日月光投資控股 (ASX) 2020 Q3 法說會逐字稿

Hello. I am Ken Hsiang, the Head of Investor Relations for ASE Technology Holdings. Welcome to our third quarter 2020 earnings release. Thank you for attending our conference call today.

Please refer to our safe harbor notice on Page 2. All participants consent to having their voices and questions broadcast via participation of this event.

I would like to remind everyone on this call that the presentation that follows may contain forward-looking statements. These forward-looking statements are subject to a high degree of risk, and our actual results may differ materially. For the purposes of this presentation, our dollar figures are generally stated in New Taiwan dollars, unless otherwise indicated.

As a Taiwan-based company, our financials are presented in accordance with Taiwan IFRS. Results presented using Taiwan IFRS may differ materially from results using other accounting standards.

I am joined today by Dr. Tien Wu, ASE Holdings COO; and Joseph Tung, ASE Holdings, CFO. For today's call, I will be going over our financial results, Tien will be providing a market overview, and Joseph will provide a recap in our guidance. We will have a Q&A session following the prepared remarks.

As with the rest of 2020, the third quarter has proven that there is never a dull moment, especially in the electronics industry. The third quarter was remarkably eventful for us. Typical seasonality ran through August with run rates reaching historical highs. However, in September, the U.S. Bureau of Industry and Securities' Export Administration Regulation, EAR for short, went into effect. As a result for our ATM business, we commenced to replace capacities left open by exiting business. This process occurred much more quickly than anticipated as our overall loading levels snapped back to full utilization.

During the quarter, we took 2 charges related to the EAR, one, in cost of goods related to unused inventory; and one and nonoperation for interface boards used specifically for EAR-impacted customers. We also experienced near-term highs in the value of the NT dollar relative to the U.S. dollar.

Meanwhile, our EMS business ramped up a bit later in the year, in line with a deferred seasonal pattern. We don't believe the EMS business has yet peaked for this manufacturing season. In total, we ended the quarter strong, with strength across our core SiP, advanced packaging and wirebonding products.

Please turn to Page 3, where you'll find our third quarter consolidated results at the holding company level. In this section, we will generally defer business explanations to our ATM and EMS P&L discussions. Intercompany transactions between our ATM and EMS businesses have been eliminated during consolidation.

For the third quarter, we recorded fully diluted EPS of $1.54 and basic EPS of $1.57. This means that on a year-to-date perspective, we have fully diluted EPS of $4.01 and basic EPS of $4.12, exceeding full year 2019 EPS already. Consolidated net revenue was $123.2 billion, representing a 15% increase quarter-over-quarter and a 5% increase year-over-year. We had a gross profit of $19.7 billion with a gross margin of 16%. Our gross margin declined by 1.5 percentage points quarter-over-quarter and 0.3 percentage points year-over-year. The sequential margin decline is primarily the result of EAR impact, a stronger NT dollar environment and a higher EMS business mix. The year-over-year decline is primarily the result of EAR impact and the strong NT dollar.

Our operating expenses increased by $0.2 billion during the third quarter to $10.6 billion as a result of higher operating expenses in our EMS business unit, offset in part by lower operating expenses in our ATM business unit. Despite the increase, our operating expense percentage declined 1.1 percentage points sequentially and 0.5 percentage points year-over-year to 8.6%. This amount is currently trending below our 2018 operating expense percentage.

Operating profit was $9.1 billion, up $0.7 billion sequentially and year-over-year. Sequentially, operating margin declined 0.4 percentage points to 7.4% while increasing 0.3 percentage points year-over-year. During the quarter, we had a net nonoperating loss of $0.1 billion. This amount primarily consists of net interest expense of $0.7 billion and an EAR-related tooling impairment of $0.7 billion. This amount was offset in part by net foreign exchange, investment and sale of equipment gains.

Tax expense for the quarter was $1.8 billion. The effective tax rate for the third quarter was 20%. Net income for the quarter was $6.7 billion, representing a decline of $0.2 billion sequentially and an improvement of $1 billion year-over-year. We believe it's important to note the operating margin impact of the strengthening NT dollar and the EAR-related write-down. Removing the inventory charge while using the second quarter exchange rate, we estimate an operating margin of 8.6%. And similarly, using the third quarter 2019 exchange rate, we estimate an operating margin of 9.7%. Without inclusion of the EAR-related inventory and equipment write-down totaling $1.6 billion, we would have basic EPS of $1.84 during the quarter.

On the bottom of the page, we have again provided key P&L line items without the inclusion of PPA-related expenses. Consolidated gross profit excluding PPA expenses would be $20.6 billion with a 16.7% gross margin. Operating profit would be $10.3 billion, with an operating margin of 8.3%. Net profit would be $7.9 billion with net margin of 6.4%. Basic EPS excluding PPA expenses would be $1.84.

On Page 4 is our ATM P&L. It is worth noting here that the ATM revenue reported here contains revenue eliminated at the holding company level related to intercompany transactions between our ATM and EMS businesses. The third quarter for our ATM business was incredibly busy. Usually, that's good. We do a lot of work and we get a lot of revenue for it. We had a different kind of busy this quarter in, which we were busy stopping EAR devices, retooling factory lines and restarting replacement devices. This happened at a friendly pace. As a result, despite the EAR disruption, we still saw measured growth. That's actually a significant achievement given the size of the business lost. More about that later from Dr. Wu.

After the third quarter finished, our ATM business outperformed our initial expectations rather significantly. In retrospect, we are somewhat surprised at the efficiency of the supply chain and the pace at which our capacity refill has happened. The refill was not the linear recovery we had initially anticipated. Instead, business snapped back with new products rushing to replace vacated products.

During the quarter, our ATM business continued seeing a strengthening NT dollar environment, in which the NT dollar appreciated 1.6%. Given that our orders are generally denominated in U.S. dollars, while our factory costs are mostly denominated in NT dollars, a strengthening NT dollar brings a higher cost structure for us. If the NT dollar stays strong for a longer term, we believe that we may be able to adjust pricing to compensate and purchase relatively cheaper U.S.-denominated machinery equipment. Shorter-term movements are more difficult to position.

For the third quarter 2020, revenues for our ATM business were $71.8 billion, up $2.3 billion from the previous quarter and up $3.9 billion from the same period last year. This represents a 3% increase sequentially and a 6% increase year-over-year. Our ATM revenues came in somewhat ahead of our expectations due to stronger snap back of revenue post-EAR impact.

Gross profit for our ATM business was $14.5 billion, down $0.5 billion sequentially and down $0.2 billion year-over-year. The sequential and year-over-year gross profit decline was primarily related to a onetime inventory write-off of EAR-related customer substrate of $0.9 billion and a stronger NT dollar.

