Alchip Technologies, Limited (3661) Earnings Call Transcript & Summary

Okay. Hi. Analysts, fund managers and media friends, welcome to Alchip Technologies second quarter investor conference meeting. For this meeting, Johnny and I will cover our second quarter operating results and the future outlook. And yes, more than glad to have you here. Okay. So [ routine ], it's the safe harbor disclaimer. Okay. This meeting will be in English. Johnny and I are all Chinese native speaker, so if you want the Chinese presentation slides or you can ask questions in Chinese. So if you need Chinese presentation slides, please go to the [Foreign Language] You can download the Chinese version of the slides there. And if you don't want to ask a question early, you can write down your questions through the new message function. We can read it and we will answer after and read in English. [Operator Instructions] The video and all the content of this meeting will upload to [Foreign Language] about 2 to 3 hours after the end of this meeting. So I'll invite Johnny to start the meeting with very short company introduction in case maybe some of you may not know Alchip very well.

Good afternoon, ladies and gentlemen. I'm Johnny Shen, President and CEO of Alchip Technologies. Yes. Thank you for joining our investor conference meeting, and we appreciate the opportunity to share our Q2 results and provide guidance to our future business outlook. For those of you who are not familiar with Alchip, allow me to do a brief company update. Our company was founded in 2003, IPO in 2014. Current market cap is around USD 4.5 billion. Head count, 600 employees. Majority of our employee are engineers. Last year, our revenue number is $460 million. Obviously, this year, we grow quite a bit. The first 7 months, I think we already overachieved these numbers. Our market focus is HPC and AI. And recently, we had the automotive category as a new focus as well. Since we found the company, we successfully taped out more than 500 designs. And among 500 design, 60 -- more than 60 is FinFET. And within 60, more than 15 are [indiscernible] CoWoS related design. Our capacity is at 20 to 30 a year. Right now, majority of revenue is 80% of revenue in HPC AI area. We are TSMC VCA member since 2009. We recently just joined the TSMC 3DFabric member last year.

Okay. Please don't use your sketching function in Zoom. And the pen view, you just draw lines on the slide. Everybody can see the lines. So can you erase the line you just draw on the slides? Thank you.

All right. Let me give everybody a quick cap on the Q2. Q2 marks another record-breaking quarter with revenue of $258 million, operating income $28.4 million, net income of $23.9 million, resulting in an EPS of TWD 10.16. While we are -- our Q2 is at historically high, but our income and EPS has not met our expectations. The primary reason is lower NRE mass production revenue ratio, which caused a notable reduction on gross margin. Additionally, we are facing account receivable issue with one of our accounts in Japan. We follow our auditor's guideline. We accrued a certain bad debt allowance as operating expense in Q2. Fortunately, fundraising progress for this customer has improved. We have already begun to receive part of the payment. As a result, our operating expense were returned to the normal level starting from this quarter. As for the future business outlook, mass production demand from HPC and AI remained robust. In addition to improved demand increase from our largest customer, several other customer has significantly increased their demand as well. Overall, we anticipate higher revenue forecast than our previous projection this year. The actual amount of growth is tied to a specific material year-end capacity constraint. Another normal piece of information to share, we have begun to receive purchase order for next year. The production demand from U.S. customers for next year even surpassed this year's figures. That suggests the momentum of revenue growth is likely to continue through 2025. Update for the automotive business. Similar to HPC and AI sector, our progress in automotive business is also promising. We have secured few design wins, have a fruitful pipeline of design from many carmakers in China and also in U.S. Our main focus is the ADAS L2 to L4 type of ASIC application, utilizing either N5A or N3AE technologies. Notably, we have achieved a significant design win from a Tier 1 carmaker in China last quarter. As I mentioned before, even automotive project has a tighter constraint with longer design turnaround time compared to other applications. However, auto project can generate substantial NRE during the design phase and produce a considerable amount of [ NT ] revenue during the production phase. We firmly believe the automotive business will be another major revenue driver for us starting from 2025. A quick update for geopolitical risk management. After successfully effectively diversifying our business concentration beyond China to other regions, in Q2 our revenue from Asia Pacific region is less than 20% now. For China's business, we are adopting cautiously approach by closely cooperating with major suppliers that involve reviewing customers' background, their end user, design spec before even accepting the projects. We continuously believe in and support the business from China as long as they meet compliance rules and regulations. In terms of our workforce, we have implemented a very aggressive hiring plan to bolster our engineering support in Japan, Taiwan and Southeast Asia. Our newly opened Malaysia office is already staffed with a fuel engineer. The hiring strategy efforts and -- to provide comprehensive and cost-effective solution to meet our customers' requirements. As a conclusion, we are confident to say our business is at best healthy state. We anticipate 2023 will be another record-breaking and outstanding year for Alchip. Thank you.

