay. How about the trend for next year? Do we have any estimates?
It's probably too premature to talk about next year, but I would think the percentage will be very similar to this year.
My second question is more on structural profitability. Given all the moving parts, depreciation trend, et cetera, for next year, can TSMC still maintain a similar kind of structural profitability with the 16 nanometer ramp pulling slightly?
We continue work on productivity and cost reduction. From what we can see now, we believe we can maintain the same level of structural profitability next year versus this year.
Great. Thank you.
Now we are coming back to the floor. Our next question comes from the floor from Credit Suisse, Randy Abrams.
Okay. Yes, thank you. I want to ask one follow-up on the 16. If you could talk about what you expect the revenue contribution now, given you're pulling it in earlier, and do you still see any material swing factors that could swing that ramp up at this stage?
Excuse me, Randy, you are saying the 16 FinFET?
For 16 FinFET for next year, when you ramp it up in second half, % of revenue, and also if there's any material swing factors up or down that could swing the magnitude of that ramp up.
Probably too early to comment on the revenue side. I would think that, give you a hint, the ramp-up will be a little bit faster than 20SoC. We have a customer committed a high volume product tape out already.
Congratulations. Okay. My follow-up question, actually just more looking at your outlook. Given the comments you made on inventory, I think you put the chart as not quite as lean coming out of fourth quarter. It's still coming down in Q4. Given you have a strong ramp-up in 2020, should we expect as we go into first quarter, one, if you think you'll use first quarter to build product again? Like if we have a lower utilization, you may build product early through first quarter. Also, if you expect a little bit of a gap or slowdown coming out of fourth quarter.
I would prefer not going into the first quarter next year. As I just show you, the inventory cycle is two days below seasonality, it's going to be a mild correction, not as severe as last year. That should give you some idea about the first quarter.
Thank you.
All right. Next question also comes from the floor. It will be from Morgan Stanley's Bill Lu.
Hi. Thank you. Can I just start with a housekeeping question? That inventory chart that you showed, can I just ask you if it includes, you've got a big system house customer that we cannot track, but perhaps you can. Does it include the system house?
That's our fabless customer. The chart shows fabless.
Fabless.
Yeah.
Okay, great. The second question is a comment that I think Mark and Lora have both talked about, which is smartphone is going to drive growth for multiple more years. I think that's a little bit different from what is the consensus view, which is that you're going to get unit growth, but a lot of it going forward is going to be more lower end. Lora also said just now that percentage-wise, smartphone is going to be about the same next year. Can you talk about why you think it's going to drive growth a few more years? What do you see that gives you that confidence?
Well, yes. Smartphone revenue next year appears flattened out total. We see next year is still going to be an increase. Our growth are propelling what TSMC can grow is mainly from the market share in that smartphone segment. That we think even the 20 nanometer going forward, and 16 and 10, we think our market share in the smartphone will continue to grow.
Sorry, I thought Lora's comment was that percentage of total revenue, smartphone is not going up next year, right? Versus last couple of years, it's been increasing. Next year, smartphone is not going to outgrow the overall company.
I thought Lora said is giving up actually is that it'll be an increase.
All right. Thanks.
Yeah, you forget that we're going to grow next year, the company.
Sure. No, I understand that.
It's a growth above the average.
Yes. All right. Next question comes from the floor. It will be from Citi's Roland Shu.
Hi, good afternoon. I think my first question is still 20 nanometer gross margin. Lora, if I have read you right, you said the increasing volume after you ramp up new technology actually will be a key factor for bringing up the gross margin overall. Also, I look at in the past, you said it takes about seven to eight quarters to bring up the new technology gross margin to corporate level. Actually, seven to eight quarters also was around the time you ramp up the total revenue to above 20%, probably 20%, 23%, 25%. My question is, for your 20 nanometer, in 4Q, I think that is on the revenue now, that is above 20%. Is this 20% actually a big threshold level for you to bring up the gross margin?
I think with this, the 20% total revenue contribution, I think the volume is big enough. Maybe there's another factor to have a low margin for 20 nanometer.