Gross profit margin for our ATM business was 20.2%, down 1.5 percentage points sequentially and year-over-year. Margin decline was primarily attributable to EAR and NT dollar impact.

During the third quarter, operating expenses were $7.7 billion, down $0.1 billion sequentially and $0.6 billion year-over-year. The sequential and year-over-year declines were driven by lower administrative costs. Our operating expense percentage was 10.8%, down 0.5 percentage points sequentially and down 1.4 percentage points year-over-year.

During the third quarter, operating profit was $6.8 billion, representing a decline of $0.4 billion quarter-on-quarter and an improvement of $0.4 billion year-over-year. Operating margin was 9.5%, declining 0.9 percentage points sequentially and improving 0.1 percentage points year-over-year.

For gross and operating margins, the onetime EAR inventory write-off had a 1.2 percentage point impact. We estimate that the strengthening NT dollar also had a 0.8 percentage point impact to gross margin sequentially and a 2.7 percentage point impact year-over-year.

Without the impact of PPA-related depreciation and amortization, ATM gross profit margin would be 21.5% and operating profit margin would be 11.1%.

On Page 5, you'll find a graphical presentation of our ATM P&L.

On Page 6 is our ATM revenue by market segment. Not much has changed here.

On Page 7, you will find our ATM revenue by service type. As stated earlier, capacity snapped back to running near full outside of certain capacities requiring longer NPI time, and those NPIs are in process. From this chart, you can see here that our wirebonding business is performing particularly well. We believe that our wirebonding business is seeing a resurgence of demand. More about this from Dr. Wu later.

On Page 8, you can see the results of our EMS business and its associated revenue by application. For our EMS business, the third quarter usually represents the peak quarter in terms of seasonality. However, we anticipated a somewhat delayed manufacturing cycle. With that in consideration, demand for our EMS business was stronger than anticipated, driven by strong SiP demand.

During the third quarter, we had revenues of $53.1 billion, increasing 34% sequentially and 5% year-over-year. EMS revenues increased quarter-over-quarter primarily because of our seasonal business ramp. EMS revenues increased year-over-year primarily as a result of stronger demand for SiP products, offset by a somewhat later seasonal product cycle.

Our EMS gross profit was $5.1 billion, improving $1.4 billion sequentially and $0.7 billion year-over-year. The sequential and year-over-year gross profit improvements were driven primarily by stronger customer demand for SiP-related products. Gross profit margin for the EMS business came in at 9.7%, an improvement of 0.3 percentage points sequentially and 0.8 percentage points year-over-year. The margin improvement is primarily the result of product mix changes.

Our EMS business unit's third quarter operating expenses were $2.8 billion, increasing $0.3 billion sequentially and $0.4 billion year-over-year. Operating expenses increased primarily as a result of increased employee profit sharing. Operating expense percentage was 5.3%, dropping 1 percentage point as compared with 6.3% last quarter and increasing 0.5 percentage points year-over-year.

Our EMS operating profit for the quarter was $2.3 billion, representing a $1.1 billion improvement sequentially and a $0.2 billion improvement year-over-year. The sequential operating profit improvement was primarily due to increased seasonal demand. Our EMS operating margin was 4.4%, which is a 1.3 percentage point improvement sequentially and a 0.3 percentage point improvement year-over-year.

On the chart on the bottom half of the page, you'll find a graphical representation of our EMS revenue by application. Our consumer segment picked up seasonally. And this season, we expect to add an incremental SiP product. We would expect that this segment continues to pick up into the fourth quarter.

It is again worth noting that our EMS business unit runs under the name Universal Scientific Industrial and is traded as an A share on the Shanghai Stock Exchange under the ticker number, 601231. We currently own 75% of the company, which translates roughly to USD 5.9 billion.

In regards to our regulatory filing to complete our acquisition of Asteelflash, the current COVID-19 resurgence in Europe is impacting the duration of our regulatory reviews. As of yesterday, France announced its second country-wide lockdown. As a result, we currently expect the completion of our combination with Asteelflash to be somewhat delayed. We now expect for the regulatory process to complete before year-end.

On Page 9, you will find key line items from our balance sheet. At the end of the quarter, we had cash, cash equivalents and current financial assets of $61.8 billion. Our interest-bearing debt increased $7.1 billion to $224.6 billion. Total unused credit lines amounted to $255.6 billion. Our EBITDA for the quarter was $23.2 billion. We continue to target a net debt-to-equity ratio of 60% to 65% by the end of 2021.

On Page 10, you will find our equipment capital expenditures. Machinery and equipment capital expenditures for the third quarter in U.S. dollars totaled $415 million, of which $288 million were used in packaging operations, $73 million in testing operations, $52 million in EMS operations and $2 million in interconnect materials, operations and others.

2020 is providing us an unusual situation. We understand that there is an expectation of perfect fungibility, where replacement business uses entirely the same package types or tester platforms and requires no incremental tooling. And in such a perfect scenario, incremental capital investment becomes completely unnecessary when replacement business comes onboard.

Unfortunately, this perfect scenario almost never happens. One customer may have used a fan-out process when another one uses bumping and flip chip. Customers may even use the same model tester, but with different instruments or configurations. As a result, in many cases, we have to make smaller investments on tooling our instruments to load previously purchased larger investments. In addition to facilitating the refill, we are seeing a significant pickup in our demand in wirebond-related capacities. As a result, we do see the need to invest in our wirebond lines. Tien will speak shortly on this also.

On Page 11, we have a brief year-to-date recap. All information here is presented on year-to-date terms. Holding company revenue grew 15% year-over-year on U.S. dollar terms. ATM revenues grew 19% year-over-year on U.S. dollar terms, with gross margins improving 1.9 percentage points. Removing the impact of currency and EAR-related expenses, gross margins improved 4.2 percentage points. EMS revenues grew 12% year-over-year and U.S. dollar terms. Year-to-date EPS is $4.12.

For an update of the overall market environment, I'll now turn the microphone over to Dr. Tien Wu.

This is Tien Wu. I would like to offer you a business environment.

If you look at Page 12, the assembly capacities are tight. In particular, wirebond capacity is extremely tight. The tightness situation, we expect that to last at least until Q2 of next year. In this particular environment, because we're seeing the tightness in wirebond as well as in all other assembly capacities, we believe the ASP environment will be friendly in 2021. In fact, we will start seeing the margin improvement starting Q4 this year, but we do expect the ASP environment to be more friendly in 2021 comparing to the previous years.

