[Foreign language]Welcome to Taiwan Semiconductor Manufacturing Company first quarter 2018 earnings conference and conference call. This is Elizabeth Sun, Taiwan Semiconductor Manufacturing Company Senior Director of Corporate Communications, and your host for today. Today's event is webcast live through Taiwan Semiconductor Manufacturing Company's website at www.tsmc.com. If you are joining us through the conference call, your dialing lines are in listen only mode. As this conference is being viewed by investors around the world, we will conduct this conference in English only. The format for today's event will be as follows: First, Taiwan Semiconductor Manufacturing Company Senior Vice President and CFO Lora Ho will summarize our operations in the first quarter, followed by our guidance for the second quarter. Afterwards, Lora Ho and Taiwan Semiconductor Manufacturing Company's Co-CEO Dr. C.C. Wei will jointly provide the key messages. Taiwan Semiconductor Manufacturing Company's Co-CEO Dr. Mark Liu will host the Q&A session, where all three executives will entertain your questions. For those participants on the call, if you do not yet have a copy of the press release, you may download it from Taiwan Semiconductor Manufacturing Company's website at www.tsmc.com. Please also download the summary slides in relation to today's 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 our CFO, Lora Ho, for the summary of our operations and current quarter guidance.
Thank you, Elizabeth. Can you hear me okay?
Well, good afternoon everyone. Thank you for joining us today. My presentation will start with the first quarter highlights and follow by the second quarter guidance. In the first quarter, due to seasonality, our first quarter revenue in US dollars decreased 8.2% sequentially to $8.46 billion, which was at the middle of our guidance range. In NT dollars, revenue declined 10.6%, reflecting 2.7% appreciation in NT dollar against US dollar. Gross margin increased 0.3 percentage points sequentially to 50.3%. As our cost improvement efforts and inventory valuation benefit offset the lower capacity utilization and unfavorable foreign exchange rate. Total operating expense decreased by 2.1 billion NT dollars, and represented 10.8% of revenue versus 10.4% in the prior quarter. Thus, operating margin slightly decreased 0.2 percentage points to 39.0%. Overall, our first quarter EPS reached TWD 3.46, and ROE was 23% for the quarter. Now, let's take a look at the wafer revenue contribution by application. During the first quarter, communication and industrial standard decreased 19% and 4% from the prior quarter respectively. While computer and consumer increased by 30% and 9% respectively. Now, let's take a look at the revenue by technologies. 10 nanometer process technology contributed 19% of our total wafer revenue in the first quarter. The combined 16/20 contribution was 22% of total wafer revenue. 28 nanometer and below advanced technologies accounted for 61% of total wafer revenue. Moving on to the balance sheet, we ended the first quarter with cash and marketable securities of 684 billion NT dollars, an increase of 35 billion NT dollars from last quarter. On the liability side, current liabilities decreased by 16 billion NT dollars. On financial ratios, accounts receivable turnover days increased two days to 42 days. Days of inventory increased 11 days to 63 days, due to an increase of raw wafer and a 10 nanometer wafer pre-built before the capacity is converted to 7 nanometer. Now let me make a few comments on cash flow and the CapEx. During the first quarter, we generated about NTD 161 billion cash from operations and spent NTD 72 billion in capital expenditures.
Our free cash flow was NTD 89 billion. We also repaid NTD 17 billion of corporate bonds and prepaid NTD 34 billion of corporate bonds and interest. As a result, our overall cash balance increased NTD 24 billion to reach NTD 578 billion at the end of the quarter. In US dollar terms, our first quarter CapEx expenditure was $2.45 billion. I have finished my financial summary. Now let me turn on to the second quarter guidance. Based on current business outlook, we expect second quarter revenue to be between $7.8 billion and $7.9 billion, which is a 7%-8% sequential decline, but 11.2% year-over-year increase at the midpoint of my revenue guidance. Based on the exchange rate assumption of one US dollar to 29.20 NT dollars, our second quarter gross margin is expected to be between 47%-49%.
Our second quarter operating margins is expected to be between 35% and 37%. Also, in the second quarter, we will again need to accrue the 10% tax on undistributed retained earnings. Due to higher accumulative translation loss as a result of NT dollar appreciation against US dollars, which will reduce our retained earning tax. Therefore, our second quarter tax rate will be between 18%-19%. The tax rate will fall back to 10%-11% level in the third and the fourth quarter, and the full year tax rate will be about 12%. This concludes my remarks. I will now make comments on capital, capacity, and profitability. I will start with CapEx and capacity. At our last conference, we stated our 2018 CapEx budget to be between $10.5 billion-$11 billion. We now see our CapEx to be between $11.5 billion and $12 billion US.
The increase is due to, number 1, we plan to spend about $500 million more to increase our maskmaking capacity to support our customers' higher tape-out activities. To spend about $300 million as prepayment for our High-NA EUV tools. Going forward, we expect our annual CapEx in the next few years will be ranging between $10 billion and $12 billion US. Let me explain how we are able to support a 5%-10% long-term growth with a similar level of annual CapEx as the follows. For the existing capacity, we are able to grow capacity through productivity improvement. That is, for the same tools, the output can increase every year through our engineering efforts and innovations. This way, without spending fresh CapEx, we are able to grow capacity to support growth.
On average, we are able to grow our overall capacity by a mid-single digit each year through the productivity improvement. Our marketing strategy with respect to the existing capacities, which grow in productivity perpetually, is to seek enough demand to fill this growing capacity at all times. As for the new capacity, which is mainly leading-edge technologies, we are very careful in planning the right level of peak capacity, knowing that initial capacity we build will grow continuously to a larger size due to productivity improvement. To summarize, through productivity improvement and careful planning of new capacity, we are able to support revenue growth of 5%-10% with annual CapEx at a $10 billion-$12 billion level in the next few years. Let me make a comment on profitability.