Actually, this is a unique way of associating with the percentage revenue to the gross margin. You talk about a 20 nanometer as a separate node. When we look at 20 and 16, we feel it's the same node because 16 is a continuation of 20 nanometer, and they share a lot of the same equipment and same process, almost. If you combine the two nodes together and, well, I cannot say if 20% is threshold or not. What we look in the company is we look into how efficient can we run this node and what will be the scale, and what's our engagement with the customer, and how effectively we bring down the capital efficiency, bring up the capital efficiency, so on and so forth. We don't usually link that to a percentage of contribution to company revenue.
Yeah. For the volume-wise, actually, I think it is big enough now. Is it possible to 20 nanometer, actually, to bring up the 20 nanometer gross margin to corporate level less than seven to eight quarters, maybe three quarters, four quarters?
It's not possible because the 20 nanometer will be very soon migrate to 16. 16 will take the momentum from 20. It's going to run very fast, not only on the volume, also on the profitability side as well.
Okay. Little bit complicated to me. Okay, I'll switch to second question. Second question is, I think TSMC actually delivered a very good job. I think from 2010 to 2014, you have two strategic financial goals. One is for your PBT to grow more than 10% in CAGR. The other one is the ROE to be above 20%. I think from 2010 to 2014, I think TSMC has been out. That is outperforming above your goal. How about the expectation for next five years, 2015 to 2018 or 2019? Are you still comfortable with this double digits in PBT growth going forward? Thank you.
We believe we'll still grow faster than the semiconductor, and actually, in the past few years, we've been growing 2x of semiconductor. We continue to gain from the market share in the next five years. That's what we believe. Our financial objective will remain unchanged. PBT CAGR bigger equal to 10% and ROE bigger than 20% from 2015 to 2019.
Thank you.
All right. Let's go back to the call. We'll take our next question from the call. Operator, please proceed with the next caller.
Thank you. Your next question comes from the line of Brett Simpson from Arete Research. Please ask your questions.
Yes. Thanks very much. This is a question for Mark or CC. I'm just looking at the end market focus for TSMC. PC has never been a big part of your business. When I look at AP for next year, the transition to 64-bit and the introduction of FinFET, is the AP moving into a world where it can realistically address mainstream PCs now that we've got Windows 10 coming and we have Chromebook ramping up and Android is moving to 64-bit? How do you see this AP evolving into mainstream computing? Maybe just a follow-up to that, can you maybe just compare and contrast if you're an Arm fabless chipmaker building these APs for computing, how would you compare the cost and the performance using 64-bit Arm and your FinFET plus versus Intel's Broadwell?
Do you think it would be cost competitive versus Intel's Broadwell, and do you think it'll be performance competitive versus Intel Broadwell? Thank you.
All right, Brett, let me repeat your question. You basically are asking us whether or not the migration of technologies and then with the arrival of Arm-based 64-bit core, the fabless today can design application processors that can going into the mainstream PC market. If they could, how would they compare with the existing incumbent players such as Intel in terms of cost and performance? That's your question, right?
That's right, Elizabeth. It's really comparing next year's AP to Intel Broadwell, whether you think the new Broadwell platform from Intel, which is 14 nanometer, how might we see fabless players running TSMC FinFET Plus and 64-bit Arm versus those platforms from Intel from cost and performance perspective?
You are specific to 2015. You are asking 2015.
That's right.
Okay.
Thank you.
Well, we certainly hope the ARM-based core can get into PC faster. However, the recent trend seems to be slowing down, and we do not count on that. Our customers are continuing to make this ARM-based core into a mainstream PC application. I think it's more than power and the performance of those chips. There are lots to do with the ecosystem around the x86 core. This is up to our customers and their customers, how we together get into this. Definitely, this industry needs alternative in the PC world.
Great. Maybe just a follow-up for Lora. If I look on the balance sheet at TSMC, there's a large amount of construction in progress. Capital has been spent but is not effective capacity. Can you maybe just give us a sense for how this trends over the next three or four quarters, and how depreciation might trend as well on the back of that? Thank you.
Your question is looking at our balance sheets with a very large item called construction in progress, which of course eventually will become capacity, and you want to know what is the trend of leading from this large account in construction in progress into the impact of our depreciation?