Okay. Before we go into the next slide, let me repeat again. Please don't use your sketching function to draw lines on the screen. Okay. For this page, this is our second quarter quarterly income statement schedule. In the second quarter, the total revenue is $258.6 million, which is 37.5% quarter-on-quarter growth and 155% year-on-year growth. And for the operating income side, for the second quarter, our operating income is USD 28.4 million. It is a 35.7% quarter-on-quarter growth and a 47% year-on-year growth. And as a result, the net income for the second quarter, we already announced a couple of days ago, is USD 23.9 million, translating into EPS of TWD 10.16. So for the breakdown in the second quarter. Again, our HPC remains our majority application revenue. More than 80% of our total revenue exposed to the HPC area; followed by the consumer, 10%; and the niche market, 5%; and networking, 1%. For the first half this year as a whole, the number is identical: 81% HPC and 10% consumer, 8% niche and 1% networking. For the revenue by process node, I would say that Alchip remains the industry leader in terms of the revenue exposure to the most leading-edge technology node. For the 7-nanometer advanced or more advanced process node, our self-exposure to them is 87% for the second quarter this year, followed by 28% -- 28-nanometer, which is 6% revenue exposure and 4% to 16- or 12-nanometer process node. And the [ rest ] is the 40-nanometer or more legacy process nodes. So for the first half this year, 81% chose 7-nanometer or more advanced process node and the rest accounts for close to 20% of our total revenue. And for the region, as Johnny mentioned -- as our CEO mentioned, the exposure to China is trending down. Of course, there are many reasons behind, but obviously, North America will be our major -- will be the region for our major revenue exposure. For the second quarter, revenue to North American customers accounted for 62% of our total revenue, while the revenue exposure to Asia Pacific accounted for only 17% of our total revenue. And for the first half this year, 63% revenue exposed to the North American region while 20% to Asia Pacific, 8% to Japan and the 9% to other regions like Middle East and Europe. I don't understand why there is still people drawing lines on the slide. So please don't. For the second quarter '23, the revenue came higher than we expected. There were 2 reasons behind this higher-than-expected second quarter revenue. First of all, some of you may remember that in the first quarter earnings call meeting, we said there was testing program review for our major products in the first quarter. So some delayed parts for these products to our North American customers delayed shipment from first quarter to second quarter. Secondly, our biggest customers keep on pushing us to speed up the shipment today. And we try very hard to ship as many as possible parts to this customer in the second quarter. So that's the major 2 reasons for the higher-than-expected revenue in the second quarter. And by comparing to the first quarter this year, second quarter has slightly higher NRE exposure. In the first quarter, our NRE revenue accounts -- accounted for about 17% to 18% of our total revenue. For the second quarter, the NRE revenue accounts for around 20%, 21% of our total revenue. And you can see the profit. I knew that many of you thought that the second quarter profit performance is slightly lower than you expected. Here are the reasons. The second quarter gross margin is 21.4%, given high shipment of the AI chip to the North American customer. And of course, the [ higher ] percentage of production revenue did cause pressure to our blended gross margin performance. That's the main reason for the gross margin front. And for the operating profit. The operating profit, which is obviously impacted by onetime bad debt allowance accrual for Japanese -- for Japan customer. This debt allowance could be reversed in the second half once customer pays it. And the good news is, first of all, this customer has very good track record in the past. This is not a new customer to us. It is a long-term -- this customer has been a long-term customer for our Japan subsidiary. And secondly, this customer is paying the receivable already, part of it. And we feel pretty confident that we can gradually recoup the payment in the second half of this year. And the nonoperating income post declined quarter-on-quarter numbers given the weakness of RMB. Let me explain here again that, to us, our functional currency is U.S. dollars. So every other currencies like renminbi, Japanese yen and the NT dollars depreciate to USD, we will report some foreign exchange loss. But it is essentially a good thing to us because we use those currencies as the expense on the salary to our employees. So essentially, when the other currencies to USD depreciate, we get a real benefit although we may record some FX exchange loss. For the 2023 business outlook, Johnny already mentioned some of them and I will do it very quickly. I believe most of you will ask about other questions in the following session. First of all, the AI chip shipment to North American customers get stronger. First of all, the customer keeps on requesting us to speed up the shipment to them. And secondly, for the third quarter, I already hint to some investors that we may have mild revenue for the decline in the third quarter because of the constraint by cost and we believe the issues will resolve in fourth quarter. So the decline could be very slight quarter-on-quarter in third quarter. And the shipment outlook for the overall second half and 2024 as a whole, the shipment we expect to get stronger and stronger. And previously, we guided for the major products currently shipping to our North American customers, we may have 30% quarter-on-quarter -- year-on-year shipment decline in 2024. But for now, the most updated forecast from this customer is that we may have year-on-year growth for this AI chip shipment to them. And for the CoWos capacity limitation, currently we are talking with TSMC and we believe the outlook is good currently for the major products to our major customers that we already got the full support from TSMC for next year. And in addition to the biggest customer this year, we believe our other customers will post heating up momentum starting from the second quarter. There will be another AI chip customer who will most likely to post strong revenue growth to us in second half or in 2024. And there are other 4 -- there will be 4 other high-volume projects to enter the production phase sequentially starting in fourth quarter this year, which includes consumer products, AI chips and main chips for mobile devices. And for the NRE [ parts ], we remain our positive outlook towards the NRE growth in the second half and in 2024. The reasons are, first of all, the process [ node ] projects keep on moving to the 5-nanometer or even 3-nanometer. And in the second half and in the year 2024, we expect the automotive, especially the ADAS project, to keep on flowing in as we may win projects other than the one we just won in the second quarter from a major China automaker. And I guess that's the -- I already come to the whole outlook part for this conference meeting, and we will go into the Q&A session. [Operator Instructions] Thank you.