TSMC's profitability is determined by the following factors: leadership technology development and ramp-up, pricing, capacity utilization, cost reduction, foreign exchange rate, and technology mix. I will go through the changes for our second quarter gross margin compared to second quarter 2017 and first quarter 2018 respectively. Compare second quarter 2018 gross margin using the midpoint of the guidance I just mentioned, which will be 48%. Compare that with second quarter 2017 gross margin, which was 50.8%. The decline of 2.8 percentage points gross margin rate is explained by 1.3 percentage point negatively due to exchange rate. NT dollars has appreciated 3.6% from second quarter 2017 to second quarter 2018. In addition to that, there is a negative 1.5 percentage point due to unfavorable product mix and inventory valuation, which is resulting from a utilization change.
By compare second quarter 2018 with first quarter 2018, where our margin was 50.3% last quarter, the decline of 2.3 percentage point gross margin is explained mainly by the favorable two percentage point of inventory valuation that was embedded in the first quarter 2018 gross margin, which I have explained in the last quarterly conference. To conclude, our leadership in technology development and ramp-up remains solid. We continue to maintain our competitive price, and we are making good progress in cost reduction. We will continue to work on filling our capacity. I finish my remark. Now, let me turn the microphone to C.C. Wei for his comment.
Thank you, Lora. Good afternoon, ladies and gentlemen. Let me start with our near-term outlook. We concluded our first quarter with revenue of NTD 248.1 billion, or $8.46 billion, in line with our guidance given three months ago. This result was mainly driven by a strong demand from high-performance computing, such as cryptocurrency mining, and increases from both automotive and IoT, but offset by seasonal decline in smartphones. Moving into second quarter this year, our business is expected to be affected by continued softer demand from smartphone segment. This decline is expected to be partially mitigated by the strength in HPC. Our revenue in US dollar is likely to grow by only about 10% over second quarter last year. We forecast our fabless DOI to stay slightly above seasonal level, but will track seasonal pattern.
For the whole year of 2018, we forecast the overall semiconductor market, excluding memory, will grow by 5%, while foundry is expected to grow by about 8%. We forecast TSMC's 2018 revenue in US dollar will be about 10%, rather than the previously indicated 10%-15% due to the smartphone weakness and the uncertainty in cryptocurrency mining demand. Now, let me talk about the long-term business growth driver. We are optimistic about the development of some of the industry's mega trends, particularly AI and 5G communication. We believe these two will help semiconductor to spread its pervasiveness into our daily lives. Both AI and 5G will create new usage models and spur new waves of demand for both of the existing and emerging applications with increasing silicon content. We expect TSMC to benefit from these industry mega trends in all four of our growth platforms.
In mobile, silicon content of the smartphones will increase due to increase in functionality, such as facial recognition and new usage such as AR, VR, and 3D video. In HPC, we expect AI will boost the attach rate of accelerators using data center for today's mid to teens level to about 50% by 2020. In automotive, the use of the new safety-related functionalities, such as ADAS, and eventually autonomous driving, will drive the increase in silicon usage. In IoT, AI will proliferate into broad-based client devices across many applications, such as smart voice assistants or electronic appliance management. This again, will increase silicon content. We believe our four growth platforms are well-positioned to benefit from the longer-term mega-trend of AI and 5G. Our leadership in advanced and specialty technologies, as well as our advanced packaging solutions, should enable us to capture the future growth opportunities well.
Let me talk about the N7 ramp-up. TSMC's 7-nanometer technology in terms of performance, power, and area density, as well as its schedule, is leading the industry. So far, we have already fabbed out more than 18 customer products with good yield and performance. More than 50 products tape-outs has been planned by the end of this year from applications across mobile, server CPU, network processor, gaming, GPU, FPGA, cryptocurrency, automotive, and AI. Our 7-nanometer is already in volume production. I'll talk about the N7+ and the EUV. We believe we can extend the success of our 7-nanometer, N7, to its enhanced version, N7+, which will have 20% better density and greater than 10% power reduction. In N7+ we use a few EUV layers to replace immersion lithography process. As a result, fewer masking layer can be used.
The N7+ will use more than 90% of the same tools with N7 and N10, where we have fine-tuned all the advanced equipment to their optimum condition during their ramp-up, we believe we can leverage our production learning to N7+ and enjoy the industry's best defect density among competitors at comparable technologies. Our N7+ silicon results today are very encouraging. Not only we have demonstrated equivalent or better performance and yield on both 256-megabit SRAM and on product-like test vehicle when compared to N7 baseline, we have also demonstrated tighter distribution of electrical parameters in layers where EUV is applied. Since we maximize design rule compatibility between N7 and N7+, our customer can minimize IP porting effort. A few customers have already planned to tape out N7+ with us in the second half of this year, and more in the first half of next year.
Our N7+ volume production is planned in 2019, which remains unchanged. We have made ready multiple EUV scanners to support not only N7+ development, but also N5 development. At N5, with more extensive use of EUV, we have obtained consistent double-digit yield on 256-megabit SRAM, as well as a larger test chip. Our silicon data has proved all the benefits we expect from process simplification with EUV. Besides the silicon development, EUV technology continue to mature toward high volume production with improving source power toward the 250 watts goal, which we expect to achieve in a few quarters. Good progress continue to be made in the EUV infrastructure in the last few months. They include photoresist, mask defect and yield, particle defects, and transmission. We are confident that EUV can meet our goal of 2019 volume production for N7+, and 2020 volume production for N5.