That's right. Thanks.
Okay. We are building our new facilities to run the 20 nanometer and 16 nanometer, which will be located in our Tainan site. The number you see in the balance sheet is associated with those building constructions, and some of them are equipment purchase. From what I can tell you, the depreciation change, and this year, with the TWD 9.6 billion CapEx, the depreciation year-over-year change will be roughly 30%, which has actually been lower. I was telling somebody 35% earlier. Now it's come out to about 30%. With TWD 30 more CapEx for next year, as I was talking about, the depreciation increase will be much smaller than this year versus last year. It will be in the mid-teen range, certainly below 20% increase year-over-year.
Thank you very much.
Okay, now let's come back to the floor, Andrew is eagerly anticipating me. We'll have Andrew from Barclays.
Thanks, Elizabeth Sun. Hello to both CEOs and Lora. Today I'd like to ask in Mandarin. I'd like to ask, when you mentioned earlier about a faster ramp, do you mean that next year's Q3 and Q4 revenue from 16 nanometer will be bigger than this year's Q3 and Q4 revenue from 20 nanometer?
Okay, I have to translate your questions into English. Andrew's question is: Given there will be a faster ramp of 16 nanometer, does this mean that next year's Q3 and Q4 revenue from 16 nanometer will be bigger than this year's Q3 and Q4 revenue from 20 nanometer?
No.
You have to quickly define
Actually, if you look at what we announced early this year, we said that the 20SoC is in production. That means we start with our wafer production. The significant shipment actually is in the third quarter. You know that for this kind of cycle time, it will run through the line, and then get in the packaging, and then get the revenue. 16 nanometer fab actually is longer because they have more masking there. When we start at the end of the second quarter, you can estimate that what will be the volume shipment as translated into revenue. That's what I can.
Is it just delay for a quarter or what? Are you saying that 10 nm will become next year's Q4 and then next year's Q1
About two quarter late.
Huh?
Two quarter difference.
The lead time or the shipment.
One and a half quarter, maybe, to be exact. All right?
To be exact, you're saying that next year's revenue contribution is still quite small.
In fourth quarter, we will see contribution.
Right. Last time you said that fourth quarter is single digits. Has that changed?
High single digits.
Okay. Thank you. That's all I want. The other question. This time using English only. For 10 nanometer ramp-up, you mentioned will be starting from year 2012 to 2016. Is that similar timeframe to ramp up like 16 nanometer ramp up next year, or further delay another one or two quarters to generate revenue?
Yeah. Okay. I said we will try to enable our customers to be able to ramp up in 2016. I think it's toward the end of 2016.
Really, it's still up to the customer's ramp. I think in terms of revenue, it will be much lower than 16 nanometer in 2015. The real volume will happen beginning of 2017 and on.
Thank you.
Next question also comes from Steven Pelayo.
When I look back at 28 nanometer, you had seven quarters of solid absolute dollar growth sequentially. Now when I look at 20 nanometer, you're already at 20% of revenues in two quarters, such a steep ramp. Do you suspect that 20 nanometer in absolute dollars on a quarterly basis can continue to grow every quarter like the first seven quarters that you saw in 28 nanometer?
Yes. We still see the 20 nanometers continue to grow next year.
Even quarterly in the first quarter, second quarter with seasonality, you still will see sequential-
Yes
dollar and dollar growth in 20-
You might have some kind of seasonality, on the average, yes, still growing.
I understand, yeah. Okay. Then, Lora, a question for you. I know the board is going to talk about a dividend next year in February. You have about TWD 300 billion in cash, you're generating about TWD 40 billion a quarter. How much cash do you need to run this business? We're trying to all figure out what kind of magnitude. Could you afford TWD 4? Could you afford more even? It seems like you could. Maybe you could talk a little bit about what are the inputs that go into that decision on where you would like to take the dividend.
I think the key input is the sustainability of free cash flow generation. Consider the potential CapEx for the future years and operating cash flow we can generate with that kind of business growth. That's the key decision factor. I probably cannot tell you what's the magnitude because we need to discuss with the board. As I said earlier in my comments, we feel that we are affordable to raise the dividend level, and the board will decide the numbers.