And for the first question, [ Has ], please?

Okay. Congratulations on the good results. My first question would be on your seasonality for 2023. Following a strong growth in first half, you mentioned sales will decline mildly quarter-on-quarter in 2Q due to CoWos constraint. Could you try to quantify the magnitude of the decline? And do you still expect fourth quarter to grow sequentially? And maybe you could also provide your expectation on NRE versus turnkey mix for second half and your full year sales guidance?

Okay. For the third quarter, we expect revenue decline in single digit. And for the second half, I will still say the total revenue for the second half will be higher than the first half. And the question...

For NRE ratio.

On the NRE ratio in the first half, I already mentioned in the presentation. For the second half, as you may know that on the milestone management is quite important the execution and scheduling. I cannot give you a very firm number. I can only say the ratio of our NRE revenue. The percentage of the NRE revenue to our total revenue will be higher than the ratio in the first half for the second half. Usually, the fourth quarter is the high season for our NRE revenue. Go ahead, [ Has ].

Okay. Sorry. What about the full year sales guidance update?

Full year sales update. The Taiwan Stock Exchange posted and I don't know -- they're pressuring us not to guide the revenue numbers. They say it's obviously -- previously we guide we're trying to achieve more than USD 800 million in our previous earnings call. And right now, we can say that USD 800 million target is unrealistic. We definitely will go over this target. The most possible revenue goal for us will be -- I can just give you the range, $900 million to $950 million or even better. [ Has ], am I answering your question?

Yes, yes. That's great. And my second question is regarding your project pipeline for your U.S. customers. Would you be able to share the contribution from your largest U.S. customer in the first half of this year and your full year expectation from that customer? And if possible, across your U.S. customers, in 2024, what is the growth outlook? Would you be able to provide some positive and negative factors impacting your U.S. business next year?

I will say this way, I will say this way. For the revenue contribution by our biggest customers this year, we expect this customer to account for around 40% to 50% of our total revenue. And for next year, because I haven't done the numeric forecast yet, so I can't give you a number, but for the revenue they can contribute to us, previously we guide we may have year-on-year decline in 2024. But right now, I will say the revenue contribution from this customer will post year-on-year growth in 2024. And for the other projects, I will say, for next year, 2 AI projects -- AI chips project to another North American customer will contribute us pretty good. I cannot disclose the numbers, but comparing with this year, the growth will be very, very significant. And you may know that there are other projects like the image signal processor chip to Japan-based customer, which is considered high-volume project in next year. Actually, the shipment will likely to start in the very early of the first quarter next year. And we will also have 2 HPC projects to China customers. And one main chip project, the application process -- processor chip to -- for mobile devices to another North American customer and the shipment will likely to start in the middle of next year. In the same time, the chip to -- for the same customer for the smart device will also start its production in the middle of next year. That's the overall picture for the second half and next year.