Let me move to 16 FinFET and 12 FinFET. We introduced our 16 FinFET in 2015, and then 12 FFC in 2017. Compared with 16 nanometer, our 12 nanometer technology deliver better density, performance, and more efficient power. With continued improvement in yield, customers are gradually moving from 16 nanometer to 12 nanometer. Among all the product tape-outs we received for this year on the 16/12 node, about 30% are for 12 FFC. The application of 16/12 includes mobile, GPU, AI, networking, FPGA, consumer application, and automotive. Our capacity of 12/16 node is being fully loaded currently. Let me update Nanjing Fab. We expect to expand our business in China. Our Nanjing Fab will enhance our support to local customers. Because of high demand on 16 nanometer, we had advanced the startup of Nanjing production. The production of our Nanjing plant starts this month.
Both yield and performance are comparable to our 16 nanometer Taiwan Fab. Our advanced packaging now, let me talk about the last item. Our advanced packaging technology, InFO and CoWoS, are becoming more important for our customer to reduce their product footprint, package thickness, while enhancing the performance. TSMC's InFO is in its third year of volume production. We have expanded its capability to cover large die size so that we can integrate two or more chips together into one package. This is particularly useful for HPC products to obtain optimal cost performance benefit. On CoWoS, we have observed a growing number of tape-outs from HPC customers in graphic and networking segment. As a result, we are increasing our capacity now to support the demand. Thank you for your attention.
All right. This concludes our prepared statements. Before we begin our Q&A session, I would like to remind everyone 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 answer your question. 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. Mark will host the Q&A session.
The first question will be coming from Credit Suisse, Randy Abrams.
Okay. Yes. Thank you. I want to ask the first question about the demand and pricing environment. Just first on pricing. In first quarter, the shipments were down just 1% and revenue was down 8%. I'm curious if that's mainly the mix factor because 10 fell off, or if there's any change in pricing maybe on some of the more mature nodes. If you could talk a bit more on the demand change versus you say in January, how much from smartphone? From the crypto you talked about uncertainty, how much is coming from that market, and what are the uncertainty factors you're seeing on the crypto market?
Can you answer that?
Pricing.
Yeah. Lora? Please.
Yes.
Randy, your question is about the first quarter versus fourth quarter, quantity versus pricing. It is true that the first quarter pricing was affected by the product mix because we have very high utilization on the 16 and 10 nanometer, and that's the case.
I think Randy also has a part of the question about the change in our forecast compared to January.
Okay. CC, about the product mix. The product mix, actually, the first quarter we just mentioned, weakness of seasonality actually. Let me say that smartphones and all other segment actually, start continue to grow. Okay. In the HPC, automotive, IoT. However, that our revenue, a lot of portion is rely on the smartphone. That's why we have a mixed effect. In the technology, our smartphone use most leading edge technology, and that's why a few % drop that's a big drop in the revenue.
The second part of that question, The second question was just about the crypto. If you could talk a bit more about the uncertainties, if you've seen that sustain or you're seeing more uncertainty. Just the factors you're seeing in that market. The second question I'll have is on 7nm plus. If you could discuss how you're seeing that ramp in terms of steepness at this stage relative to we've seen steep ramps of 10-nanometer and 16 nanometer before. If you're seeing steep ramp going into 7nm plus and also just how important it is to have a steep ramp. In terms of how important it is to have that steep ramp, just in terms of the learning as you try to bring customers onto EUV.
How important is it to have a big volume ramp on 7nm plus
7nm plus ramp
for your EUV development?
Okay. Let me answer the cryptocurrency. We see still very strong demand in the first quarter from cryptocurrency in the first quarter. During the second quarter, this is start of the second quarter, we see some weakness on 28-nanometer, but the rest of the technology is still very strong on cryptocurrency. Does your second question is on the 7 plus ramp?
Yeah. It's two parts. How steep you're seeing or how much demand inflection are you seeing for 7 plus? How important is it to TSMC to have a wide adoption in terms of the learning, like to move customers to 7 plus in terms of learning as you look ahead to five and even against your competitors?
The 7 plus, I mentioned we are using a few layers of EUV, and the EUV progress is very well. Our customers start to adopt it. How important? It's important. We expect that the N7 plus will start to ramp up in the second half of next year quickly. Actually, it's ramping up quickly because of smartphone business.
Thank you.
I think 7nm+, we have EUV development, as C.C. just addressed, progress smoothly. We have demonstrated 7nm+ yield equivalent to 7nm. The rest is volume. I think the volume, we don't need a huge volume to prove EUV because all the development is in place. The biggest volume ramp will be in N5, which in 2020; already in 2019 have some small volume. N7+ will slightly prelude that, but we don't need as big a volume to enable the N5. No, we just need some volume to prove that. N7+ provide better density, better performance for our N7 customer following their N7 products. Either they stay on N7, go to N7+, or they can from N7 go to N5, depend on their product needs.
Next question will be coming from Deutsche Bank, Michael Chou.
Thank you. First question is regarding your revised guidance for 2018. Can we say you will still have the same 7nm sales ratio target this year, 10%?
Close.
Yes, around 10%.
Around 10%.
Yes.
It's still the same outlook as you expected three months ago.
As I said, the smartphone, it will be weak, continue to be soft that we predicted today. It's the loading, the 7nm. It's a little bit, one point off.
Okay. Is that fair to say 7nm sales portion may be slightly below your previous target? Or you still think this should be close to 10%?
Very close.
Second question is regarding your structural profitability. Do you think this year your structural profitability will improve or be the same as last year?
I've said several points that are associated with the structural profitability. Technology, pricing, cost, capacity utilization, exchange rate, and product mix. On this one, there are certain things we can control. Number one, for exchange, we cannot control. Product mix always go with the customer's demand. While we have seen some weakness in the first half for the mobile and there are certain technology of our existing capacity, maybe not so full for the whole year. That would be one factor to our overall structural profitability. I think our objective to maintain or improve structural profitability remain unchanged. The market situation and some uncertainty, we have to be mindful.
Is that fair to say your biggest uncertainty will be FX rather than the other factors?
FX at least explains half of the deviation for profitability. The other half, I think it's mainly utilization, product mix, and other factors.