I'm sorry, if I could just sneak in a follow-up to that first question to you. If you do have some seasonality in 20 nanometer in the first half next year, do you worry at all about the margin implications there if you're not fully utilized on that very expensive capacity?
Very hard to answer your question. Although I say seasonality, I don't expect too much of a drop, if there's a drop. You can see the smartphone selling very well, right? I did not say which one. I'll let Lora to answer this question because you know
It's too early to give a guidance on margins.
Yes. We think the seasonality Since last year, we have this mechanism of a pre-built, work with our customers, right, to smooth out the utilization. For the 20 nanometer, we intend to do that. I think that in terms of product complexity, it's much simpler, which we should be able to minimize the impact.
All right. Next question comes from the floor. It will be from UBS, Eric Chen.
Hi. My first question regarding to the 20 and the 16 FinFET, the technology. You mentioned the equipment in between the both FinFET are pretty similar. I assume the yield rate at the 15, sorry, 16 FinFET, even 16FinFET+, is pretty high. I'm just wondering why our equipment move-in is so conservative compared to our competitor. I mean, compared to our competitor at the 14, the FinFET technology in term of equipment move-in.
Eric's question is, he assumed that we will have a very good yield rate on the 16 nanometer FinFET. I think that's correct.
Yeah.
Given that, why are we so conservative in equipment move-in?
All circle, like Samsung, GlobalFoundries, they mentioned they have like 50,000 wafer at the end of this year.
Let me answer your question. First, your impression of we slowly move in 16 FinFET equipment. I think Lora just mentioned that we are going to spend also a big CapEx next year. From this year, next year, we invest on the 16 FinFET and 20SoC also. Some of the tools, actually a high portion of the tool are common for these two node. No, it's not slowly moving, as you said. Yes, we did ramp up the 16 FinFET behind our competitor. Yes. That's why Chairman say that we are going to have a smaller market share. Our situation improved as time goes by because of our manufacturing maturity and the good yield performance and also the customer pull-in their demand. We decide to pull in the ramp-up schedule.
We are In a very high gung ho to bring up the 16 FinFET. I said we have more than 1,000 engineer preparing for the ramp up. That give you a hint that we are full speed, actually, preparing for that.
Okay. We believe you are going to be very aggressive under the 16, the FinFET capacity for next year. Your competitor right now is very aggressive, probably will keep aggressive. Internally, the TSMC do have the scenario, the analysis say the 16 process of FinFET technology, in term of the capacity wise, probably will over capacity in the year.
16.
16, I'm sorry. Year 2016. The over capacity and the 14-
Over capacity?
We build the capacity with the customer demand.
We have a confidence, as I said, we already have customer committed high volume tape outs to us.
Okay.
We build the capacity according to the demand, and we have a confidence to gain large market share 2016.
The year will be the key, right?
Yes.
Also for Lora, regarding to 20 and the 16 nanometer FinFET. I made notes. The equipment is not big different. We assume that the depreciation pretty much allocated for the 20 nanometer process. In other words, can we expect the gross margin for 16, the process, in fact, the pick out time, the above average, probably will much faster than we thought.
That's the right assumption. Yes.
Okay. Let's say that if we talk about the second quarter for 20, can we assume that probably fourth quarter for the 16?
I don't know.
Okay.
We only look at the two nodes together.
Okay. My second question.
You already have two questions.
Okay.
Yes. Now we have to go to someone else. Sorry, we can come back to you.
Okay, sure. Thank you.
All right. Next will be from Dan Heyler.
Thank you. Just a few quick housekeepings. Lora, could you comment a little bit on the linearity of the CapEx next year in terms of first half versus second half? You normally share that with us.
Next year is not decided yet. Let's see.
I imagine front-end loaded.
Actually, I don't have number with me. I intuitively think it will be front-end loaded because we need to get a capacity ready for the 16 nanometer ramp.
Okay. kind of similar to this year then probably, right? In terms of the weighting.
This year is not that front-end loaded. It is slightly front-end loaded. I don't have a number with me.