Okay. A quick follow-up. What about Habana Gaudi projects, which Intel mentioned a strong pipeline? And back to 3Q, how much of your business was impacted by the CoWos constraint?

[ Has ], it's better not for -- clearly, it's better for us not talking about the names of our customers and the project names of our customers in the earnings call.

Okay. So maybe I'll just ask a different way. What about your U.S. semiconductor companies start-up projects? How much of the volume expectation or the opportunity there do you expect in second half of this year or even in 2024? And I also have a separate follow-up on the third quarter. How much of your business was impacted by the CoWos constraint?

For sure, the revenue contribution is not comparable with the contribution by the current customer -- for the customer you mentioned and -- but it's quite significant. I cannot give you the numbers. For the CoWos, I think Johnny can talk about the CoWos situation.

Okay. Yes, to be honestly, our revenue, the total revenue, especially the mass production revenue, is dictated by the CoWos capacity. I think for the past months, we've been working with the supplier very closely. Fortunately, we get -- we can say we start to get a very reasonable support. Yes, the reason -- okay. The reason behind that, as we all know, in terms of CoWos, we are -- right now, we are one of the biggest customers for the supplier right behind maybe top 3 customers. If you look at us, we are representing more than 5 customers, yes. So suppliers also want to diversify their CoWos concentration. We've been quite successfully [ grant ] a certain capacity to support our current biggest customer and also the customer you mentioned above. And next year -- right now, we have an R&D project to work with them. If this CoWos solution proven, we will start to receive very fairly amount of allocation, not only for the #1 customer but also for other customers. So Q3, we still suffer a bit. But starting from Q4 and next year, we expect everything will be very smooth.

Okay. Got it. And a quick follow-up to my second question would be on your largest U.S. hyperscaler customer. Beyond the existing 7-nanometer AI project and a few other consumer projects you mentioned, could you provide some update on the new AI projects with the largest U.S. customer?

No. [ Has ], you know that we cannot provide the details.

Okay. No problem. So probably -- okay, Johnny, do you have anything to add or...

The specific part and the customer name we cannot mention, but obviously, the AI inquiry is getting higher and higher. Many system companies start to do the ASIC chip, for sure. And surprisingly, many start-up companies they are used to face a lot of problem recently. The fundraising situation for them become much, much better, so they come back to us as well. So I think the AI opportunity, I can say, for the past quarter is incredibly high.

Okay. Let me give the others opportunity to ask questions. Charlie, please.

Congratulations for great results and business outlook. So probably just use some kind of public comments information to ask a few follow-ups. So I believe in your AGM, right, when you answered to media's question whether you're confident to secure next-generation project from the big customer, you said the confidence level is high, right? So I'm just checking that whether the kind of confidence level is changing.

Okay. Yes, I mean the confidence level is better. Better, yes.

Okay. So better than...

Better than before. I think this particular customer continuously give us more constraints and also the more [ game rule ] is we try to -- we already follow one by one. So right now, we have a hundred engineer ready to do outside of China for their request.

I see. Yes. And I don't want to be too specific, right, but when do you expect you can see the first N3 NRE revenue to contribute to Alchip?

Currently, I believe probably we have chance in the second half this year for N3 revenue.

Okay. Okay. Sorry, I wouldn't go into further detail, but that should be very, very sufficient information. And over the past 10 minutes, I already got that 2 questions from investors. They seem to be a little bit confused about your [ yearly ] revenue updates because you kind of mentioned the previous USD 800 million, right? But I believe what you just said is that it's much better, could be even above USD 950 million. So let me test your limits. Do you think you can get to like USD 1 billion in 2023?