Is that fair to say that your 28-nanometer UTR slightly below your expectation, so that could lead to some margin downside?
It could.
Okay. Thank you.
Next question will be coming from Sinolink, Andrew Lu.
Last time I came back, year 2010, the chairman give me three questions to ask. Can I give this a exception?
You have two.
What was number one?
The second quarter guidance, the full year guidance, is this factor in the ZTE recently been banned by U.S. to sell the chip to ZTE? Is this factoring into your model? That's my first question.
The second half this year?
Almost second quarter.
Second quarter.
Second half also as well.
No, we haven't. We got this news just yesterday. We think the effect is very minimal. Currently, we are still under study what is the impact of ZTE suppliers. For first glance that we look at, we have a very wide customer portfolio. ZTE, depends what they're supplied from. Being everyone's foundry, we have very widespread. I think the impact will be softened much, much more. We think that minimal impact on the second quarter. You won't see the number change.
Which quarter? Sorry.
Second quarter you're talking about?
Okay.
Yeah.
Okay. Thank you. The second question I have. Okay. In terms of cash dividends, I estimate every year we got additional TWD 2 billion-TWD 3 billion additional cash, even though we pay a lot of cash dividend. Are we going to give a much higher payout ratio compared to in the past few years? Even this year, by factoring the new cash dividend, we still generate additional cash. Right now, I think it's over TWD 20 billion on hand cash already. We are keep going up every year TWD 2 billion-TWD 3 billion. That's my question. Thank you.
We have said many times our dividend policy to be sustainable and gradually increase dividend. We haven't really tied to a payout ratio per se, but we do look at free cash flow generation, next 12 months, next 24 months. We use that as a basis, to decide how much more dividend we are going to gradually increase.
It's TWD 20 billion to date, and I've said the CapEx will be TWD 10 billion-TWD 12 billion. We do have a capability in the future to increase a little bit more than we have been increasing in the past.
Next question will be coming from Citigroup's Roland Shu.
Hi, good afternoon. First question is, C.C., do you still hold the view that 10-nanometer total revenue will continue to grow this year?
The answer is yes.
Okay. For 10-nanometer to continue grow on top of this 10% year-on-year revenue growth, 7 nanometer around 10% grow from scratch. Both 7 nanometer and 10-nanometer probably total contribute more than 10% of the total growth this year. Does that mean that for 28-nanometer and 14 nanometer and above technology, the total revenue is not going to grow? Is that right?
You are doing a very good basic metric calculation. I don't comment on that.
Okay. What's the reason for this 28 nanometer and the 14 nanometer and above does not grow? Is that because the end demand is soft, or is this due to the capacity constraint, or we are losing market share to other?
I would say that a lot of 28 nanometers usage has been advanced into the more advanced node. They move to more advanced node. The utilization definitely is not what we expected last year. All I can say is we remain very competitive, and we are developing some derivative technology to serve all the customer. We move into the 22 nanometers to help our customer can get a very good cost performance benefit. We are maintaining our market share.
Okay. Last year, actually, we also expanded about 15%-20% of the capacity for 20 nanometer. Is there any problem to fulfill the capacity this year?
Last year, we increased the capacity because of a very high demand. A lot of customer did not have enough wafer. That continue into this year's first quarter and probably half of this quarter. As they move forward faster than we thought. That's why that we see a little bit weakness in the second half.
Okay. How are we going to utilize this new added capacity for 28-nanometer?
Oh, we have a lot of new application being developed. Trust me, that one day you will be fully loaded again.
Okay.
I hope it's as quickly as possible, technology product development takes time.
Okay. Thank you. For second question is for Nanjing Fab. Now, we are already at start mass production, and we have 20,000 wafer per month capacity. Is there any plan to further expand the capacity in Nanjing Fab?
20,000 wafer per month is still our plan. We plan for this year all the way to next year. We ramp up so quickly, actually, it surprise us.
Okay, for the further capacity expansion in Nanjing, are we going to buy the new equipment, or are we going to continue allocating from Taiwan?
We probably still at the first phase, still moving the equipment from Taiwan.
First phase means the 20,000? Yeah.
Yeah. Majority of 20,000.
That is ongoing or that is already done?
Is 20,000 capacity moved, yes.
Let me give you an idea. We are fully loaded, so I cannot move any equipment right now. Anything we move will lose the capacity.
Thank you.
Next question will be coming from UBS, Bill Lu.
Hi. Thank you very much. First question is for Lora. Lora talked about productivity improvements leading to mid-single digit capacity increases. Two-part question. One is that capacity increase pretty even across all the mature nodes, or is it certain nodes that you're seeing a bigger increase? Second part of the question is that mid-single digit, going forward, what was that number in the past?
I didn't hear your second question, sorry.
If you look at productivity improvements in the past-
Oh, mm-hmm
What has that number been?
Okay. Let me elaborate about the productivity. As you can imagine, when the new technology just launch, in the first few years, you have very significant productivity improvement. When it gets mature, we have stable productivity improvement. When it gets very old, you can have limited productivity improvement. The 5% I was just referring to is an average for the company, but it's not the same for every technology. Your second question is compare with?
Yeah, that mid-single digits per year-
What has it been in the past?
Oh, very similar. We have been doing the mid-single digit. Maybe some year is 1% more, some years 1% less. More or less, it's in the mid-single digit range. Yeah.
I guess I'm not really sure why that impacts your CapEx plans then, because it's always been like that, right?
It does, because another factor other than the productivity is, we also very carefully plan the peak capacity and watch very carefully for the capacity migration. C.C. was talking about technology where we use 90% commonality of tools. If you remember what I said a couple of years ago, the number was 70%, 80%. By doing that, actually enhanced the productivity improvement, especially for the leading-edge technology. That's one contribution to that.
