[Foreign language] Welcome to TSMC's third quarter 2015 earnings conference and 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 Senior Vice President and CFO, Ms. Lora Ho, will summarize our operations in the third quarter, followed by the guidance for the fourth quarter. Afterwards, TSMC's two Co-CEOs, Dr. Mark Liu and Dr. C.C. Wei, and CFO Lora Ho, 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.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 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 summary of operations and current quarter guidance.
Thank you, Elizabeth. Good afternoon, everyone. Welcome to join us today. I will start with the financial summary for the third quarter, followed by the guidance of the fourth quarter. In the third quarter, demand for TSMC's wafer was essentially flat with the second quarter, mainly reflecting customers' cautious inventory management. However, a stronger-than-forecasted U.S. dollars against the TWD caused our third quarter revenue to exceed the high end of our guidance given in July. On a sequential basis, revenue increased by 3.4% to TWD 213 billion, gross margin decreased 0.3 percentage point to 48.2%, as the benefit from higher capacity utilization and the favorable exchange rate were outweighed by unfavorable inventory management and the margin dilution from 16 nm. During this quarter, we ceased TSMC Solar operations and incurred a loss of TWD 2.8 billion.
Among the total, about TWD 400 million was recorded in cost of goods sold, and TWD 2.4 billion in other operating expenses. The loss dragged down our operating margin by 1.3 percentage point and reduced our EPS by TWD 0.08. Operating margin in the third quarter decreased 0.6 percentage point to 36.9%. Without this one-time solar write-off, our operating margin would have been 38.2%. For non-operating items, ASML shares disposal gains were TWD 3.7 billion, or TWD 0.13 in EPS. For comparison, in the second quarter, we recorded TWD 20 billion disposal gains on ASML and Vanguard shares. With all these items, our third quarter EPS was TWD 2.91. Let's take a look at revenue by application. During the third quarter, communication, computer, consumer, and industrial increased 1%, 15%, 3%, and 11% respectively. This number reflects the change in favorable foreign exchange rate already.
In terms of revenue by technology, we are pleased to report the first quarter of 16 nm volume shipments. 16 nm and 20 nm combined contributed 21% of our total wafer revenue in the third quarter. Moving to the balance sheet. We ended the third quarter with cash and marketable securities of TWD 525 billion. Current liabilities decreased by TWD 108 billion, mainly as we paid out TWD 117 billion of cash dividend in July. On financial ratios, accounts receivable turnover days decreased two days to 42 days. Days of inventory decreased by three days to 59 days. As we now have lower days in work-in- process inventories, and we shipped the 20 nm wafer we previewed last quarter. Lastly, I would like to make a few comments on cash flow and CapEx.
During the third quarter, we generated TWD 118 billion cash from operations, invested TWD 70 billion in capital expenditure, and paid out TWD 117 billion cash dividend. Additionally, we received about TWD 15 billion from disposal of ASML shares and borrowed TWD 28 billion in short-term loans for currency hedging purpose. As a result, our cash balance decreased TWD 13 billion to TWD 516 billion at the end of the third quarter. I just finished the summary of the third quarter financial outcome. Let me turn to the fourth quarter outlook. Due to the weaker than expected end market demand, customer continue to manage inventory cautiously. Meanwhile, 16 nm will ramp strongly and contribute more significant revenue in the fourth quarter.
Based on our current business outlook and exchange rate assumptions of 1 U.S. dollar to TWD 32.71, we expect our fourth quarter revenue to be between TWD 201 billion and TWD 204 billion. Gross profit margin to be between 47.5%-49.5%, and operating margin to be between 36.5%-38.5%. You may notice that the midpoint of gross margin guidance is slightly higher than third quarter gross margin, despite a lower revenue. This is because we expect our cost improvement and favorable inventory valuation adjustment to fully offset the impact from lower capacity utilization. In addition, the expiration of the remaining hedging contract of ASML will contribute about TWD 0.01 per share in the fourth quarter, and that will be end of the year total accounting treatment. This concludes my remarks. Let me turn the podium to Co-CEO Mark Liu for his comment.
Good afternoon, everyone. I'd like to give you some remarks, some messages here. My message has three parts. First, I'll talk about near-term demand. I'll describe the background color, I'll talk about TSMC's long-term growth drivers. The third part is leading-edge technology development update. First, the near-term demand. This year, due to a weaker global economy, a stronger US dollar environment, and a volatile financial market, the electronic device market has been negatively impacted, resulting a lack of growth in the overall semiconductor market. In addition, we see the unexpected slowdown of the economy in China since the Q1, resulting a continued sluggish smartphone demand in China. This led to our relatively flat revenue growth in the third quarter. We think the inventory level in the fabless industry will be still above seasonal normal by about 10 days at the end of the third quarter.
Looking forward, we see the active inventory reduction actions in the fabless companies continue into the fourth quarter. Nonetheless, our accelerated ramp-up of 16nm FinFET Plus technology supporting high-end smartphone market did uplift the otherwise weak wafer demand. We forecast a moderate decline, about 4%-5% of revenue from the previous quarter. We estimated the fabless industry inventory will likely settle to a seasonal level towards the end of this year. Albeit some uncertainty on this still exists. Given the fourth quarter guidance, TSMC should be able to deliver a double-digit growth at about 10%-11% year-over-year in 2015. Thanks partially to the stronger US dollar. Our 2015 forecast of the growth of smartphone unit shipment is 10%. PC for the industry is -6%. Tablet, -14%. Digital consumer electronics, -6%.