Charlie, the reality we are -- I'm facing is if you say a number, the second day the media will quote the number. They don't care about the context. So -- and clearly, we'll read the numbers for us. So we mentioned that the second half revenue should be higher than the first half. You can calculate that for the first half, we have $188 million for the first quarter and we have $258 million -- yes, $258 million in the second quarter. And if you double the numbers, I think we are pretty confident and comfortable for delivering higher numbers than 2x of the first half.

Okay, okay, okay. Yes, I think maybe some investors just need that clarification. So no doubt that this year is a very strong year. But how about next year? I think you already provided a lot of context, right? Big customer going down next year; other U.S. customer, mass production, other for chip, right? But if we want to put those in kind of a numeric number, right, I think you should have some sense, right, about the next 2 years revenue CAGR. Can you provide that number?

Okay. For next year, I would say this way, I think every company is seeking for growth. So do us. So I would say, this year's revenue is like a baseline for us. And for this year, previously I worried that -- the thing worries me the most is how can we make up the gap generated by the year-on-year decline of the major products we ship to our major customer. And right now, fortunately, last year, the contribution for this product will not go down. We may even have a certain amount of increase for this product next year. And besides of that, as I mentioned, there are about 4 to 5 high-volume projects into production next year. So for the production part, next year, definitely will be a very promising year for us. And for NRE, of course, during, I would say, the -- during this [ AC ] market status and we have a very good position to win projects, there is no reason for us to have year-on-year decline for NRE numbers, especially when the process node is still moving. So all in all, I cannot give you numbers for the year-on-year growth right now. But I would say, first of all, definitely, the growth rate, the percentage would not be that high compared to 2022 to 2023. But obviously, we can deliver pretty strong numbers next year.

Okay. Yes, Charlie, let me try to add more. Just like Daniel mentioned, please try not to expect us at the same growing rate compared to this year, unless another project here [ homegrown ]. Otherwise I think, the growth rate will not be as high as this year. But in terms of mass production, our #1 customer give us a higher forecast. I think that's very good news for the company. In addition to that, we already started we see many orders from other customers. I think that's -- that, for sure, will boost the MP revenue. The NRE, I think that area is under our control. The MP, honestly, is a little bit out of our control. It depends on customer demand. It also depends on year-on-year constraints and also capacity from supplier. NRE, I think that area, we have confidence to continue to grow because more project pipeline and also the technology migration. For the past 10 years, our NRE keeps growing. So I don't worry about next year. I think, for sure, NRE will grow to a certain degree.

Got it. And last one, I will move back to the queue. Gross margin and operating margin trend, because I know this quarter kind of hit the kind of low for gross margin. But do you think the gross margin can improve next year? And also operating margin, I think this is most important metric for your business model. What would be the long-term goal for your operating margin?

Okay. So I have been keeping -- I've been keeping on explain the business model or the structure of our book to the investors. I will say, for example, for this quarter, you can see that our gross margin is pretty similar with the number in first quarter for the gross margin front, but we have much better operating margin in second quarter, even impacted by the AR accrual. So it has been a very [ debatative ] topic for us, and therefore, the discussion with you that because the production revenue is, to us, is like an effortless revenue and the gross profit for the production will trickle down directly to the operating level, which means as long as we have the production, the profit will go directly to operating and we will have higher EPS. To me, personally, I want to make more money. So to me, I don't really care that much about the gross margin. I care about how much money we can make. But I can understand that the capital market and from an investor point of view, gross margin is really important. So do us, and so we try to balance the gross margin and the EPS. The most effective way for us to boost the gross margin is to have more NRE revenue. We try, and we think, starting in the second half, the gross margin will improve. And starting from next year, I would say, for other high-volume projects, the gross margin of that will be better, will be higher than the current product we are shipping. So for the growth -- for the direction of the gross margin, I am pretty optimistic on the future trend.

Okay, okay. Yes. So what's the target for long-term operating margin. Is that -- any [ clue ] for this?

I don't have a target. We can record $2 billion, $3 billion, $4 billion production margin, I'm happy to have it. Yes. And at that time, I think our operating margin will be very high.

Okay. So just to make sure, what was the pro forma operating margin in 2Q? And -- or can you tell me what was the write-off number in 2Q, so we can calculate?

Let me give you -- let me give a range because we cannot say it in the earnings call. It could cause trouble from the authority. The range could be around USD 5 million to USD 8 million.

USD? USD? U.S. dollar?

U.S. dollar. Okay. And then next one, Szeho. Szeho, please.