Second question is on cryptocurrency. This is pretty new to many of us, and I feel like TSMC has been pretty conservative in terms of looking at it in the short term. That's the right thing to do. If you look at it for the next couple of years, 2019, 2020, as you think about your capacity planning, as you think about what customers to support, this is a market that is changing so fast. How do you think about it strategically?
Can we
Okay.
We look at the cryptocurrency's market price. It can drop from $20,000 US down to right now, $8,000. It increased from $1,000 to $20,000. This kind of uncertainty, that is what we are talking about. In terms of capacity support, unless it is sustainable demand, we will not increase the capacity because of this kind of uncertainty demand coming out, unless it's sustainable. In which I know that our customer is developing a lot of things on blockchain technology, AI. They are doing very well, and we expect that those cryptocurrencies, mining, those things will slowly move to AI area.
Do you have an estimate for, if you look at the crypto plus blockchain, how fast is that demand going to grow next several years?
First, I always say uncertainty of cryptocurrency mining, how can I put them together to give you a very good forecast? No. We don't.
Let me add. We increase our capacity in lieu of the invested capital, ROIC, return in capitals. If a demand is a spike demand, we'll be very careful. We try to support our customer in every way. What cryptocurrency developing is, because of price sliding, their demand of high-end technologies increases. The tail end becoming not that productive for them. This is the changing. In one way, we try to support them with a more advanced technology as we quickly, as time goes. Secondly, we support them on our available capacity.
Thank you very much.
Next question will be coming from Morgan Stanley's Charlie Chan.
Thanks. Hi, CFO. I actually want to follow up Bill's question on your productivity and CapEx. First of all, can you comment on recent chats about the EUV throughput issue, and is that related to your more volatile CapEx range? This is my first question. Thanks.
Can you repeat the question?
I think Charlie is asking whether or not we can give additional color about the rumored EUV throughput problems. If that's the problem that causes our CapEx to be more volatile.
The EUV progress, as we said, we move smoothly, and we got encouraging result. That's why I say we have confidence to support the N7+ volume production and the N5 volume production in 2020. Now you're asking about why we increase the CapEx, because we want to buy a very advanced EUV tool, called High-NA tools. That require that we put some deposit to book the machine slot. That's what we did. We are happy that the High-NA machine will be great for us to use in the future. That will further improve the efficiency and hopefully that increase the productivity, lower down the cost.
Thanks for the clarification.
Just make sure it's a High-NA in EUV. It's not the same EUV we're doing that. It is a next generation EUV tools. We work with ASML to book the tool early. Those tool currently is targeting at technology beyond 3 nanometer.
Yes. Thanks for clarification. Investors in general are a little bit nervous about this EUV progress, because it's associated to so-called Moore's Law progress, right? The rumor is that the daily throughput is now around 1,000 wafers per day. Right? It is not that economical. My worry is that whether that can affect your 7nm+ cost structure, because you still need to use EUV. The layer of the EUV usage could be fewer than your previous expectation. Anyway, that is what market is chatting about. My second part of the question is about the end markets. Okay, it's very difficult market to predict, we understand. Can you clarify your comment on the weak smartphone demand? I think now China smartphone, at least you see some seasonality, right? Why that is not reflect to your revenue guidance upside?
According to our analysis, we think some of your customer shrink their die size significantly this year. Is that affect the wafer demand for you at 6nm? This is my second question. Thanks.
The question is quite long, but let me give you some explanation. Yes, we do see China as a market start to pick up on the smartphone. In TSMC, in our smartphone market segment, there's some very high-end smartphone is a little bit soft. That's why we projected that it's going to be continuous softening. Okay. It's not because of China market. That start to pick up slowly.
Okay. Thank you.
Now, we would like to go to the lines for questions. Operator, could you please get to the first caller on the line? Thank you.
Our first question today comes from Brett Simpson from Arete Research. Please go ahead.
Thanks very much. I just had a follow-up question on crypto for C.C. Wei. You mentioned it was strong, crypto was strong in Q1. Can you confirm if this is double digit % of sales or not? Then just looking at your revised outlook for 2018, can you help us how we should think about crypto in second half? Do you expect crypto to be flat or up or down versus the first half in your revised outlook? Thank you.
All right. Brett, let me just repeat your question again. You were asking about the cryptocurrency and where C.C. said the demand was growing in the first quarter. You are asking if first quarter demand accounted for 10% of the total first quarter revenue, then second half cryptocurrency demand, will that be larger or smaller than the first half cryptocurrency demand? That's your question, right?
That's it. Yep.
All right.
That's right. Thank you.
Okay. Let me answer the question carefully, because I already said the uncertainty of cryptocurrencies are mining as we forecast in the future. So far, we still think that cryptocurrency in the second half still have a higher demand than the first half. They are using more advanced technologies. They will move into the seven nanometer that I can share with you. Okay.
Can you confirm whether it was double-digit, 10% of sales or not in Q1 for crypto?
I won't give a specific number of how much of a percentage it was in terms of the revenue. I'll say it's increasing.
Okay. Thank you very much. Just to follow up on 22 nanometer ULP and 12 nanometer ULP, can you give us an update on timing, and whether there's been any change in how you're thinking about the ramp of 22 and 12 nanometer ULP? Thank you.
Dr. Sun, repeat the question.
Well, Brett want us to give him an update on the development of 22 nanometer ULP and 12 nanometer ULP, ask if we have any changes in the plan, such as timing, schedule, market.
No, we don't change the plan. It continue to be a very important technology to TSMC. If we change anything, we speed up the progress. That's what we are doing. Because of where our customer is moving from 28 to 22 and from 16 to 12, we had to speed it up of all the progress.
Okay, thank you very much.
Operator, we can go to the next caller on the line. Thank you.
Our next question today comes from the line of Mehdi Hosseini from SIG. Please ask your question.