Let me talk about the sectors from our perspective. In the communications sector, we have 5G driving a portion of the growth. We also have a lot of WiFi 6 standards driving the multiple upgrade cycles in communication as well as in automotive. We have seen a slowdown in automotive in Q1 and Q2. In Q3, we're seeing a remarkable recovery and strength in automotive sector. We believe this will be reflected in Q4 as well as next year. Computing and consumer demand have been strong, and they remain strong for Q4. We also believe the computing and the consumer demand will be strong for 2021.

A lot of investors are asking us about the COVID-19 effect, and I would like to share our perspective with you. The COVID-19 has created new value for technology products. Namely, the technology products are viewed to be an alternative to reduce medical risk, which I will elaborate a little bit more; also, facilitating social connectivity in addition to the traditional value of digital efficiencies.

What I was referring to is because of work from home, because of learn from our home, people are buying IT products to minimize and reduce the medical exposure and medical risk because of COVID-19. Because of COVID-19, the social connectivity will be augmented by the IT products. So in this scenario, we're seeing 2 fundamental changes in consumer behavior. First, more people tend to buy IT products. Second, people are willing to pay higher price to buy IT products with performance. It is in this regard we believe we are seeing, during the COVID-19 days, why the IT products in almost all sectors are showing particular strength, especially the communication products.

So in that regard, with the 5G, with the WiFi, with all the upgrade cycles, which were permeated through the cell phone, the automotive as well as PC, Bluetooth, and obviously, IoT, we believe there are fundamental changes in the way people are willing to spend money to buy IT products.

Also in the product mix, we're clearly seeing the product mix as well as product complexity increasing. In particular, in the wirebond, the number of stacked dies that we're doing now is more than before. The type of product we're doing for RF, for analog, and we're seeing more multiple dies require wirebond, which we have not seen before. So in this particular cycle, it's not just the volume, it's also the number of dies, the number of wires as well as the complexity. On top of that, because this product, some are going to medical devices and automotive, the kind of quality requirement is also different from the consumer. That is why for quality wirebond service, we can command a premium in this year, and I believe, in 2021 as well as years going forward.

Let me comment about the OSAT overall CapEx expenditure. If you go back to 2018, 2019 as well as the 2020 number, which we do not have the complete view, you will see that the overall OSAT industry, we believe, has been underinvested. In that particular scenario, ASE has been spending CapEx in 2018, in 2019 as well as in 2020. Going forward, we will accelerate spending in Q4 such that we can put more capacity in place in anticipating as well as the -- fulfill the commitment based on our customers' demand. Going into 2021, we will be moderating on the CapEx, which Joseph, our CFO, will talk a little bit more.

Based on the new design wins as well as all of the long-term contracts and forecasts, as of today, we believe we will have a strong 2021, and we're quite optimistic about that. Once again -- which I'll be happy to talk to you about the -- our growth as well as market share gain. In our calculation, if we exclude the EAR impact, we believe ASE is gaining shares in all sectors, in all package types.

Please turn to the next page, Page 13. Ken already talked about the painful experience and the sudden nature of the EAR action. The EAR affected ATM revenue in Q1 and Q2 were about 20% compared to our overall ATM run rate. In Q3, it was down to about 13%. In Q4, it will be 0. Greater than 75% of the lost revenue and the lost capacity has been backfilled, which Ken already talked about it, by many other customers with the help through all of our partners throughout the supply chain and customers. The remaining 25% will be backfilled by end Q1 2021.

So as of today, we're confident that our year-over-year growth in 2021 will still be positive. In other words, all of the tooling, all of the asset disposition has been completed either in Q3, or the latest will be completed by Q4 of this year. Ken already talked about it. We had a onetime write-off, which was included in the Q3 number that we just reported. With all of the additional work in Q4 or Q1 of next year, we will not incur any additional charges.

With that, I will turn the floor to our CFO, Joseph Tung.

Okay. Good afternoon, everybody. And this is Joseph. Before I get into the further comments on the financials, I would like to give a very brief comment on quarter 3.

Certainly, we have successfully managed through a very chaotic third quarter and came out with a much stronger-than-expected quarterly results. It was with close collaborations among different operating units that we can now substantially reduce the negative impact of the EAR restriction and quickly regain our momentum going forward.

Now with that, I want to take a few minutes to update on the progress we are making and some of the financial targets we set out to achieve at the beginning of the year. First, on OpEx ratio. Our target is to go back to 2018 level of 9.4%. In quarter 3, it had come down to 8.6% from 9.7% a quarter ago, way below the target level. Now with continuous effort going into Q4, I strongly believe that we will reach -- we will not only reach our target, we're more likely to exceed that.

Second, on OPM, or operating margin, we stated we shall see 2% improvement in the year. And I believe we are ahead if we take out the negative impact from NT dollar appreciation and EAR-induced inventory write-down. Although it is difficult to quantify, the overall improvement of operating margin came from synergy -- part of the overall operating margin improvement came from synergy created between ASE and SPIL through increasing coordination. Going forward, we will further deepen and broaden our -- such coordination on various parts of our operation, including capacity alignment, business development, procurement and R&D.

Thirdly, on CapEx. After heavy investment in the past 3 years to support our strong business momentum going forward, our CapEx in 2021 should start to moderate. At this point, I believe 2021 CapEx amount should fall between 2018 and 2019 levels. We're more leaning towards 2018 level. With the improved profitability and reducing CapEx, our cash flow position in 2021 will see good improvement. Therefore, it will allow us to increase our cash dividend payout to no less than $3 per share while still reaching our deleveraging target of 60% to 65% net debt-to-equity ratio by end of 2021.

Okay. Now with that, let me give you our fourth quarter guidance. Based on our current business outlook and exchange rate assumptions, management projects overall performance for the fourth quarter of 2020 to be as follows: In NT dollar terms, ATM fourth quarter 2020 business should be similar with the first half 2020 level. ATM fourth quarter 2020 gross margin should be similar with first half 2020 level.

On EMS, in NT dollar terms, fourth quarter revenue, the business sequential growth rate should be similar with the average of second and third quarter 2020 levels, while EMS fourth quarter 2020 operating margin should be slightly better than the average of second and third quarter 2020 levels. It is a bit complicated, but I'm sure everybody will figure that out. Thank you very much.

Now we would like to open the floor to Q&A.

(Operator Instructions) The first to ask questions Gokul Hariharan, JPMorgan.

Congrats on managing through the situation in Q3 and the good results. Quick question on IC ATM. Could we talk a little bit about -- we talked about some price adjustments potentially. And could we talk a little bit more in detail about what are the discussions you're having with customers? And given that this year, currency has clearly affected margins and taken away a fair bit of the margin improvement, could we talk about the conversations that you're having with customers on potential price adjustments for the currency side? That is my first question.