Yes. I think the first question, yes, based on your interaction with your customer, the major customers, if there's a choice between, let's say, the third-party of the share of IP and proprietary IPs, would they prefer...

Szeho, sorry for interrupt. Your voice is not -- I can barely hear your voice. Can you be louder?

Is it better now?

Yes, yes, yes, much better.

Yes, yes. Sorry. Yes. Actually, first question regarding the IPs. When you deal with your customers if there's a choice between, let's say, the third-party of the share of IPs and proprietary ones, would they prefer one over another if there is an option for them to choose?

Okay. Szeho, I believe it really depends on the customer strategy. Different customers have different strategies. But for the IP part, I would say Johnny can give you some color about the IP part. We are going -- our thinking about the IP part.

Yes. Talk about the HPC and AI. To be honest, the IP is very fixed. For the interface part, it's a very straightforward IP. People use [ HBM ], SerDes, PCI and et cetera. I think, for us, our IP strategy, I think, remain the same: we will not make any connectivity or mixed signal complicated IP to compete with our partner. We are working very high with IP partner to provide the ultimate solution to our customers. Yes, I think that's the right way to do because if we start to do the IP even a portion, it's no way for a company like us to provide a total IP solution for our customer. But if you touch the IP, you will end up to compete with major IP partners. Right now, we have a very, very good relationship to our #1 IP maker in the industry with current design and also with the future road map. Nowadays, I mentioned many times, SoC is very complicated. Everybody should focus on their SoC area. I think there are a dedicated IP house doing IP for them. There is a dedicated foundry. And we are the one doing the back-end implementation. We have confidence that nobody doing better IP than us because we have such approval track record, more than 30 tapeouts. In terms of CoWos, yes, we are pretty much -- in terms of experience, we are pretty much top 3 in the industry. So IP strategy remain the same. Unless later on, when we grow to a certain degree, we may consider about using -- that we evolve for any specific deal. But right now, we are quite confident to work with our partner-provider on comprehensive solutions.

I see, I see. Very clear. And my second question, right, the company is definitely benefiting from the AI mega trend, right? So I just wonder if the company starts to deploy AI in our workflow to improve design productivity?

Szeho, can you please slowing down because your voice is... not...

Sorry. Yes, maybe I repeat my question, yes. I think AI, is the company starting to deploy AI in our design workflow to improve their productivity?

You are saying if we are deploying the AI into our design flow?

Right, right, right. Do you use some AI technology, yes?

Okay. Yes, let me try to answer that. I think right now using the artificial intelligence to reduce the workforce, to reduce the turnaround time, I think that's a hot topic. Yes, for us, for the mainstream technology, yes, if we're doing -- I'm just using design -- current design as an example. For the mainstream technology, 28-nanometer, a few years ago, like 5 years ago, when 28 become the most leading-edge technology, we need about 40 people to do 1 piece of [ that ]. Right now, if we want to do a 28-nanometer design, we only need 8 people. Because of a stronger EDA tool because of our AI technology, they have a lot of reference, they have a lot of data, statistical data, to improve the design turnaround time. But in the other world, if you tackle the most leading-edge technology like N5, N3, unfortunately, the database is so few. There are no database. You have to do the design, everything from scratch. AI engine cannot do any inference. They don't have a statistic to try to solve, try to inference. So when we're doing the 5-nanometer design, yes, we need about 60 people. When we do the 3-nanometer design, we even need 100 people. So using AI, they can improve the machine turnaround time. But in terms of design, because they don't have current statistical database, so there's no way to do the significant improve. When the 3-nanometer become mainstream, I believe, IBM, they only need like 10 to 20 people. But right now, we need a lot of workforce in order to do the most leading-edge technology. AI can help you to solving the existing program, but AI cannot being created to solve the future problem. I don't know, did I answer your question?

Yes, yes. It's actually a hypothetical, yes. Hypothetical, yes, but it's very helpful, yes.

Thank you. Okay. The next one is Robert. Robert, please.

Can you hear me?

Yes.

Yes. Loud and clear.

So I've got a couple of questions. The first one is on automotive. So it's quite encouraging to hear that you have won automotive project from Chinese customers. So I'm just wondering, how big it is for the engineering resources that you have to dedicate compared to the normal HPC project? And also how should we think about the size of the turnkey revenue? And how fast will it start to contribute to your [ funding ] at the earliest?