Okay, great. Thanks. We will come back to this. On the capacity growth then, this year you grew by 12%. Could you share with us what kind of capacity growth we should expect next year? Because you did allude to some 8-inch expansion, and then you've also got, which I'm sure is quite small, but in addition to that, your 20 and 16 nanometer capacity growth. Thanks.
As I said, this year we will grow about 12% in total capacity. Next year, about the same.
Great. Is it fair to assume that very little 8-inch addition, that's more kind of peanuts, or could we see bigger 8-inch?
TWD couple hundred million. It's not a peanut.
Okay.
More 8-inch than last year, we'll see.
TWD two million?
Cannot be that specific.
TWD 200 million.
$200 million.
I guess what I wanted to think a little bit about the magnitude of the revenue trends in the fourth quarter, you're growing in low single digit. The non-20 nanometer capacity, how much is that? Is that growing or is that down in the fourth quarter in terms of the non-20?
Yeah. In take out 20, it's down. It's down, yeah. For our case, little bit more than seasonal because we are forecasting the DOI will be below.
It looks like down low single digit is what it looks like?
We don't know. Probably close to double digit decline.
Double digit decline for non-20.
For non-20, yeah.
Okay. Great. Then finally, just short No? She's cutting me off. Okay.
Right.
I'll come back.
Next one goes to, well, the question will be coming from Credit Suisse, Randy.
This one might be more for C.C. I just want to see if you could give an update on the InFO, and the fan-out wafer-level packaging, just your initiatives on doing some of the integration. Whether next year there's many applications or it'll take time to develop, would come more the following year. What type of applications you're seeing the first interest?
It takes time for the customers really to adopt the new packaging because that affected their design, architecture, and also everything. Application wise, it will be still in the mobile devices.
Your question is?
It's probably more the following year. Next year, probably not much revenue, but I guess to get to TWD hundreds of millions in revenue.
The significant volume will be in 2016. We already work with the customer on that.
Okay. Then one quick follow-up on China. There was some discussion earlier about China, and now you have competitor foundries putting up fab there. Given a lot of attention on local manufacturing and even the Qualcomm going SMIC to have local China manufacturing, does it make sense at some point to consider a fab in China? Are there restrictions, or are those removed if you wanted to have a 12-inch fab in China?
We have the fab in China today. It's eight-inch, and we are expanding that capacity also. This actually is today and the following months, and we have space next to that fab also. When the demand is needed, when the option is the most cost effective, that's one of the consideration. As to 12-inch, no. I think so long as N-2 and below, I think at least in the local government, restriction is not there.
Thank you.
Now we are going back to Eric for his follow-up questions.
Elizabeth, thank you very much. Regarding to the 10 nano FinFET, ASML have a conference call last night, Intel regularly, they mentioned that TSMC is quite a leading player. I just would like to know what's your strategy? I mean, your equipment moving in '16, probably not so aggressive, but your strategy of the 10 nano is so aggressive. How you look at the future outlook?
On EUV, you mean, the question?
In term of the 10 nano.
Our current 10 nanometer, it does not using EUV. All the technology developing for 10 is still using multiple patterning technology.
Okay. I saw you use two, right? EUV and the one you mentioned.
We are working with ASML to develop EUV tools.
The opportunity is to have a follow-on process simplification using EUV, the masking layer simplification. That's some way to go. Could not catch our initial ramp of 10 nanometer. The current challenge is still the power of the source and the availability. It's still very encouraging that source power is continually increasing, and other factors as mask defects and mask technology and the photoresist have a lot of improvements. We are looking at the follow-on insertion point after the 10-nanometer ramp for the cost reduction or for the process simplification.
Oh, okay. The way you go for the 10 nano FinFET is the same as the Intel way or different way on the Intel?
In terms of the not using EUV or in terms of-
Yes, in term of-
Right. Intel appeared to announce they don't use EUV on their 10.
It's a different way, right? I'm sorry.
I don't know Intel technology that well.
Okay. The reason I ask that, there is one argument talk about the 14 nano FinFET is the way Intel to go and Samsung go. Once they go to the 10 nano, if they go the different way, probably we will see the two different way in the 10 nano FinFET and for the client to choose. That probably will bring the risk on the either side. I don't know this argument makes sense from your point of view. Does that make sense?