Second is we do hear that there is some degree of capacity shift happening towards non-China OSATs. What does ASE see in terms of your customer base, both for advanced packaging as well as wirebond? And how should we think about capacity build for ASE, given that you also sold a small fab in China while announcing meaningful investments in [Kaohsiung]?

To answer your first question, we will not comment on any specific customer engagement. However, in the overall scenario, the wirebond gap as of today is anywhere between 30% to 40%. It is not a 3% to 4% gap. The gap is quite substantial. ASE today, we have the largest installment base in wirebond as well as the best efficiency and quality. A lot of customers, either they're moving business from somewhere else, or just based on organic growth or the new product ramp, the demand is very, very strong.

The kind of conversation that we're having is, for example, there is a lead frame and substrate cost increase and the gold price -- there's a gold price [go adder]. Whenever there is an expedite delivery, the GAAP cost will be reflected. Particularly, we are engaging with multiple customers in a much longer service contract. For example, we will talk 6 months, 1 year, and in some cases, we're talking about 2 years take-or-pay.

In other words, all the capacity we have put in place since 2018 has been put to use. Right now, we're cautious in terms of adding capacity because we understand the tightening situation. We need to be a little more cautious. But all the capacity right now we're adding do have long-term service contract behind it. And we believe now even with the pace that we're having, the industry supply -- excuse me, I'm sorry. Are you asking a question? Or are you making a comment?

No, no. I'm on mute.

I don't think that was Gokul. That was the operator. Operator, can you put yourself on mute?

Yes. I'm sorry. I apologize for the interruption.

Okay. Right. So in that particular scenario, we believe the assembly equipment lead time right now is quite long. And I don't believe we are overbuilding capacity in any stretch of the imagination, which is why I made a comment the tightness situation will last at least to Q2 of next year, right? So that answered the first question.

For the second question, I can't really comment which one of our customers could be moving from China. I believe the movement is both sides. Now obviously, we do have customers who intended to have more of the supply chain be outside of China. I mean we do have seen those cases, but I believe we also have other cases.

Just one confirmation. So 30% to 40% is the supply-to-demand gap in wirebonding, right?

As of today, that is correct. But I am making a comment on behalf of ASE.

Next one to ask questions, Bruce from Goldman Sachs.

So my question is mainly for Tien. I think that from our perspective, we are happy to see that the management is commenting on some of the dollar content growth for the packaging. Can you elaborate more? Because what we saw here is that for something like 5G smartphone chip, we see a lot of dollar content growth from the wafer side, from the production side. But for the packaging and testing, maybe a lot of different testing, but the entire dollar content for the packaging is not growing or it's growing a lot slower than the wafer at the foundry side. So can you comment a little bit more about what kind of dollar content growth for the packaging moving forward at an industry basis?

That really is a very difficult question because the -- I think, in general, people are complaining that the packaging is becoming a higher percentage of cost within the total. The -- I believe if you comment based on the automotive, clearly you see more semiconductor in the latest, the 5G phones. You've seen the AiP. You have seen the front-end module. You have seen the PA. You're also seeing more semiconductor content.

But in terms of the packaging versus the overall increase in semiconductor content, my perspective is you will see that percentage going up. I don't believe the packaging costs should be going up as fast as foundry because the foundry, the investment versus return, the business model are completely different. But I do believe, because of the complexity and the nature of the heterogeneous integration, you will see the packaging content, when you normalize things, I do believe that trend is going up.

Yes. But what's difficult for us to see that from the revenue growth side because -- for example, like smartphone chip, it's so clear for the wafer side, but it's not clear at all for the packaging side. So it's difficult for us. Can you give us certain quantitative analysis for that?

I don't think I have that number. My apologies.

No worries. So the next question -- I'm sorry?

Yes, go ahead.

Okay. So the next question is that one of your supply chain got serious fire a couple of days ago. What kind of impact to your business? Because one of the key customers is your key customers as well. So what kind of business impact you included in your forecast for the fourth quarter?

Well, thank you for asking that question. We will not comment on that particular incident because we're still going through the clarification. So we're waiting for further report.

But having said that, we do have multiple important customers who are affected by this particular fire. But just like all of the supply chain scenario, when you have a position, chances are, it is easier for us to go through the supply chain and get replacement as well as get priority. During the last 2 days, we have collaborated with all of our customers. So I can tell you with confidence in Q4, our revenue has already been factored and the impact is less than 1%. So that has already been factored in.

Today, we're working on the alternative supply for all of the key customer in the -- of our global, the substrate suppliers. We believe the situation is complicated, but like everything in the past, this is not the first time we encountered supply chain disruption. We had much worse scenario and we handled that pretty well. In this round, I hope our particular partner in Taiwan can recover soon. But as of today, the situation is manageable.

Okay. Can I ask please one more question, which is for the ASP? Management was talking about ASP will be friendly in 2021. Does that friendly situation only happens in wirebond or it happens in chip testing or overall? And what about the SiP pricing environment?

All right. I guess the best way to answer is the -- it's more friendly in wirebond. Other products are also friendly compared to last year. All right. My apologies. That's pretty much all I can tell you.

How about SiP?

SiP, same because chances are when you have like an allocation in a particular package type, it basically cascaded down to all other products.

Now the line is open to Szeho Ng, China Renaissance.

My first question is regarding the CapEx. You mentioned that next year's CapEx will be going back to the 2018 level. That would represent quite a sharp drop compared with this year's level. So I just wondered which areas are you holding back the investments for next year.

Well, I already made a comment. If you look at the ASE in 2018, we started a major investment ramp. We call it the super cycle. In 2018 we invested. In 2019, when everybody backed off, I forgot the number, the -- I think our number is $1.4 billion, $1.5 billion. We ratchet up. If you go back to the 2019 total, all the other 19 OSAT combined, we spent USD 700 million less. I think ASE spent $400 million or $500 million more.

As a result of that in 2020, even in the early days when people are not exactly sure, we just got the -- another super cycle. That's why Q1, Q2, Q3, we managed to spend good amount of dollar. I'm not sure we had reported a number but -- we did. But in Q4, we will continue to spend.

Now after the 3 years of super cycle, we believe we have all of the technology. We have all of the basic elements in place. Of course, what we can do is we can continue to expand in 2021 and beyond. But another alternative is just look at all of the capacity based on what we have and then look at it from a synergy perspective, look from a customer base perspective and how do we optimize and maximize the capacity that we already put in place.

Also, mind you, over the last 3 years, a lot of the CapEx are going to the automation. So as of today, ASE has 18, the line of factories. And we're trying to explore and trying to expand the efficiency advantage of those line of factories. So between all of these things, our current view is we should probably more moderate. We're not really reducing CapEx. We're reducing CapEx back to the normal level, but we're not doing the super cycle anymore, so if that helps you.