Okay. Robert, for the automotive projects, I would say the supporting effort not only from back end but also for the front end. So we will partner up with a front-end partner to do the NRE, for sure. And the supporting effort is obviously, higher-than-normal HPC project. So the total contract value for automotive, especially in the ADAS project, will be higher-than-ordinary HPC type of projects, for sure. And for the yearly contribution from mass production, okay, my position is not to give very, very optimistic view to investors because it has not happened yet. But we do receive forecast from customers. The forecast is quite significant. Too good to be true, let's say this way. So I would rather not to talk that into numbers for the contribution. But I can guide that the yearly contribution is quite significant.

Okay. Let me add to that. Automotive design, I think we need additional resource not only for the design phase, front end, back end, for the safety manager, security party, they all need additional resource. And also many new tools need to be implemented to do the design. So design turnaround time longer and design effort is higher. But honestly and personally, I think about it's really worth it. Just like Daniel mentioned before, the amount of forecast they give us is incredibly high, yes. I really hope we all have multiple #1 customer in the near future. In order to diversify the concentration for the automotive, I think right now we have a chance to win multiple accounts. I think 2022 -- 2025, as I mentioned before, I think automotive becomes another strong revenue driver for us.

Robert, we have all these questions for others that [ key tap ] principle.

Okay. So just a question of CoWos capacity. If you are looking at 2024, how is the CoWos capacity allocated to you compared to second or third quarter this year?

I cannot give you the numbers and I can only say for next year the CoWos capacity is tight. It is the reality. But as Johnny mentioned, our position within TSMC is the best than ever. So we have very good support from TSMC for the CoWos capacity allocation.

Okay. Got it. Second question. Socionext. Socionext also compete with you on the same space on the auto side. So is there any -- if you look at the 2 when you're competing with that, is there any project they have, like they have stronger opportunity in winning? And what kind of projects do you have a big opportunity of winning the type of project in the auto compared to Socionext?

Okay. In the earnings call, we had better not mention the names of the customer and the projects of the customer in the pipeline. It will cause big trouble to not only us, but also our customer. Yes.

Yes. Understood. Socionext is -- okay, in terms of competing with project with Socionext, do you see them as a competitor or they are not a competitor because you work on different projects?

Yes. Socionext is a very respectable company. And its strength is in automotive application, for sure. But I would say for the [ physical side parts ], I think we are the leaders in this industry. So we don't rule out any possibility to work together with Socionext.

And the next one is Jeffrey. Jeffrey, please.

One quick question for me, a follow-up on automobile. Would I be right to assume that if you do an advanced ADAS project and get through NRE and mass production, that, that would be like a several year -- way longer than your normal projects for mass production for auto?

Yes, automotive project, especially for the ADAS because there are several type of -- different products for the automotive-related project. So for the ADAS, usually the design turnaround time will be much longer than a typical HPC project.

But would also the mass production length also be several years, will be a much longer than normal mass production project?

Yes, for sure. We expect the automotive type of project, the product management cycle will be like 4 to 5 years. And the next one is Susanna. Susanna, please. Susanna, we cannot hear you. [ Susanna Shu ]. You may have some connection problems, so let's jump to...

Joey.

Can you hear me?

Sorry. [ Susanna ], okay, please.

Sorry. I mute myself. I would like to ask, TSMC 3Q CoWos capacity should be slightly up Q-over-Q. So may I ask why do you expect your revenue to decline Q-on-Q due to the capacity constraints?

You are asking about why we can't get support from CoWos from TSMC?

TSMC is the...

Yes. For 3Q.

For 3Q, we don't have a problem. The CoWos actually for this year is pretty much set in the CoWos allocation. Of course, we have some flexibility, but the range is not wide.

Yes. I think he said -- her question is TSMC continuously increase the capacity, but why we are facing Q3 declines on -- am I...

[ Susanna ], because your voice is not very clear, so we are guessing your question. And are you asking why TSMC keeps on expanding the CoWos capacity, but in the same time the CoWos capacity is not enough for allocating to customers? Are you asking this?

My question, can you hear me?

Yes, please.

Once again, my question is, yes, TSMC keeps expanding its CoWos capacity, especially in 3Q and 4Q as well. But you expect the -- your revenue decline Q-o-Q because of the CoWos constraint on the [ TCT ]. So I would like to ask what is the reason behind.