That's our plan. I don't know what Intel's detail plan is, and they haven't announced their 10 nanometer details either.
Okay. I see.
We just make the best of what into our most beneficial way to implement the EUV.
Okay. Very clear. Thank you, Dr. Liu.
Follow-on questions from Andrew Lu.
Very quick two questions. First one is, 28-nanometer high-k metal gate versus the polysilicon percentage on 28 in Q3 and Q4, any roughly numbers?
Probably 70 to 30.
For both quarter?
That's good. In Q3 is probably more than 80% are high-k metal gate, less than 20% are polysilicon. Getting to Q4, polysilicon has a big increase from our demand. It's getting to
Temporarily get to 70/30 in the Q4, as C.C. just mentioned.
Because of customer, right?
Because of customers.
Okay. The next one is, I remember last time Lora mentioned the CapEx of sales for next year is going to be higher than this year. This time changed the tone, saying the CapEx next year will be slightly higher than this year. I just want to ask whether the CapEx itself is also higher than this year for next year?
I don't think I said last time that next year will be higher than this year. I didn't say that. I probably said will be lower than last.
say that.
It'll be lower than last year. That's probably I have said.
Okay.
Given the total CapEx, as I just indicated, will be slightly higher than TWD 10 billion, but we have not decided how much it's going to be. We don't decide the revenue of 2015, so it's probably too mature to comment on the capital intensity for next year.
Thank you.
All right. Follow-on questions from Daiwa's Rick Hsu.
Hi, thank you so much. It's Rick Hsu from Daiwa. Just a quick follow-up question, sometimes Dr. Morris Chang gives some color for beyond one quarter. Would you give a little bit color for your Q1 outlook? We'll talk about inventory correction. This year is very mild and demand is still pretty strong, especially for these edges, has been in good shape. Just some color for next year, Q1.
Well, Q1 normally is a low season. You will see sequential decline for the experience we had before. I cannot comment more on that. I don't think Q1 will be very different from that pattern.
Let me add some color for that, if you want.
Yeah, sure. Of course.
We don't expect strong inventory adjustment more than what we see in Q4. Okay? I think this is what we see from our demand forecasting.
Follow-up question from HSBC, Steven Pelayo.
Just a little bit, thinking about the next three years or so. The last three years, you guys have clearly benefited 28 nanometer dominance, smartphone industry growing at 40% compounded growth rate. You guys have grown 20, 25% per year, I think, in the last three years or so, and I got to do the calculation again for this year. Maybe you're actually above that. When you think about the next three years, I think people are forecasting smartphone CAGRs maybe more in the under 15% type range. Do you think, and you guys gained a lot of market share, I think, in the last couple of years as well. Are laws of large numbers catching up to you guys?
Without smartphone growth and that significant increase in silicon content per phone, do you think these growth rates that you've been enjoying, this 20% plus for multiple years, are headed lower over the next few years? To what magnitude?
You cannot expect 20% growth for every year. We are confident we can grow double digits next two years.
Short follow-up questions from Dan Heyler.
That last one that you let me sneak in there. I didn't forget it. On the 20 nanometer, coming back to that, the fourth quarter contribution for this year on this percentage of sales. Will 20 grow one year later? Fourth quarter next year, what contribution to sales do you expect 20 per se to be? 20 nanometer next year. Because you've obviously pulled in 16 nanometer, there's new dynamics. I'm wondering if you're going to see any growth year-on-year from 4Q this year to 4Q next year. Will the TWD value grow year-on-year?
Good question. I think next year when we ramp up our 16, 20SoC, the demand will be flat, and the percentage will be lower.
Right.
Okay. Does that answer your question?
Yes, it does. Thank you.
Okay. I think pretty much we have answered the most critical questions in your mind, and we would like to conclude our conference and conference call right now, and thank you for attending our session today. Before we conclude, please be advised the replay will be available within three hours from now. Transcripts will be available within 24 hours from now, both of which will be available through our website at www.tsmc.com. Thank you for joining us today. We hope you will join us again next quarter. Goodbye and have a good day.