Thank you. One follow-up clarification, one question. Just coming back to your comment regarding days of inventory that were up during Q1. I think Lora mentioned that it had to do with 10-nanometer products, especially with the smartphone. Just want to make sure, the weakness that you see from a smartphone market, especially the premium products in the second quarter, that has to do with inventory digestion. Perhaps as the new products ramp in the second half, would the premium smartphone show a rebound in the second half? In other words, is Q2 kind of a digestion, inventory digestion period? I have a follow-up.
All right, Mehdi, let me try to repeat your question and see if it's correct. You are asking about the increase in our days of inventory. Then you are asking if the demand of the premium smartphone picks up in the second half of the year, does that mean that we will begin to digest inventory in the second quarter?
Yes.
Thank you.
Okay.
The days of inventory for TSMC in second quarter was 63 days. First quarter, 63 days. I explained because we are ramping down 10-nanometer capacity. We will massively ramp up seven nanometer capacity. We are in a migration period. We on purposely build more inventory on 10-nanometer. We can move the equipment to seven nanometer. We can save the CapEx. That's the thing. You're asking whether our days of inventory will go down. Not necessarily, because we are ramping up seven nanometer, which has a much longer cycle time. I do not expect our days of inventory will go down or will change much in second half of the year.
Okay. Very clear. Thank you. One question I have regarding EUV and migration. I understand that seven nanometer and EUV insertion doesn't really have much of a change to the design rule like a place route, but there would be a material change at five nanometer. In that context, when would you be able to provide those design libraries developed for five to the design community so that they could better evaluate the cost and benefit of EUV at five?
Well, Mehdi, you are asking us if because we are using EUV, there will be some material changes at seven plus nanometer. Do we provide design library or all these other design ecosystem support to incorporate those changes?
Yes.
Yes. Just to clarify, I think some of those changes would impact 5nm more so than 7nm+.
Please repeat your question again.
Sure. Okay. I think as we migrate from 7nm to 7nm+, with the inclusion of EUV, there may not be much change to design, especially with the back end of the design place and route. I do expect material change at 5nm. In that context, I think those design libraries developed for 5nm would be very critical. It helps design community to better evaluate cost and benefit. Want to understand, want to hear from management, when would those design libraries be provided so that we could better evaluate demand for 5nm with increased EUV insertion?
It is-
Is that clear?
the design library with respect to N5, 5 nanometer. Given 7+
Yes
7 are pretty much done.
Yes.
N5.
Yeah. N5
Yes
N5 technology development is well on track, and some of our customers function block already designed in.
Let me comment. I capture the first question is N7 to N7+, the design porting, we have demonstrated with several customers showing that the porting work is relatively easy. The only major difference is if you want to increase the density, you just change the standard cells, which have a tighter density. The rest is we have a porting algorithm to support our N7 customers. N5 is a new design, so we thus provide the design ecosystem start from early. Today is well on track, and we do support our customers on N5 design when they call upon.
Okay. Does that mean that you have already finalized the number of layers that EUV is going to be inserted for?
Your question is EUV insertable on N7?
My question is, if you already have provided cell libraries, does that imply that you have already finalized the layer count for 5nm that would include EUV?
Oh, you are asking us if we have finalized the number of layers that we will be using EUV at N5? Have we finalized that number?
Yes.
Yes. It's already finalized.
Is that closer to 10?
We do not disclose that. Thank you. We are coming back to the floor. Okay.
Thank you.
Thanks, Mehdi. Coming back to the floor, the next question will be coming from CL Securities, Sebastian Hou.
Thank you. My first question is to follow on the CapEx raise for this year. Lora already mentioned about the CapEx raise. Half of that is driven by the mask making.
Mask making.
Mask making, because of the higher demand from customer design activities. Can you elaborate on that?
Okay. Actually, we have seen increasingly a big increase in customer tape-out requirement. We found out our e-beam capacity is insufficient. Those tape-out in business is very good business for TSMC. In order to facilitate customer for their product launch, we need to add e-beam capacity immediately. That's the $500 million I was referring to.
Sebastian, let me add some color on this. We do see an increase of tape-outs, and we do see the increase a lot from AI networking areas. Some are from the fabless company, but some are from the system startups and some from even the cloud service providers. In this area, in the AI area, we think we are casting a bigger net in accommodating more customers today. How will that fade into volume? We yet to be see, but we are very happy to see the increase of tape-out and participation of new product designs in more companies.
I see. Thank you. Just a follow-up on that, because last quarter, you offered the expected seven nanometer tape-out by the end of this year will be 50. This quarter, C.C. already mentioned again you were going to be 50. If we just, judging from the number, it doesn't increase. Now, you're raising the mask making CapEx. Does that mean that actually the actual numbers internally you look at is actually increasing? Real tape-outs?
Yes, on other nodes.
Second question is that if I just do some calculation on the guidance revision for this year. Earlier was 10%-15%, now it's about 10%. That probably, if we use a midpoint, probably about close to $1 billion of revenue decline. Remember the last quarter, chairman talked about the first half year-over-year is about slightly above 15%. Now if we bake in the new guidance for second quarter plus the first quarter, we already know is about 12%. That's about like $600 million fall in first half. Earlier mentioned about the smartphone going to be weak, which means that there will probably another $400 million downward revision in second half this year. Would that be mainly driven by smartphone again or the other applications?
I think I just described it as a continued weakness of the smartphone demand and the uncertainty in the cryptocurrencies mining. We judged it down.
Just to follow on the uncertainty of cryptocurrency mining. Can I assume that even though it's hard to predict, but can I assume that you are baking in a more conservative assumption right now for second half compared to-?
Yes.
Okay, thank you.
Next question will be coming from Goldman Sachs, Donald Lu.
Sorry. First question is, what is the new guidance now for 7nm and the 10-nanometer as a % of revenues by Q4 this year? That's my first question. The second question is going back to this more tape-out activities. I think, in previously when you move on to a new node, the tape-out activities has been decreasing. I mean, more markets consolidation, et cetera, if I'm correct. Why suddenly you see there's more activities from different customers? Is there a reason for that, or is there going to be a new trend for the industry?