But of course, any business environment can change. If our customers are demanding we need to do more, then we do more, right? This is just based on -- and we also believe our OSAT competitor will probably realize this again, and they might spend more in 2021 going forward. So it really depends on how other -- I mean just do a contrarian approach. When people are not investing, we invest. We just want to stay ahead of the curve. But if we need to invest, we'll continue to invest.

Okay. Got you.

And a little bit of backdrop of the moderating CapEx in 2021 is that we are actually upping our CapEx this year going into Q4 because of the overall retooling and realignment and also the -- we're trying to fill the gap that exists today on wirebonding. So the overall CapEx for this year is -- actually is going to exceed our original target.

Okay. Got you. And second question, when we go to the heterogeneous packaging level, yes, so what role are you going to play in that area?

I apologize. You're asking the testing? Can you repeat the first part? Because it wasn't very clear.

I'm talking about advanced assemblies, let's say, in the heterogeneous packaging. What roles are we playing when we go to the heterogeneous packaging industry or the packaging?

All right. When we talk about the heterogeneous integration, it's such a generic term. Everyone uses it, right? For the packaging industry, what we're referring to is silicon passes, sensors, optical, audio, all kinds of intrinsic, different components and functionality. And we will call that the heterogeneous integration.

Now if you really look at all of the SiP product that we're building, whether it's in the optical space, in audio space or consumer space, we tend to have a very diversified functionality and integrated through packaging. So this is slightly different from the SoC approach. Or when you put memory chips and logic chips and pack them together, that's also HIR, but it is slightly different because the chips are chips. There's large chips, small chips and different functionality of chips, but they're still chips.

So the chips, the mechanical and physical behavior will be very, very similar, which is why it is expensive but it's easier to do integration. For the packaging, you really have to think about generically, intrinsically different materials, different physical, different mechanical, material characteristic, and we put that together. I think that's what packaging HIR was referring to.

Okay. All right. Got you. And my last question maybe on the financial part. What is the company's gross margin sensitivity to the FX? Any ballpark indication would be helpful.

For each percentage point appreciation of NT dollar, that will have a -- on the ATM side, will have a 50 basis point impact on the gross margin. But on the -- at the holding level, the impact would be 30 basis points.

Next one to ask questions, Roland Shu from Citigroup.

My first question is you talked about for your EAR order impact. 75% of demand has been backfilled and the 25% will be backfilled in first quarter next year. So question is what kind of demand had been delayed backfill to first quarter next year? And how confident are you for this demand will be backfilled in first quarter this year?

Okay. The -- we're seeing strength from all sectors, which I've reported. But given the nature of the capacity that became available, right, after the September 15, we are entertaining a lot of the customers from our communication sectors. Now we have also done retooling. For example, the -- a lot of the bumping facility capacity can be retooled to serve customers in other sectors. The 75% has already been backfilled. We have a very high confidence the remaining 25% will be backfilled by -- in Q1 of next year because we are currently in a capacity constraint.

It's a matter of how fast can we retool, how fast can we do the CapEx, upgrade the card. There's a lot of details that we're doing through. So the -- I think we have been extremely busy for the last few months just trying to work with all of our customers. But the customer demands are strong. And all of them have given us very strong indication, either by contract or by a firm commitment that this will be the case.

Understood. So this is mainly constrained by the capacity. I think my second question is similar to this. When I look at your 4Q guidance, especially for IC ATM. When I combine your 4Q IC ATM revenue with 3Q, together, I think the second half total ICT revenue increased about low single-digit percentage point year-over-year. However, if we compare with the TSMC and UMC, I think TSMC, its second half revenue is growing by 17% year-over-year. And for UMC, its second half revenue is growing by 13% year-over-year. But for your revenue, ICT revenue only grow by low single-digit percentage point.

So is this because of the capacity constraint? Or is this because of the competition, or you have other reasons you have this lower growth in second half this year?

The capacity and the -- let me answer it this way. The -- I talked about the EAR impact. The run rate in Q1 and Q2 was 20%. In Q3, it dropped to 13%. In Q4, it dropped to 0. So literally in Q4, we had 13% run rate taking out. And we're trying to compensate the revenue by changing the capacity in real time. And that has been a challenge for the operation unit. The customer qualification, given the complexity and also the quality requirement, that also requires the timing.

So overall, if you look at the full year performance, and I think we are reasonable, comparing to the OSAT industry. If you take the EAR effect out, I think our revenue was higher than all of the other competitors, especially the way we count the SiP revenue. We will count the assembly portion of the SiP revenue, where the higher material content of the SiP revenue, we will include that in the USI unit.

So when you combine all of these facts, then you understand the -- from the assembly perspective, we are gaining shares and we're not lagging behind. The EAR does cause a onetime headache, but we're behind that right now.

Now when you ask comparing to TSMC and UMC, I don't think that is an apple-to-apple comparison. Technology content is different. The investment is different. So I will not compare directly between the foundry and the OSAT.

But in the past, these are highly correlated, correlation of the foundry and the OSAT business revenue, especially for UMC. I think from their -- most of their product that need to go to the OSAT are for the assembling and testing. So I think for UMC's number, I think probably this actually is relevant.

Okay. Well, let's repeat again. Excluding the EAR, the second half of '20, we're up 20% in U.S. dollar terms. I'm not exactly sure the UMC number, but you should -- you can make that comparison.

Yes. I think that's more of an apple-to-apple comparison between us and UMC.

Understood. Okay. May I ask how about the utilization in 3Q and 4Q for all product lines?

In Q3, in terms of assembly -- in terms of packaging, we're about 80% plus. In terms of testing, it was around 80%.

Yes. How about the Q4?

Going into Q4, I think packaging-wise, we will remain at 80% plus, while tests will come down a bit from -- to about 70% to 75%.

What's the reason for testing revenue -- utilization to go down? Because we have more interest or...

I think part of the EAR impact is that the EAR-impacted customer has a higher turnkey ratio.

Understood. Okay. Okay. My last question is one of your key customers have been -- delayed launch is a smartphone by several weeks. So how do you look at your EMS seasonality in first quarter next year?

EMS Q1.

I think overall -- I think Ken mentioned that there's a kind of a delay in terms of getting -- for EMS to get to its peak quarter. And so with that as a backdrop, I think quarter 4, we will see a very strong quarter in the EMS. And quarter 1, of course, there will be a seasonal decline, which is the normal pattern. But all in all, it still is going to be a very -- relatively speaking, a very strong first quarter, comparing to previous years.