The semiconductor industry growth is about 0%, our fabless industry growth is -5%. The next part, I will talk about TSMC's long-term growth driver. In the past three to four years, the growth of smartphone market has propelled TSMC's growth. Recently, we see the total unit growth of smartphone appears slowing. The silicon content in all segments of smartphone continues to increase. This silicon content increase particularly shows in a high-end smartphone. For example, the unit growth of high-end smartphone this year will be about flat, the silicon content of the high-end smartphone will still have a mid-teen percentage growth. We see the bifurcation of high-end and mid-low-end smartphone markets continues, the high-end smartphones, which provide richer features, higher performance, and lower power consumption, will continue to drive the demand of leading-edge technologies. We stand to benefit from this trend.
Smartphone will continue to provide growth momentum for TSMC in the next two to three years. TSMC holds our corporate mission of being the trusted technology and capacity supplier in the logic semiconductor industry for years to come. We work closely with many of our partners and collaborate with them to produce the best product in many market segments, each with differentiations of its own. From that, we see several drivers of future demand growth. It is very encouraging. Three areas that we will soon add to the growth of TSMC in the coming years. First, computing market. In addition to the smartphones, we see faster growth of processor unit demand in many applications. With the internet traffic near exponentially increasing, people are to extract intelligence out of the mountains of data.
Extensive computation not only needs to be done in the data center, much must be done in a network, as well as the distributed edge of the network. Image processor, CPU for automotive, CPU and GPU for augmented reality and virtual reality are examples. Second, fabless system company. We now see a new breed of fabless system companies coming to join the play of semiconductor product innovation. With the enhanced SoC integration and system software and hardware integration, new usage models with brand-new markets are opening up for the semiconductor industry. We have been working with these fabless system companies for several years. Third, Internet of Things. As billions of things are connected to Internet, with sophisticated computing and analysis, they soon will become intelligent enough to create new user experiences and to improve efficiency in many aspects of life, work, and leisure.
Recently, we see fast innovation taking place in areas such as cars, drones, robots, wearables, and smartphone devices. Working with innovators around the world, we continue to see the insatiable need of performance of leading-edge technologies. We also see the need of advanced packaging technology to increase the overall system performance. Thus, we will continue to increase the pace of our leading-edge technology development. Leading-edge technology will continue to be the major driver of TSMC's future growth. The third part, I'd like to update you about our leading-edge technology development. Our 10-nm technology development is well on track. This technology has a logic density of 2.1 times of its previous generation. That's 16 FinFET Plus, with performance of 20% enhancement or a power consumption reduction of 40%. During this quarter, we will freeze the process and begin technology qualification for our 10-nm technology.
Customer product tape-outs will soon begin in next spring. TSMC's 7-nm technology will fully leverage of our 10-nm EUV learning. The progress of our 7-nm is well on track as well. The migration from 10-nm to 7-nm provides substantial improvement in performance, power, and density. We are very happy to inform you that fully functional SRAMs on our 7-nm have already been demonstrated. On advanced packaging development, our InFO technology will enter high volume production with our 16-nm technology next year. We are currently working on the second-generation InFO technology for several projects of system integration on 10-nm and 7-nm. Above is my key messages. Thank you for your attention. I'll turn the podium to C.C. Wei.
Good afternoon, ladies and gentlemen. I will update you the status of our fab operation on the following topics. First one, 28-nm. Due to customers expediting their inventory management, as Mark just pointed out, the demand for our 28-nm has been reduced the fourth quarter. As a result, 28-nm utilization rate came down from above 90% in third quarter to a level below 80% in the fourth quarter. However, the outlook for 28-nm remains very promising thanks to our newly developed 28 HPC and 28 HPC+ technologies. These two new technologies not only suit the mid to low-end smartphone requirement but also highly useful for other applications such as a Wi-Fi, digital TV, set-top box, and image signal processor, and et cetera.
Overall, we expect we will ship similar amount of 28 nm wafers this year as we did last year, and we expect to ship more 28 nm wafer next year. We have increased our market segment share in 28 nm this year. We believe our technology and cost advantage will enable us to compete well. Going forward, we expect to continue to protect our market segment share and gain a good profit. Next, on 20 nm. Of generally weaker demand, as Mark described, and several customer has accelerated their product migration to 16 FinFET Plus, the demand for 20 nm this year falls somewhat short of our estimate made at the beginning of this year. That said, we continue to stand by our early prediction that revenue from 20 nm this year will be at least double the level we have last year.
Given the demand outlook trends and the high ratio of common tools, we will continue to convert 20 nm capacity into 16 nm. Let me update you the 16 nm. We begin high volume production of 16 nm in third quarter as planned, and saw revenue contribution from 16 nm become better than we expected earlier. 16 nm yield improvement is progressing very well, setting a new record internally for ramping up a new technology node. In addition to 16 FinFET Plus, we are developing 16 FinFET C for the low application. I'm sorry. Today, we have already completed the first phase of 16FFC and obtained good result. Between 16FF+ and 16FFC, we will have around 100 product tape-outs from about 40 customers before the end of 2016, with very comprehensive portfolio including mobile, networking, CPU, FPGA, consumer, and GPUs.
We believe our 16 nm portfolio, including 16FF+ and 16FFC are very competitive. We anticipate continued ramp-up into 2016 as multiple customers in both mobile and consumer applications will drive the production in parallel. Similar to 28 nm, we believe 16 FinFET will become a long node supporting multiple high-volume market applications. TSMC's technology advantage should allow us to capture a large majority of this demand. Let me update InFO. We have completed the construction of the new facility in Longtan and ready for InFO's volume production. The manufacturing equipment move in is on schedule, and also we target volume ramp-up at second quarter next year. Compared to existing package scheme, TSMC say InFO can bring greater than 20% reduction in overall package thickness, 20% speed gain in performance, and 10% better in thermal performance for power dissipation.