Lora will answer the first question.
Donald, your question regarding the revenue contribution from seven and 10 in the fourth quarter. Is that right?
No, Q4 this year.
Q4 this year.
Yeah, forecast.
Q4 this year. Okay. 7 nanometer, C.C. mentioned for the whole year will be about 10%, but since we are ramping heavily from third quarter through fourth quarter, we expect the 7 nanometer will contribute more than 20% in fourth quarter revenue. Meanwhile, we are ramping down 10 nanometer. 10 nanometer contribution in first quarter was 19%, and we expect that number to be single digit in the fourth quarter this year.
Well, Don, I think it's a new trend. I think earlier, we see the customers consolidating. Now we see even smaller company doing the design and tape-outs. That, I think particularly in the AI and networking area. I think the system architecture today is much freer, much more customized for each new application than before. Therefore, each system company, even service provider, are willing to participate for the silicon design in order to optimize their system performance or service quality.
Sorry, just to follow up. Is that because the barrier to entry are lower, or is that those system companies and startups have too much money?
I think both. The barrier definitely lower because all the EDA tools is much more matured today. Also, the venture capital supply of money is definitely, in certain segments is increasing. Mostly, I think the system spec, the chip specification currently is up to the ingenuity of the system designer, not just an off-the-shelf existing products.
Thank you.
All right, next question will be coming from HSBC, Steven Pelayo.
Just a couple quick ones here. On 20 nanometer softness, you mentioned it was a lot of the crypto guys looking to accelerate to more advanced nodes. I'm curious, beyond crypto at 28-nanometer, is it also broader-based weakness at 28-nanometer beyond crypto?
Yeah, we still have You are talking about 20 nanometer?
28-nanometer.
Twenty-eight.
You mentioned that being relatively softer as crypto.
Well, let me give you some example. Actually, this year, we see the highest tape-out number from 28-nanometer in 28-nanometer's history. This year, the tape-out is still higher. Okay. That give you some color on As I said, some of the product move to a high-volume product. Most of them are mobile. All others, they move into this area. The tape-out number is still as high as last year, and last year is the highest number. That is at 28-nanometer. Other than cryptocurrency, there is a lot of thing we are talking about, automotive, IoT, and some of the ISP, those kind of thing moving.
Understood. I just want to make sure the breadth was still there. The other question I have is if you could talk a little bit about manufacturing cycle times going from 16 nanometer to 12 nanometer to 10-nanometer to 7nm. How long is this? When do you need to be starting wafers now? I especially want you to talk about it relative to, I know Elizabeth says we shouldn't pay attention to monthly sales, but you just did a 60% month-on-month, and that is unprecedented in my model. How are we to look at kind of the monthly volatilities as well as the lengthening manufacturing cycle times?
We are improving our manufacturing cycle time from node to node. We continue to improve. There are more layers on each technology node. They increase dramatically, like from 60-some to 70-some to 80 layers. For the cycle time, our fab improved quite a bit, but I won't give you the actual number.
Thank you.
Stephen, you're asking the month-to-month revenue?
I was just surprised at the volatility in the March monthly sales.
Okay. I am like Elizabeth. Don't pay too much attention to it. It's achievements based on customer requirements. There's a seasonality and there's no known requirement. Different customer may be different. It's very hard to predict monthly revenue.
Hi there.
All right. Sorry. The next question will be coming from Nomura's Aaron Jeng.
Hi. After listening to these discussions, I got one question, which I didn't expect at the beginning of the call. Can I say that the [blended hour IC per tape-out] of TSMC will be coming down in the future because of the new trend that more tape-outs from the small IC design system companies? They might not be able to generate big revenue per project. This trend is new, but on my understanding, it seems to be kind of a down trend versus what TSMC was seeing in the past cycle. Don't get me wrong, TSMC is still outstanding. Just compare with what you saw in the past, the new trend seems to be unfavorable to probably anyone in the market.
Well, if you look at our e-beam margin, we welcome more tape-outs, just for the mask making. Of course, I think the trend, we see it as more designer getting into the new product designs. Okay. They are early entrepreneurs. That's how we grow our customer base back beginning. If you look at the cryptocurrency designer, they are very smart people, and the mining machine, they are very intricate system designs. They break all the barriers, pushing the envelopes. Those designer were coming to later on today in cryptocurrency, later on in the AI or blockchain and so forth. That is our ecosystem, part of our ecosystem, we are very happy to see our ecosystem grows to pave way the future product innovation.
Thanks. Well understood. One follow-up. Another angle to support my judgment is that you are increasing the high end of your capacity per year to $12 billion versus in the past, which was $11 billion. Your revenue target is still to grow by 5% to 10%, which means you are spending more for the same target. That's also another angle to support my judgment from the call. That's why I got this thought.
The EUV investment is for the future. Next technology beyond 3 nanometer. Mass making is for the purpose what I said. It's possible that we don't see immediate return, but that is our long-term investment target purposes.
Thank you very much.
Next question will be coming from JPMorgan's Gokul.
Thank you very much. I have one question on AI, HPC, et cetera, and the growth going forward. I think CC mentioned that you're expecting 50% attach rate for accelerators in data centers going forward. That's a pretty high number compared to probably single digit today, or probably low double digit. Couple of years back, you had given a guidance of 5%-10% with half of the growth coming from mobile and half of the growth coming from HPC and other areas. Could you update the component of that guidance? Does it mean that now smartphone is probably not going to grow from here? Is that how you think about it? Almost all of the growth comes from new areas like HPC, IoT, automotive.
Given the very aggressive numbers that you have put out, some of your customers, like NVIDIA, have put out as well a couple of weeks back, what is the tipping point that you need to see to be more confident on areas like HPC?