But then the kind of the seasonality you said will be similar as the previous years. Is that right?

I think for quarter 1, coming off a very normally strong fourth quarter, the seasonal movement will be very similar to previous years.

Are you speaking about EMS or...

Yes. Yes, I'm talking about EMS specifically.

That's all we have there.

Okay. Okay. Got you. So similar?

There will be a similar type of movement coming off a very strong -- abnormally strong fourth quarter.

Yes. Just to add on to that. Our -- it's also a little bit early on the EMS side for the first quarter. So at least from the ATM side, we do see probably an improved environment, better than typical seasonality. But for the first quarter on EMS, we don't have as much visibility on that in that front.

Next one to ask questions, Randy Abrams, Crédit Suisse.

Actually, maybe one quick follow-up on that point. You mentioned the new consumer SiP project. Just curious when we should factor that in because I think you just mentioned seasonal decline. But is that a product coming in early? And is it material enough when it comes in, it could allow you an incremental boost beyond seasonality?

I think we're having a very strong year for SiP, both in terms of overall volume as well as obtaining new projects. I think we set out to say we have a target of new project incremental revenue to be $100 million a year. It seems that we will more than triple that target this year. And momentum continued to be strong. We continue to gain or engage in some of the new projects going forward. And although we are not giving any numerical numbers at this point. I think in the sense of overall SiP momentum, it remains to be strong.

Okay. Great. I guess there's been a lot of questions on competition midterm. But your initial view pipeline, do you still see expansion? And -- like that target $100 million, I guess, at this stage, how do you see project momentum and also second source risk on existing pipeline into next year?

And I think we will always have -- when the product becomes more volume-oriented, naturally, the customer will like to have second source. And I think that's just the nature of the electronic industry.

So I can give you 2 indicators. First, on the major customer that we started SiP with, we have been growing our SiP overall revenue with that particular customer. And the result is obvious now with all the reporting that we have given. The second indicator is how fast can we enable other customers with meaningful revenue. And we have been struggling with this since day 1.

And today, we're reporting we have 15 projects, 8 customers, other than the major one, with revenue in the range of USD 300 million. We're looking at USD 100 million last year, and we would like to exceed USD 200 million this year. I think we're slightly ahead of the target.

Now going forward, in 2021 and beyond, how can we enable more customers? How do we grow revenue? That remains to be a challenge. But in that front, we continue to explore SiP, new projects, the new ideas with all customers, including the original one.

Okay. That's great color. To clarify, so $100 million last year. So you added $200 million this year to get over $300 million? Just wanted to clarify.

Correct.

Okay. And the second question, I want to ask a little more on the wirebonding because you have such a large industry base. But to go 30% to 40% short, I guess it gets back into the question of how you're assessing the overbooking. Because demand may -- I guess on applications, it doesn't seem to be growing that fast. So to open that up that much of a shortage.

But if you could give a sense how much wirebonders you're adding in? Maybe -- because wirebonding is so broad-based, are there a few major application within that, that's really contributing to the shortage right now?

All right. When people understand the -- this is the nature of the industry. When people understand there's a capacity constraint, everybody will come in and try and secure their share. So if you're asking me of the total demand versus the capacity we have you saw in place, we have a 30% gap, how of much of that is overbooking, I can't really answer that question. But I'm telling you, I'm pretty sure, when the customers are pushing us for the capacity commitment, there's naturally overbooking or exaggeration built in.

Having said that, the kind of phone calls that we're getting right now are not about overbooking and not about forecast upside. In multiple situations, we're talking about line down situations. So we understand the line down, and our job is to make sure there is absolutely no line down situation. Otherwise, the impact is much greater.

So you take that portion out. Then you look at the real demand cycle, based on all of the end market sell-through of product, based on the historical, all of the adjustment, and that will take another portion out. With all of the exercise that we have gone through, we believe the wirebond shortage is real. Whether it's 30%, I do not know. But based on today, the gap is indeed 30%, and that will lead to a more friendly environment when people talk about ASP, and that's real.

But how fast can we compensate the gap, whether it's 30%, 40% or 20% or 15%, and that needs to wait until at least Q2 of next year because the capacity delivery based on tooling and all of the other line balance, that's how long it will take to get to a meaningful level. That meaningful level, in my view, is not 30%, 40% gap. However, it will take a physical amount of time to get to a line balance situation, where we can see the things a little bit more clear. But right now, it's -- clearly, everyone is fighting for capacity.

Okay. And a follow-up to that. I think you guided next year range. Maybe I missed it, but do you have the range like where fourth quarter is, like to get to the full year this year? And if you have the bonders that you expect to add at this stage?

Can you say that again? I'm sorry, I lost you.

Yes. Yes, I was curious, the fourth quarter this year CapEx. Because you mentioned you're pulling in spend, so where you'll end up this year on CapEx? And then how many bonders do you plan to add?

Our overall CapEx last year is close to about $1.6 billion. And I think this year, we'll be roughly $200 million or above it.

Okay. Okay. And then I just want to ask one other question on the balance sheet. I think you talked about meeting target, raising dividends. Is there within that a monetization? I think a couple of calls ago, you talked about monetizing balance sheet. So I'm curious if you have any efforts or that's part of -- or any other type of fundraising you need to do?

We're not -- we don't have any plans for any further fundraising. But in terms of -- it really depends on the cash flow situation. I believe that it will greatly improve next year. But having said that, monetizing some of our assets is still one of the options that we have if we need to go to that route.

Next in line is Rick Hsu from Daiwa Securities.

This is Rick. So my first question is on the wirebonding demand, which you guys said is very strong. Can you elaborate that a little bit more in terms of demand drivers, in terms of what kind of key products and applications that are driving your very strong demand for wirebonding?

We're seeing the wirebond strength from almost all sectors, and I can give you a few examples. For example, on the WiFi, we used to deal with like single-chip WiFi or 2-chips WiFi. And right now, they're becoming in the 3 or 4 chips. And I guess the notion is when there's a new WiFi standard, like WiFi 6, 6A or 6A-plus, the easiest way is just add another die. And that will inevitably require adding more wires, all right?

Now in the industry, when gold becomes very expensive, to use copper wire even though everybody claims they can do copper wire bond. But when you want to do stacked die, fine-pitch, 1,000 wire range, the short wire loop, it becomes a quality issue, how many people can really do this.

The automotive guys in RF analog space, when the WiFi infotainment goes into automotive or go into the autonomous driving, the quality requirement based on traceability also become different. So how many lines that has been automatically qualified using copper wire to do multiple stacked die, that becomes an issue.