InFO technology is capable and well-positioned to enable next generation mobile applications. Right now, we continue to work with major customers on completion of their product qualification. Meanwhile, we are developing the next generation InFO process, as Mark just said, for the future application. Our expectation of InFO contributing more than $100 million quarterly revenue by 4Q 2016 remains unchanged. Let me update on specialty and eight-inch fab. In order to meet our customers' requirement on the more and more demand, we have developed specialty technology based on our logical processes. We have been doing this for more than 15 years by now. Applications such as MEMS, power management, CMOS image sensor, high-resolution display driver, and high-precision analog have been the main drivers for the specialty technologies.
After these years of effort, demand for these specialty technologies has grown substantially and account for more than 70% of all the eight-inch fab business. Some of the specialty applications have also moved to 12-inch fab, of course. We expect this trend will continue, and we are preparing sufficient capacity both in eight-inch and 12-inch fabs to address the need for specialty technology. Thank you for your attention. Now I turn the podium to Lora.
I have a few comments, let me start with 2015 CapEx. Six months ago, we gave our 2015 CapEx guidance in range between $10.5 billion-$11 billion. Due to a combination of factors including efficiency gains, capital investment program changes, capital deployment schedule changes, and foreign exchange rate changes, we now expect our 2015 capital budget to be about $8 billion. Of these changes, about 33% or one-third of the reduction of CapEx is due to operating efficiency gains. That led us to spend less money but still have the same output. Another one, about 30% of the reduction, is due to changes in the investment project, including the conversion between 20nm and 16 nm. Another one, about 20% of the reduction, is due to changes in capacity schedule. Lastly, the remainder, about 17% of the total reduction, is due to the strengthening of U.S.
dollars against euro, Japanese yen, and NT dollars. We expect our 2016 capital budget will be higher than this year. Next one is regarding our structural profitability. In the last few years, we have been able to maintain or improve our structural profitability despite making heavy capital investments. This is mainly due to three reasons. Number 1, operation innovations leading to productivity efficiency and better asset effectiveness. Number 2, very careful planning and build-out of capacity. Number 3, fast yield learning when ramping new nodes. All these three factors allow TSMC to enjoy a large cost advantage, which leads to a better structural profitability. TSMC's cost reduction efforts are as intense as our R&D efforts. A result of our cost reduction, we have been able to maintain our structural profitability in spite of price pressure, and we expect to continue to do so in the foreseeable future.
To talk about our solar operations. There have been a lot of changes in the world solar industry since we started our solar operations six years ago. Capacity overbuild in the solar module industry, coupled with the substantial price decline, has made TSMC Solar's future highly doubtful. Despite hard work at TSMC Solar throughout this six-year period, we have decided to terminate its operations at the end of August this year. Owing to this change, we have evaluated financial impact across all accounts. The total impact is about TWD 2.75 billion, which reduced our third quarter operating margin rate by about 1.3 percentage points and reduced our third quarter EPS by about TWD 0.08. TSMC will continue to honor all product warranties that have been offered to existing customers, and we also extended employment offers to all employees working at TSMC Solar in Taiwan at the time of the closing.
My last comment is regarding China investment. In the last several months, we have been actively evaluating the potential investment of a 12-inch fab in China. The consideration includes the positive development of semiconductor market in China and a large pool of engineering talent. On the other hand, the manufacturing cost will be higher in China than in Taiwan due to lack of economy of scale and many other reasons. As of now, we are still evaluating the potential investment. This ends my remarks. Thank you.
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 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'd like to ask a question, please press the star then one on your telephone keypad now. Star and one. Questions will be taken in the order in which they are received. If at any time you would like to remove yourself from the questioning queue, please press the pound or the hash key. Let's begin the Q&A session.
Our first question comes from the floor, it will be from Deutsche Bank, Michael Chou.
Thank you. Two questions. First question is, regarding your 16 nm FFC, would that enter mass production in the second half next year or 2017? That's the first question. Thank you.
To repeat your question, you say that it will be in the second half of next year and enter into.
2017.
2017. Our initial schedule actually say in 2017, but we might pull in because right now we have two versions of 16nm. One is shrink, and the other one is non-shrink.
Second question is, regarding your InFO, will customer use 2D or 3D in the future? It's 2D now? Sorry.
Right now, it's 2D.
Yeah.
In the future, of course, we are going to the 3D dimension.
Do you think the 3D yield rate will be similar to 2D yield rate?
We are working on it.
Okay. Thank you.
All right. Next, it will be coming from the floor again. That will be from Credit Suisse, Randy Abrams.
Thank you. My first question on 16 nm. Some of the third-party benchmarks are showing good performance, better battery life than your competitor. I'm curious if this could trigger some incremental strength for your 16, how you see the ramp-up now on 16 relative to 20, where it ramped to 20% of revenue pretty quickly.
I think Randy's question is whether or not the reported differences in the chip will lead us to have more demand.
We are very confident on our technology, of course. If you say what they reported in the newspaper, I would say that our customer already make an announcement that there is a very minimal difference between TSMC and our competitors, and in the normal usage condition. We respect that their analysis.
Okay. If I could ask a follow-up just on 16 and on the second question. How are you seeing now, then, the ramp of 16 or broadening out to additional customers? Should we expect contribution and diversification of the customer base to grow quite a lot in the next couple quarters? Do we need to wait for 16FFC for that diversification?
We expect the ramp up and probably quicker than 20nm SoC. We develop FFC, certainly we hope that the 16 nm will become a major node. We expect that more business, of course.