Mobile is still the most important segment that we are in. However, as I mentioned, the AI and 5G communications are in the future. That this mega trend that makes the HPC is very, very important. So we see a strong demand that will contribute to TSMC's revenue.
Exactly what we can expect that in the future, I would say that they are getting more and more important and higher than we thought that last year, actually. Do you want to add something?
Your question, earlier, the crystal ball we saw before, the smartphone, in 5 years, going to give us 50% gross TWD. Now it looks smaller, like a little bit more than 40%. HPC used to be 25%, but it looks like it's going to be close to 40%. HPC, as we look forward, it seems stronger than we saw before, and smartphone a little bit weaker than before. IoT and automotive, it's about the same. To be honest, this goes with the time. Even in smartphone, we look at the 5G transition. That will be another industry transition. If you look at the 4G transition, the demand has changed the whole landscape. That's what we look at today, and we'll go along with the industry doing the migration and update for you.
If I could just ask for some more perspective on this. A lot of new tape outs, increasing mass capacity, feels like the beginning of another new product cycle, typically. Could you compare how this looks like compared to, say, 10 years back at the beginning of a smartphone product cycle? Probably a lot of your current big customers were much smaller at that point in time, asking for more tape outs as well, and mass capacity. Could you compare how it feels like today versus, say, 10 years back in the beginning of the smartphone cycle? Thanks.
We don't know. It's just like smartphone. When first smartphone came, we did not know. It really depend on the usage model and application of those products. We just cast a wide net and make sure we support all the innovators.
Thanks.
All right. Questions will be coming from Credit Suisse, Randy Abrams.
Yeah, thank you. One follow-up question on the gross margin. With the midpoint at 48% in the second quarter, how should we think about second half, where you normally have the peak season? You're also implying at least half on half growth. Is it still the target? I think in the past, you've said close to 50%. Is that still kind of the range we'd get some leverage first half to second half?
I think the foreign exchange rate, if you compare with last year, will be a hit for the whole year. That's number one. The second half, we are expecting more 7nm coming on the line, which will have 2-3 percentage point dilution to corporate margin. These two factor are negative.
Yeah.
I think you should not assume everything will be same as 2017.
Okay. Then one question on the tool reuse. N7+ has a lot of reuse. For 5nm, where it's a bigger change with EUV, does that change the equation that it's different, or do you think you can keep that same conversion down to 5?
We are seeing every generation new migration will be utilized more than 90% of the previous generation tool. The convertibility is pretty high, even for 5nm.
CL Securities, Sebastian has a follow-up.
Thank you. Just one. On the inventory turnover days, if you compare it on the worldwide perspective, increased almost 20 days. Lora already mentioned about this, some raw wafer inventory pre-built for seven nanometers. I wondered how much of this is just a pre-build for the seven-nanometer ramp, or how much of it is because the raw wafer is more expensive today, so we want to buy more now?
Pre-build has higher weighting than a raw wafer. Raw wafer will be a few days. The remaining are the pre-build and the complexity in technologies. You're comparing to a long time ago, right? Say, 20 days difference.
Does that mean that you don't need to buy as many wafers in following quarters?
I don't quite get your questions.
You already prepared some of the raw wafer right now for second half ramp. Does that mean that you don't need to buy as many as raw wafers in second-
That's right
half this year-
That's right
compared to first quarter?
Thank you.
A follow-up question from Sinolink's Andrew.
My follow-up question, first one is regarding CapEx to sales ratio. I remember last time, Dr. Chen mentioned the long-term will stay at 20%-30%. For this year, we raised a little bit, it's about 32%, 33%. Are we still staying this assumption for the next few years, 20%-30% CapEx to sales ratio?
That's still true. At 30% level, around 30%.
We change from 20%, 30% to 30% now.
It was within 25%-30% anyway. This year it will be higher.
For the reason I was just mentioning, slightly above 30%.
Yes.
If you look at a couple of years down the road, it will be around 30%. Some years may be lower, slightly lower than 30%. Some years as 30%.
More like a 25%-30% range.
Yeah. That's still true.
Understood. Thank you.
All right.
My last question is.
You have another question? Okay.
Are we really gaining back the customer we are losing on 14 nanometer and 10-nanometer for seven? I did not mention any customer name.
We are the only seven nanometer provider, so you can imagine that the answer of your questions.
Thank you.
Well, I'm just going to, for the interest of time, only going to limit two questions, so one person each. Okay? Citi's Roland first, and then Deutsche Bank's Michael. One question each.
Thank you. For Citi, now you are ramping up 16 nanometer production in China. What are you going to do if customer needs to do the InFO or CoWoS packaging layer? Are you going to build the InFO or CoWoS capacity layer, or are you going to logistically, you have to ship back to Taiwan? This is for InFO and the CoWoS.
Well, that will depend on customers' demand and customers' need. We always support them as much as possible. If we can support locally, we will do it. But today.
Not today, yeah.
Today is not in our plan yet.
Okay. When? When are you going to build a CoWoS or
Depend on customers' demand.
Okay. Thank you.
Michael.
mentioned, seven nanometer will be very massive, will most of customers shift to five nanometer, or they will skip five nanometer and shift to three nanometer directly?
Your question is, are those 7nm-
Customer-
Custom-
Will shift to 5nm, or there will some customers just-
Well, we-
skip 5nm.
We developed the technology working with the customer to meet their product design, their product spec and with their product's performance. Whether they move to five or not, it depend on their product's nature. Some of them, they definitely will move. Some of them probably will stay in seven, N7+.
In the past year, 20 or 10 seems to be a short node, right? Will we see the 5 nanometer a short node or?
No. 5 nanometer will be a very long-lived and useful and very cost-effective.
Will start in 2020. Okay. With that, we will conclude our today's conference. Please be advised that the replay of the conference will be accessible within 3 hours from now. Transcript will become available 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.