When I was referring to a 30%, 40% gap, I was only talking about ASE and I was not referring to the industry, because some of the devices, they are obviously -- based on customer input, ASE is the preferred supplier, mainly because we have the base. We can ramp up more readily. We also have the original IP on the copper wirebonding. And we also have the first of its kind, which is a wirebond line down factory where we can do fully traceable, data-oriented in terms of reporting also the algorithm analysis. So the automotive and the medical customers tend to favor this type of wirebond.

Given the complexity and also the quality requirements and also the sudden increase in 5G upgrade cycle and WiFi upgrade cycles, that's why we're seeing this demand increase. And I'm pretty sure there's double booking somewhere along the line. But the end result is the -- we're having a lot of demand in wirebond, which we have not seen before. And I think it's mainly because of product upgrades as well as automotive going into the autonomous driving range. So I think there's a lot of new demand.

I'll also comment on the COVID-19. I really believe there's a behavior change because of COVID-19. People no longer look at IT product only as a digital efficiency because if you can buy better TV, a better WiFi system, you tend to stay home a little bit more. And when you stay home, you reduce the risk for any COVID-19 or any kind of virus attack.

Now from that regard, people's willingness to buy IT product at a premium, it is different. Also, the age tend to expand. I've heard very, very old people and very young people are buying premium IT products. And those are the behavior that we have not seen before. I think this behavior, because of COVID-19, is going to be here for a while. But even with COVID-19 subsiding, I think this behavior change will last much longer.

All right. Okay. That's very helpful and very comprehensive answer. The second question that goes to your -- I think your revenue guidance, ATM revenue guidance for Q4. If I did the math correctly, I think your Q4 ATM revenue would be down by around mid-single digit quarter-on-quarter, right? So if I look at the revenue guidance from the foundries who already reported, TSMC and UMC combined, they're talking about still quarter-on-quarter growth. So I wonder whether this disconnect between the front end and back end, like yourself, is mainly because the customers try to fill more with the bank than die bank. And if that's the case, if the inventory risk is ahead the industry, you guys will be safer relative to the foundry guys. This is my second question.

I will not comment on that. But what I'm telling you is that in Q3, my EAR-affected revenue 13%. In Q4, 13% is gone. So we are recovering based on 13% minus, work our way back. So there is real growth if you exclude the EAR. Unfortunately, we have to talk about this for the last time, right? But Q1, we will not talk about this anymore, right? Now with that, I think we're comfortable.

Also, I think on the margin side, in Q4, after the EAR issue is behind us, I think that we will see further margin improvement in Q4.

And I think that would be a pleasant surprise. I think the ASP environment partially will be reflected in Q4 and also the write-off will be completed. So we're just optimistic about the resetting in Q4 because, honestly, the first 9 months of this year has been very, very painful for all of us.

Next to have a question, Sebastian Hou, CLSA.

My first question is on the pricing outlook -- sorry, not the pricing outlook. It's actually the pricing -- frankly, pricing environment for wirebonding. I'm curious about how much of this is simply due to the -- on the supply situation. How much of that is to reflect the higher material costs that you have mentioned earlier?

In the discussion, you have to cover both. And I won't be able to give you a percentage, the material cost versus the assembly value-add. I can only tell you they're both -- and by the way, each customer is different. And if you recall, I think it was 2017 or '18 that we call it a recalibration of our portfolio. Some of the SiP was not generating the kind of return, and we sort of dropped that. I think in wirebond, we're actually going through a similar exercise, too.

But anyway, based on the capacity that we have and the customer engagement plan as well as the return profile, this is a great opportunity for us to look at it. What type of business we want to have a longer engagement plan, what type of business we do not want to have a long-term engagement plan. But the overall effect, we will see better in Q4, as of going forward in Q1 and Q2 next year. But this is the exercise that we're going through.

Okay. So by transferring some of the cost -- higher cost to customers, plus some of the undersupply benefit on that pricing negotiations, so net-net, the margin will be aggregate from Q4 onwards. Is that right?

That's correct. Yes. Hopefully, we can give you a better percentage in Q4 guidance in terms of the quantitative. What is that percentage adder?

Okay. Okay. A follow-on that is where is the -- if you remember or can you remind us, when is the last time wirebonding had such a big undersupply gap?

Year 2000.

Okay. The Y2K?

Yes.

Okay. Right.

This time is worse than 2000.

All right. But how do you -- how would you compare the -- I think the different background, the story that is all different. But how would you compare in terms of the customer mix, applications, customers' overbooking behavior is also part of this? And how do you think that -- how long this will last based on your best estimate?

All right. It's interesting that you're asking this question. I'm not sure how much time do you have for me to answer. But in year 2000, you have to remember, that was a very strong IDM captive environment. ASE revenue at the time was less than USD 2 billion. Our installment base was much smaller compared to today's environment.

In 2020, the outsourcing percentage has greatly increased in the last 20 years. ASE's ATM position improved from $2 billion to $9 billion-plus. Our installment base right now, our wirebond in the outsourcing market, I believe, is way over 50% on a global basis and also on percentage. So when I talk about the supply/demand balance in year 2000 versus today, it's quite a different scenario. So I think year 2000, if ASE is short by 30%, that means something. Today, if I'm short by 30%, that means someone had done something else.

I'm not exactly sure how is the other OSAT capacity constraint. I don't believe they're short as much I did. Chances are the -- I always get fully loaded first. Over the last few years, even when business is going up and down, my loading situation has always been quite full. In this kind of undercapacity situation, the other guys will get filled. But long-term competitiveness, you got to go back to product complexity, quality and also the liability you can accommodate with respect to all of your end customers.

So I do believe all of these business terms will come into play. People want to look at somebody who is highly reliable, who's got R&D pipeline, who has also got the investment appetite should they need it to ramp up new products.

So for the wirebond, I do believe there is an upgrade cycle because of 5G, because of WiFi, because of the electric car. And now we're seeing a lot of the analog devices. As a matter of fact, if you look at the 8-inch wafer demand, it's going through the roof and the fact that people are talking about ASP adjustment on 8-inch, and you will understand that. I mean, there's something fundamentally different in this realm.

Got it. Very insightful sharing on that. But on the supply side, do you see any constraints on the capacity? I think the 8-inch wafer, they have their supply concern given the secondhand equipment availability. But how about the wirebonding? Do we have the similar constraint or that's not a significant issue?

Well, the wirebonder is only one component of the wirebond line. You also have other instruments. That's why when you talk about line balance, that's why I said that it will at least take 6 months to get industry capacity to a level where people become more comfortable and talk more rationally. And I will not go into detail what the wirebond line consists of. And a lot of vendors will be involved.