My second question on looking ahead, because you talk about normal inventory by the end of the year. If you could talk the last four or five years, you've actually held flat sequentially Q4 to Q1. If we should think normal inventory, do we have restocking again in first quarter where we could have similar, the new post-crisis seasonality where things are stable? I guess if you can give a view how we look for demand and inventory into first quarter.
Well, the inventory in the fourth quarter, our estimate is close to seasonal normal. As you know, the inventory going into first quarter really depends on the burn rate in the first quarter. That estimate still is a three-month backward numbers. That's why I mentioned about some of the uncertainty at the end still exist. We certainly hope that Q1 next year will come to a normal quarter.
Okay.
All right. The next question will be also coming from the floor. It will be J.P. Morgan's Gokul Hariharan.
Hi. Thanks for taking my question. My first question is on the new opportunities that you mentioned on the computing side. Could you talk a little bit about what are the developments that you're seeing? Are we going to see the opportunity from the computing side, primarily on the data center side, coming through? Also maybe a quick stab at what you think the opportunity could be in the next, say, couple of years, on the computing side. I have a second question to follow up.
Let me repeat the question so that I make sure I answer you right. You mentioned about ask me to comment more on the computing side, whether the computing side is more on the data center or other areas.
On the client computing side itself.
Okay. Yes. I think TSMC has been supplying the processors, distributing processors in bigger shares. More recently, we see the processors growth increasing. Therefore, we think the computing market, we meant not just the data center, but also the distributed processors, namely application processor. If you look at the application processor, unit growth is much faster than smartphone growth. Microprocessors. Image processors, when you diagnose absorbed images, you need to process on-site. We still fall quite short about the capability of an automated car, for example. Network processor is also important. Of course, we include microprocessors and CPU definitely is one area. By CPU, I mean the Chromebook, that type of distributed processors. That areas we think we see growth, and we believe as the data continue increasing, the local processing capability has to increase tremendously. Yeah.
When you scope out this market, do you see this market probably exceeding what the current mobile application processor or mobile processor market is? Any kind of estimates that you can talk about within a bit more longer term horizon?
Well, it is hard to estimate because usage model, the service, those big data service, analytics services, are still yet to prevail to our lives. At this point it's not as big. I think it will take off before the smartphones growth stop.
Okay. Just my second question is on the CapEx intensity. I think previously Lora had mentioned that this come down to a 35% level in CapEx intensity. I think this year we are down to around 30%. Should we think about the TWD 8 billion CapEx this year as kind of a one-off, and then we go back to the normalized like TWD 9.5 billion, TWD 10 billion kind of levels or even higher that we saw in the last few years? Or just some initial color in terms of what kind of CapEx jump are you expecting next year as you support some of these newer opportunities as well?
I just mentioned, this year TWD 8 billion is kind of low, and next year expect to be higher than TWD 8 billion. Although it's too premature to give a precise number. Back to my earlier comment of the future CapEx intensity will be at a mid-30 level. That statement still holds. I think it will be within 30%-35% range.
Okay. Thank you.
All right. Next question will be coming from the floor. It will be Goldman Sachs, Donald Lu.
Hey. Ciao. Sorry, my first question is about CapEx. TSMC cut CapEx quite substantially in three months. Last time we had the guidance back in, I think, July. I think there are three reasons. I think the efficiency gain and conversion, those things doesn't seem like it's something you just find out in the last three months. Is there anything new that you realize that make you to change CapEx so much? I remember in the past, you always talk about CapEx today for production a year later. Do you become very cautious about something next year? Thank you.
CapEx management is a continued effort. As you say, it's just come out in one month or two months. We have seen the trend. In the last quarter, I mentioned about the CapEx, we still hold the original number, but there may be some changes. As the things evolved, we continue the effort to drive the CapEx efficiency, and we have gained a lot of experience doing this in the company. If you look at the reductions, actually it's more than half. It's truly effort. Efficiency gain, of course, with some help from foreign exchange rate. Some program change, including the migration and schedule change. That's response to the market. Some of them will be shipped to the next year.
Okay. My second question is on 16 nm ramp. Since you talk about it's going to be faster than 20 nm. Would that imply it will be over 22%, I believe, in Q1 next year of your total revenue?
Donald, our 16 FinFET, right now most of the volume come from one single customer. It will be very difficult for me to give you the exact number to say which, that it's a quarter percentage. Otherwise, I release too much of information.
Okay.
It will be faster.
It will be faster.
than 20 nm.
Yeah. That's what we say.
Okay. Thank you.
All right. Next it will be from Citigroup's Roland Shu.
Hi, good afternoon. I think it's unusual for TSMC have this within quarter pre-announcement, also give the guidance one month ahead of the earnings release. Just wondering, what have you seen by then, and how does it compare with the view now? Is it improving or is it stabilized or even deteriorating? Thank you.
Okay. Let me take this question. The announcement we did on September 23rd is a pre-alert. Actually, we have give a better guidance for third quarter, it's not really regarding third quarter. Being a transparent company, we like to be transparent. We kind of view the expectation for the rest of the year. Analysts expectation for the rest of the year may be higher than our forecast. We feel there's a need to come out and say something. That was the basis for the alert announcement.
Now for the full Q revenue guidance, low end is higher than what company guided a month ago. Does that mean that we are actually seeing more confidence on the demand side?
I recall the alert, we gave a higher range, a bigger range.
Yeah.
I was saying in the TWD 198 billion-TWD 204 billion. The range gets smaller, but toward the high-end side. As time goes by, we have more clear picture of what we are looking at in the fourth quarter.
For the margin forecast. I think in 3Q we have the margin impact by the inventory valuation adjustment and also, for this 16 nm ramp dilution. Are this going to be the key impact in 4Q or going forward? Is there any margin downside impact up on this one?
The inventory is related to the utilization changes. Not too much about how big this utilization changes. Where there's an upswing, there will be a negative inventory adjustment. Where there's a downstream, and there will be positive inventory adjustment. That has been like that for a long time. The other one is the 16 FinFET dilutions. Since we just started ramping, there will be some dilution. We expect the dilution will be smaller next year
About 1 percentage point to corporate average, and beyond that will be less than 1 percentage point. It will be smaller.
The whole year impact will be 1 percentage point?
For this year, it's 2-3 percentage points. Next year will be about 1 percentage point. After that, it will be smaller than 1 percentage point.
Okay. Thank you. My second question is for the 16 nm InFO. How many % of the 16 nm wafers is adopting InFO packaging next year?
That's a question pretty hard to estimate at this time, but I would say it's quite a portion of it.
Okay. By this definition, how are you going to build your InFO capacity accordingly?
Well, according to customers' demand, of course.
Thank you.
Understand. I think one last question. Even for the 16 nm, since the 16 nm is going to be a long node, same as the 28 nm. What is the ultimate capacity you are going to build for 16 nm?
It will be according to customers' demand again.
Thank you very much.
All right. Next, questions will be coming from the floor. It will be UBS, Eric Chen.
Okay. My first question, very quick, Lora, you talk about the CapEx cut, right? How about the capacity, the plan, and for the 8-inch, what kind of capacity expansion, and we should put on our model for this year, and how about the 28 nm process? Can you give us an update, more elaborate? Thank you.
I would talk about the company as a whole capacity plan. With the TWD 8 billion CapEx I was talking about, we expect the year-over-year capacity for the TSMC will increase by 12% combined. Mostly or majorly 12-inch.
All right. Okay. No idea for the 28?
I prefer not to disclose individual technological capacity.
Okay. Thank you. Appreciate it. Also the data way, I remember you talk about the 7nm fully leverage the 10nm, the equipment, right? Or the Dr. Liu [inaudible]. That's interesting point. When you talk about the full leverage, I mean the compatible, the equipment available to compatible the switch from the 10nm to 7nm based on your internals, the plan. Is that correct?
Eric, I think your question is asking whether or not there's a lot of common tools between 7 nm.
Yeah, right.
10 nm.
The way like between 20 and 16 nm.
Yes. We developed our 7 nm with low cost target. Therefore, when we develop this technology, we will make use of 10 nm equipment as much as possible. At this point, more than 90% of the 10 nm tool can be directly used by seven nm.
How about from the 16 to 10, how many %? Is that possible quantify?
From 16 to 10 is a lesser requirement.
Because 16 will be a long node.
Right.
Will be fully utilized.
Instead, the 10 nm will be a shorter node.
Therefore, we take particular focus on the tool migration efficiency.
Okay, that's interesting. For the 16nm, how you judge that's a long node? If we look at the schedule, probably around the five quarter, the switch from the 16nm to 10nm and from the 10nm to 7nm, right? In terms of the geometry migration schedule, almost the same. How you define the 16nm is a long node and the Sorry, the 16nm is long node or the 10nm is a short node? How you make such a judgment? Thank you.
Actually, we are working with the customer, and then according to their product's requirement and also their product's roadmap. Customer and TSMC determine which node will be the best performance and best cost effective. That's why we can say that the 16nm impact right now, it will be a long node as compared with the 20nm.
More like the survey feedback from the client.
Yeah. That's between cooperation, the TSMC and customers.
Okay. Last question. In terms of InFO, are you going to build a standard in the industry for the InFO technology?
We build InFO to meet customers' demand and to fulfill their requirement. We did not think about something else.
Okay. You will build a standard, right? I'm sorry. Anyway.
Okay.
Okay. Thank you.
Thank you very much.
Thank you.
All right. Next one will be coming from the floor. It will be Morgan Stanley's Bill Lu.
Hi. Thank you very much. Follow up on Eric's question on this concept of equipment reuse. I guess, first of all, does that change how you depreciate? Secondly, I'm not sure if you agree, but it seems to me like, it's not just a 1 node phenomenon. The next several nodes, you might have more and more equipment reuse than what you've seen in the past. I'm just wondering if you could tell me whether you agree with that, and if you do, what does that mean for your business model as far as return on investment?
Let me answer the first part of your questions regarding the reuse. Are we changing any depreciation rule? For the manufacturing tool, we use 5 years straight line depreciation, but there are some called open idle tool. That means the tool cannot be converted to next generation. For those tools, we use 3 years, we specifically identify which parts cannot be converted. Actually, with the effort from our operation people, this non-convertible part becomes smaller and smaller. Mark was talking about above 90%, very close to 95%, are convertible to next generation.
Just to be clear, if you start a tool at 20, it now converts to 16. The schedule doesn't change on the depreciation if it's converted, right?
It doesn't change.
Right. Thanks.
Can I ask you, what do you mean about business model you have in mind?
I guess, I just feel like the smartphone is now such a big part of the overall industry, and it's driving a lot of the leading edge. When that starts moving to the next generation, maybe the second wave and the third wave aren't going to be as big as they used to be in comparison, right? Maybe you convert more of 16 to 10 than you did historically, where you kept all these old nodes for a long time. I wonder if you see that as a trend. What does that mean for your business model?