But the industry throughput in terms of ramping up this kind of equipment are very slow, so that becomes the bottleneck. So in a way, it is actually helping all of us to regulate the water level. We're not adding capacity in such an accelerated pace. I think, intrinsically, the industry has a built-in buffer, a regulator in the whole supply chain sanity, and I think ASE is playing a portion. Unfortunately, we're at the bottleneck. The wirebond capacity is also part of the bottleneck. But in a way, I think this is really helping everybody. With substrate, it's not a bottleneck. So there are actually multiple bottlenecks right now for the back end. So we'll see how this is heading to.

Got it. Last question for me is just a very small question on CapEx side. I remember in the beginning of the year, I think the capacity expansion and CapEx focus was mainly geared toward the testing side. So going to next year, given that our testing utilization rate is affected more by the EAR-affected customers, so can we -- is it fair to assume that in next year, the mix will get toward assembly?

I think it's too early to say because we are working very hard to convert those testers for the other customers right now. Just give it another quarter.

Next one to ask questions Gokul Hariharan, JPMorgan.

Yes. Just to follow on, on Seb's question on the testing side. I think previously, you had indicated that your long-term goal is to increase the turnkey test ratio from 1/6 -- or 15 to 1/6 right now to 1/3. Obviously, the EAR-affected customer was one of the biggest turnkey customer. So could you talk a little bit more in terms of what are the efforts ongoing to kind of increase the -- or expand the list of customers who are using turnkey test at ASE? And could you also talk a little bit about what are you -- what is the initial feedback from those customers? And any change in that plan in terms of getting to that 1/3 turnkey test ratio?

And Dr. Wu, if you could also say what are the time line by which you get to, let's say, 1/3 of test revenues for your total ATM business?

The testing remains to be a major initiative at a group level. So we understand that there is EAR setback, and we accept that. Very painful, but we have very little choice. However, that doesn't alter the strategy that we want to increase our turnkey percentage as well as our test percentage as part of our overall revenue portfolio.

So because of the assembly capacity is short right now, it happens to offer us a better leverage. So when we talk to our assembly customer, we naturally would like to propose we'll also perform testing for them.

The EAR is giving us a short-term setback, and we have to go through different test platform. Or even in the existing test platform, we have to do a lot of operation. Then we need to have some time, but it remains to be our major initiative to increase our test revenue and the turnkey ratio.

Any time line on when you would reach that target? Is it like a 5-year target, a 3-year target?

That was never really a target. That was just kind of a normal turnkey test occupies 1/3 and packaging occupies 2/3. That's just a theoretical concept.

You're giving me a good hint. Well, I'm going to go back to talk to my team. We need to have a target, yes.

Next one, we have Bruce from Goldman Sachs.

I want to switch gear here for the question for the advanced packaging. So if you look at your advanced packaging, the revenue growth is somehow similar to the corporate average. On the other hand, TSMC is reporting that their packaging business is growing by 30-plus percent this year. So obviously, they are gaining market share. So what do you think about your future growth outlook for your advanced packaging? What kind of growth rate we can expect that?

Okay. First of all, I'm really happy that TSMC is growing and I'm happy TSMC is growing their packaging part of business, right?

I think the first and the most important clarification I would like to make is now we have to understand what wholesale market really is. In order to do OSAT service, we have to qualify for 3 situations. My apologies for this elongated answer because I really need to speak this out. Many people ask me the same question. I just wanted to clarify this with all of you.

If this is a captive market, it is not OSAT. The central memory is captive. That is not part of our serviceable available market. Intel microprocessor is a captive market. Therefore, we never attack the Intel microprocessor business, per se. TSMC packaging part of the business focused on leading-edge lithography, completely captive. That is not part of the OSAT market. Yes, it is a packaging assembly revenue, but that will be in the same category as Samsung memory as was Intel microprocessor.

Now the overall packaging revenue, which I will give you a more clarification Q1 update next year, continue to grow. As semiconductor grows, as you already said it, the packaging and test value will continue to grow. However, the petition and the division between the IDM captive vertical versus OSAT fragmented, outsourced service player based on merit compete based on value efficiency. That has never changed.

Having said this, now another way that you would do is the packaging revenue and test revenue, we're talking about the value-added service. You really have to look at the overall revenue content. How much of that is silicon-based? How much of that is memory? How much is everything else? Then you're trying to dissect that and understand what is the particular assembly value-add versus test value-add versus material contribution versus silicon contribution versus interposer versus all of the detail. Then you understand that some of the business, not only we cannot do, it is just out of our business model. This is not part of the OSAT engagement that we will do.

Can I be clear about this, that when we used to talk about the OSAT market, sometimes the advanced technology, OSAT will do the second wave business. So when the technology is mature, OSAT can do the business. But Ken was mentioning that at the certain like node, they will be the captive market by the foundry. Can you tell us where is the border line for that? Or which node -- from which node we can expect that will be the capital market for foundry?

We had the same conversation about Intel microprocessor. I mean, as long as I remember, I had this conversation for 30 years. We also had a similar conversation about the central memory. When do you think central memory will go to OSAT? And when do you think Intel microprocessor will go to OSAT? But finally, we're talking about Intel might outsource their advanced wafer either in graphics channel or whatever to the foundry. And that will become noncaptive.

I do not know the line, the division. I don't think we have a clear view about in which node, which lithography that will become the OSAT market. Even when that becomes the OSAT market, you still need to look at the value-added content versus the material content. But you still have to look at it, out of the value-added content and the material content, whether it's foundry to do it cheaper or OSAT to do it cheaper.

So once you understand there's no fundamental thesis that we can decide which business is suitable for OSAT, which business is suitable for IDM, which business is more suitable for foundry, for captive market, I mean it's a very interesting conversation. We can talk for hours about all of the details. But I really get confused because everybody asks me the same question and I couldn't understand where the question is coming from.

I think the question-and-answer are very complicated. That's why people are asking every time. So I mean maybe next time or -- we can come up with somehow the quantitative addressable market at least for the coming like 2 years, which is easier for the investor to understand.

Okay. So I'll be happy -- yes, we should prepare something. Thank you.

Okay. I think we're at the end of the conference call. Let me sum up what we have on this call. I think, first of all, we had a good third quarter. And we are seeing improved situation getting into Q4 now that EAR is now largely behind us. Business continues to be strong, and we are in a pricing-friendly environment.

Over the past 3 years, we have been making a pretty heavy CapEx investment. That puts us in a better position to capture whatever growth opportunity going forward. We're seeing synergy being created, as reflected in our improving margin and lowering OpEx ratio. We remain optimistic business prospect for 2021. And we will see improving cash flow and our financial standing going forward.

Thank you very much, and that concludes the conference call today. Thank you.