I think, the reason we are very focused on the tool compatibility is indeed what you mentioned. The second wave and third wave, some of them may not catch up as fast as in the past. We do this, it's for the maximum effectiveness to provide our technology to our customer. Many of our customers have their product schedule differently. Some of the customers will have their product launch at different times than the other, we pace our technology so that they can take the maximum benefit of a particular set of tools, with the most updated technology capability. That's our value to our customers. The added burden for us is, of course, manage the tool compatibility. I think we have the profitability target in mind to ensure that we will keep that.
Sorry. I wasn't exactly clear. I guess what I was getting at is if you have more reuse and more conversion, maybe it means lower capital spending going forward structurally, right?
Can you repeat your question or your comment?
Want my help? My interpretation of Bill's question is that because we are using a lot of those reuse because of the high common tools and therefore the need for our fresh new capital investments, actually, the burden is lighter and it should actually help our profitability and returns.
Of course. That's why we want to make it most effective of the tools.
Thank you. How do you quantify that?
How do I quantify the effectiveness?
Maybe, I don't know if you can help me with the amount of reuse today versus, last generation or how you think about CapEx intensity now versus, last generation.
For the capital intensity, it's increasing for each node. Therefore, we try every way to reduce that and make it more effective to counter that heavy loading. That's the tool migration is about. You want me to quantitize? Maybe Lora can help me.
Let me try, Bill. Remember, this question has popped up quite often. Maybe two or three years ago, we were talking about a conversion rate is more than 70%, and then we said it's more than 85%, and now we are saying it's above 90% and 95%. That's the effort we made during this process. In many ways, you have to start from the design phases. You have to look at design. You make sure those tools get utilized more than one generation. That's one thing. The other one is you need to know how to modify the tool so that you don't have to buy a new one. That's another reason. There are many things happening in the factory, and we are giving an aggressive target on this, and we measure that.
Thank you. Second question is for Dr. Liu. You had given us some of your new demand drivers, which is very helpful. Can you look at how that impacts the foundry industry? If you look at 10 nms, how big do you think is the foundry market with these added demand drivers versus, say, 16 versus 28? I just feel like we talk a lot about the ramp, right? The ramp is a little bit different now because it's one customer specific. Can you look at the overall market as it matures?
The drivers I mentioned about mostly is to show the demand for the leading-edge technology, where the computing is hunger for either has fast speed or more for mobile, lowest power. Therefore, you can see we benefit more on the leading-edge technology compared to other foundry players. I do see the trend will continue. The leading-edge portion of profit really will increase as time goes.
Do you think 10 nano will be bigger now than 28?
It's too early to tell. I think in the beginning of 28 nm, we didn't think 28 is so big either. I believe as we right now size our capacity only according to our customer's demand. As time goes on, we may see that differently. We expect this should be comparable. At this point, at early stage, it's indeed much like 28 in the beginning. It is smaller than what 28 is today.
Okay. Thank you.
Okay, I think it's about time that we will take our next question from the call. Operator, please proceed with the next caller on the line.
Your next question on the line comes from Brett Simpson from Arete Research. Brett, please go ahead.
Yes, thanks very much. I have a question for Lora on FX. Lora, can you just confirm, does TSMC sell all its wafers in U.S. dollar? On the gross margin side, can you perhaps give us a sense for how cost of sales breaks down between U.S. dollar and Taiwan so we can understand the movements in FX on the gross margin side? Maybe for Mark, regarding 16 nm and 20 nm, your sales in Q3 combined is just over 20% of sales. How do you see these combined nodes as a percent of sales in 2016, and how might it split between 20 and 16 nm? Thank you.
Let me answer the first part of the question regarding a foreign change rate. It's indeed our revenue are more than 99% denominated in U.S. dollars. In terms of purchasing, not so much on cost of sales. In terms of purchasing, I think we have about 50% in U.S. dollars and 20% in EUR, 20% in TWD, and the rest in JPY. There's the effect of revenue minus the cost impact. If you recall, we have said earlier, 1% affects impact will change our gross margin rate by 0.4 percentage point. This rule of thumb still holds today. Second part of Brett's question is combined 16 and 20 nm proportion to revenue next year.
That will be greater than this year, I can say, I cannot have a very accurate number right now to share with you.
Brett, you have to wait-
Maybe-
a little bit longer.
Maybe if I can follow up about 2016. I think you talked in your prepared remarks about growing sales next year, double-digits. I know you mentioned FX would be a big driver in achieving double-digit revenue growth next year. If I look at the market on a U.S. dollar basis, what are the demand drivers you see in 2016? Because it's clear the Android smartphone market is having a lot of price pressure, and next year, iPhones will have much tougher compares. What are you most optimistic about in demand, and is there a risk that leading-edge demand maybe disappoints next year? Thank you.
Brett's question is about next year's growth driver. Given that smartphones is experiencing very heavy pricing pressure, and it has actually been a pretty significant growth driver for TSMC in the last few years. What would be the demand driver for TSMC next year?
It's still going to be leading-edge technology, and we see mostly will still be high-end smartphones.
Brett, maybe I can add.
Thank you very much.
one element to that. Not only is it the smartphones, but also TSMC increasing our penetration in the smartphones. All right, next question will still be coming from the line. Operator, could you please go to the next person on the line?
Your next question today comes from the line of Mehdi Hosseini from SIG. Mehdi, please go ahead.
Thank you. Thanks for taking my question. First question has to do with the first quarter. It seems unusual that your customers' forecasts are still uncertain, especially after several quarters of working down inventories. Is that right to assume that this inventory correction cycle is indeed different than the previous cycles? Especially when we look at your cycle time averaging two months, and keep in mind that Chinese New Year in February coming, and your customers, even after several quarters of inventory digestion, are still not willing to refresh inventories. I want to get more insight from you on this factor, and I have a follow-up. Thank you.
All right. Mehdi, let me repeat your question. Basically, you're a little bit surprised by the length of time that our customers are going through the inventory digestion, as you said, that it has already been going on for several quarters. Given the cycle time is about average two months, why does TSMC still have uncertainty about first quarter next year?
Okay.
Yes.
Let me answer about the inventory for the whole year. For TSMC, as you recall, our first quarter is a pretty good quarter. It's flat. Typically in Q4, Q1 is a weaker quarter. We find out that actually the first quarter, the inventory actually is higher than season normal by several days. We were thinking of how this inventory will work down. Unfortunately, the demand in China turned sour during the first and go in the second quarter, coming down very fast. That surprises us, to our customers too. Therefore, the inventory really didn't work down that well. As a matter of fact, it actually increases. That puts everybody uncertain on the inventory. We look at the Q3. I think Q3 traditionally is a strong quarter, and the people still didn't really cut down their inventory that boldly.
Therefore, we see at the end of Q3, it still didn't come down too much. However, from the current global economy and perspective, people really put it to realistic view. We look at the Q4, indeed, people are actively adjusting inventory just like several other years end. They try to reach the inventory towards the seasonal level. That's why we see the weaker, particularly weaker Q4. That's the situation. This year is very different than last year, and I hope it's not going to replay next year.
As a follow-up, second question, want to revisit the 10nm. In the past, you have suggested the 10nm contribution would start in Q4 2016. Is that forecast still remaining the same? If the InFO would start generating TWD 100 million per quarter, is that part of the 10nm, or should we account the InFO as additional line source of revenue?
All right. Mehdi's question is, will we, as we said before, begin to see revenue contribution from 10nm starting in fourth quarter 2016? Whether or not the TWD 100 million quarterly revenue by 4Q 2016 from InFO included the 10nm revenue.
I'll answer the first question. Yes. We did one time talk about we want to enable customers production at the end of 2016. The first customer we have thus played it in a conservative way, and they didn't set the most aggressive schedule, and they put a production of 10nm as planned, as their product plan, according to product plan. That plan has not changed. We continue to develop technology to produce the best quality 10nm to fit that schedule still. We talk about we're going to freeze this technology at this quarter and going about the qualification and prepare to the same production schedule of 10nm next year.
I answer the second part of his question. On InFO's revenue next year, in the first quarter, as I stated, it will be more than $100 million by InFO per quarter. By InFO alone, not with the silicon wafer or others.
Thank you.
Okay. Let's now coming back to the floor. Next question will be coming from the floor, a follow-up from Credit Suisse, Randy Abrams.
Wanted to ask a follow-up. You mentioned about InFO, TWD 100 million. Could you talk about the scope of the whole back-end business? Because you're also doing bumping, the CoWoS, maybe the size of the back-end business, and as InFO ramps, what you see as potential, if you have any aspirations, how big do you think it could be 12 to 18 months from now?
InFO's revenue will be a part of the whole back end, certainly the back end of business is much bigger than this $100 million US per quarter. Cannot give you a roughly right number right now.
Okay. The second follow-up question. We had a lot of reuse this year. Could you talk about for next year, 2016, are you through with that reuse or you see potential to get some savings from 2016 or from some of the prior nodes? You think next year could be a year you get some of those reuse? Then if we're looking at CapEx TWD 8 billion this year, what that means for depreciation for next year, if it's going to be a much more modest increase on depreciation.
Randy, I did not catch up your first part of a reuse. Reuse of what?
Sorry, for the capacity. If you'll convert more capacity next year, if you see more CapEx savings from converting 20 to 16, or some of your prior nodes like 28 or lagging nodes to reuse and move tools down. You think next year is another year to reuse CapEx or reuse capacity to save CapEx next year.
In other words, he's asking whether the conversion between 20nm to 16 nm will continue into next year, and whether there will be other type of conversions. For example, converting from 28nm to some other nodes also occurring next year.
We continue to convert 20nm capacity into 16 to meet the customers' demand. I described in my presentation.
If I can make some comments on this. Actually, the conversions depend on the demand on those two technologies, and it is ongoing process. Obviously, we have started conversion this year. It will depend on next year's demand. If there's a need, we will convert, continue doing that.
Oh, depreciation.
Depreciation, I can talk about this year first.
Okay.
This year, we expect with TWD 8 billion CapEx, depreciation is going to go up by about 11%-12%.
Yeah.
Next year, since we have not fixed the CapEx yet, it will not be a very high number either.
Okay. Next questions will be coming from HSBC, Steven Pelayo.
Just a couple quick follow-ups here. You mentioned that 16 and 20 combined will definitely grow year-on-year. I'm just curious in terms of sequential trends. If about 21% of revenues, does that grow every quarter when you look into Q4 and Q1?
It will grow year-over-year. I cannot comment on the quarter. All right.
Okay, fair enough. Three, four, five years ago, I think Chairman mentioned the one number that really summarized it quite cleanly and easily for us was that this move to smartphones, we were going to see a tripling of the silicon content per phone. Lo and behold, that's really what drove TSMC on top of some share gains and really no Intel in this mobile phone world. I guess, as I'm thinking over the next three, four, five years, you guys are highlighting Internet traffic, and even that's only 11 million servers a year, not really that big. Cars are only 60 million-70 million cars per year. It's very difficult, I think, for us now to really see something that can be as significant, even with the size of a market and the silicon content increase potential.
I guess when you're thinking out over the next three, four years or so, what do you think are the biggest drivers? What's that same driver that we saw five years ago, like we saw with smartphones? What excites you most, I guess? Can you quantify it a little bit to a simple sentence like that?
Steven, this is a tough one. I mentioned all the trends of computing, actually, in IoT, mostly growth will come computing, because the sensor and connectivity are more mature technologies. There is nothing, killer apps as the smartphone today. As you know, all the computation demand
Some of the usage model come and goes. It wasn't successful. For example, we still are yet to know whether the watch will become a usage, people will get used to it or not, drone or other things, home devices. I just don't have the clear naming for that gadget. Okay. I think what I can say is just like five years ago, the connectivity and communication is a definitive trend, therefore, we see the smartphone will come, and indeed, it did. It does pay. However, going forward, how this computation come into what form of gadgets? Actually, I cannot clearly describe to you. Yeah. People still talk about the smartphone in their pocket is too hot or it doesn't take long enough time, and let alone any virtual reality or augmented reality can be fed through your smartphone.
Your smartphone will die within one hour if you're doing that. Your entertainment, people would like to watch 3D TV, but really, the kids today really looking at entertainment of a different form. They want to evolve into the virtual reality, and that could be the future form of entertainment. Do we call that movie theater? Maybe, but could be a different name. The gadget is still upon the innovator of the product industry to come out. I think, looking at our customers, there are tons of innovators. They are thinking, come out new products to becoming a next smartphone. They produce their design continuously. Let's see how that will pan out to be a particular gadget for a particular successful innovator.
I understand. I'm sorry, if I can just do one more quick one. When will 10 nm debut its revenues? As you said, it was 16 nm this quarter. I know you have tape outs coming out in 10 nm mid-next year. Do we wait until 2017 before it's a few percentage of revenue, or do we think it actually can contribute a few percent by the end of next? Do you have?
We will start production maybe the fourth quarter next year, but revenue, we expect from year 2017 in the fourth quarter.
Sorry, 20, okay.
That will be, yeah.
Okay. Thank you.
All right. Questions will be coming from Deutsche Bank's Michael Chou.
Thank you. Could you give some color for 10 nm scaling versus 16 nm? 10 nm scaling.
Scaling, you mean
Yeah
compared to 16, how density and so on, I think?
Yeah, it's a point. Let me comment on that. 10 nm scaling, we designed it very aggressively. Typical node to node scaling, if we talk about logic density, it's about 1.9x scaling. I just show you the number, the 10 nm, we designed to 2.1x scaling from 16 FinFET Plus.
Second one is, as you mentioned, your 10-nm risk production should be Q1 next year, right?
Q4.
Oh, Q4. Okay. After one year risk production, so we should expect you can enter mass production in 2017.
Yes.
Based on your R&D progress to your customer, do you think you can meet the customer's scaling requirement this moment?
Yes. We can. Yeah. The tape-out is imminent.
Thank you.
Okay. Follow-up questions again coming from Citi's Roland Shu.
Just one follow-up question for depreciation. Lora said the depreciation based on TWD 8 billion capital spending this year. Depreciation will be about up 11%-12%. You said 2016 will not be a very high number. You mean the depreciation or CapEx?
I said CapEx will be higher than TWD 8 billion.
Uh-huh.
Depreciation year-over-year will not be very big number compared to the changes of this year versus last year.
For this less spending depreciation and also the macro uncertainty this year and next year. Are you still holding your long-term financial goal for EPS to grow more than 10% for the next five years? Is this goal still hold?
10% is a good goal.
Yeah. I think previously, you talk about this long-term goal. Yeah.
It's long-term. I'm talking about five-year goal.
Yeah.
10%.
Still hold?
As a goal, still holds, but we need to work on it, right?
Okay. Thank you.
All right. There is also a follow-up question from Morgan Stanley's Bill Lu.
Hi. Just two very quick follow-ups. One is, I think Mark Liu talked about the second generation InFO that's under development. Second generation InFO. Is that still targeted for the mobile processors, or is it for new applications?
It probably will be adopted by the mobile processor first, and then applied to all other applications. We are working with many customers right now. In fact, I cannot really tell you which one will be the first one to use it.
Secondly, can you give us an update on EUV? There seems to be some talks recently that you may be exploring other options besides EUV.
Okay. Let me comment on EUV. In the past three months, we see quite encouraging progress of EUV development. We work on our 3300 tools from ASML to improve their source chamber and the pockets, and a very good progress. We have agreed to go further to moving the 3350 tools. That tool has made also major modifications, and we plan to move that in January next year, so that to catch up, not only accelerate development, but also catch up on the five nm development program.
Follow-up questions from J.P. Morgan's Gokul.
Thanks. Just one follow-up question. Dr. Liu mentioned that the system fabless companies, the new breed of companies, are becoming bigger and bigger. Could you talk a little bit about what kind of customers are these? I think currently, obviously, there are a couple of them that are really big and primarily in the mobile space. Are we seeing expansion of these companies towards other areas as well, maybe data center and that kind of areas as well?
I cannot name the company's name for you, but let me share with you the thought. Actually, this phenomenon is not new. Today, of course, the smartphone you refer to, mobile space, a lot of system company does come out, create new momentum for the demand market. In other area, like gaming, for example, it's produced by the system companies. We will see other areas for the virtual reality, for example. Some of that probably will be driven by the system companies. Cars, for example, the machine learning, that probably will also be driven by fabless and system company together. We see this a definitive trend that system developers could not find the standard product on the market, therefore, they involve on the design to optimize their system hardware and software to create more innovative products.
That we see, that's where the encouraging is, where the new players of innovation in the semiconductor industry emerges.
I think it's about time that we should end this quarter's conference call. Thank you for coming over for today's conference. Please be advised the replay will be available in three hours from now, transcripts 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. Have a good day.