Yes. It's a complicated question because there are different type of CoWos. Some are new. Some are already running for a while. So multiple reasons because for the CoWos, the technical details, I believe, we cannot discuss in the earnings call because it's related to TSMC.

We can only say TSMC is working very aggressively trying to solving the CoWos constraint issue, capacity, EO, everything. So for Q3, we are facing certain challenges for CoWos bring up, I think, working with the TSMC and related supplier very closely trying to solve. Unfortunately, the July and August is overstocked. It's not too much room for us to recover, but starting from September in Q4, I think mass production revenue will go back to the normal projected stage.

Maybe can I have a follow-up question? Actually because TSMC CoWos capacity going up, but your revenue going down. Can I perceive that you are losing the allocations for the CoWos capacity in Q3? And if yes, why are you losing...

No. Incorrect. Susanna, you're guessing. Your logic is incorrect. So for the CoWos questions, I would suggest you to ask TSMC directly. We, as a customer and as a partner of TSMC, we don't -- it's not suitable for us to discuss the CoWos into too details here. And next one is [ Joey Chai ].

So it's really good to hear that there are lots of inquiries for the project specifically. That seems extremely strong. But what will you say the percentage chance like for an NRE project enter into production? So the question is like more specifically is how committed are your customers that are currently in NRE process for production into 2024 and 2025?

Okay. Let me try to answer that. The most projects we are taking are most leading-edge technology with very high NRE. I can say because of investment, then go to the production chance will be higher. The project we are doing, 07, 05. But at the end, every company does the same thing. Top 5 customers dictate majority of the revenue. But we still -- for us, we still try to take the leading-edge technology customer as much as we can in order to continuously position us as the #1 back-end leaders. We need to build out the ecosystem. We need many projects to build out the methodology. And we need more and more track record. So when we take in a project, if it's a good project with a reasonable NRE, most likely we will take. Yes, we are not only consider about the production unchanged in the possibility. But overall, based on the current trend, the leading-edge technology, we are taking 7-nanometer, 5-nanometer, majority of them go to production. Rather the NPE revenue can give us -- can contribute to our overall revenue too much or not, I think that really depends on the customer's forecast and the customer's view.

Okay. Let me say the project I mentioned into production phase last year, of which -- that means we already knew the forecast by our customer. So that's why I said we consider those projects. I just mentioned our high-volume projects.

Understood. And so one quick follow-on. So do customers have to pay sort of penalty fee if they back off, and let's just say, in the small chance scenario?

You mean back off mean -- what do you mean back off means?

Sorry. Back off from massive production, like scheduled massive production.

No, there will be no penalty. But if that's the case, which means the customer is in big trouble. [ The funding ] investment is totally waste.

All right. Let me try to add to your previous question. I think different customer, we have a different rule for them. For example, the China customer, especially start-up customer, during the production, everything is a prepay. They have to pay us before we raise any order to TSMC. Yes, for some credible customer in the U.S., it's something -- we just prepare the material based on their forecast, but they have to pay us when we ship out the product. For the -- based on past experience, everything goes very well. If they didn't pay any payment, then the [ gain ] will be higher. For China customer, everything is prepaid.

So that's including like for the MP part, not just NRE part?

That's MP part.

And the next one is [ Bill ].

My first question will be on your NRE also turnkey mix. First half it seems like 20%, 80%. And just wondering how are we looking into '24? And also, do we have the potential target or the mix that we'll be looking at ultimately in the next 3 to 5 years? I just wanted to get a sense to see how much growth are we looking at?

[ Bill ], honestly, it really depends on the application, the customer and -- it's hard to tell. But usually, when we discuss the projects in the very beginning, we will have the last forecast by the customer because the shipment volume in the future is a key point for us to discuss the pricing of the NRE. So we will know if there will be production or not in the very beginning when we talk about the projects. Of course, customers and Alchip may have different view on the future production, but, yes, it's a process of negotiation.

Okay. Yes. Let me some more add on that. I think your question is the total NRE versus mass production ratio, right now, it's around to be 20:80. But to be straight, even we have very high confidence to grow our NRE, but NRE can be only grow linearly. It depends on the process technology improvement and also depends on our exact capacity. But as production revenue can increase exponentially, so eventually when our business getting higher and higher, I think the NRE to mass production ratio will keep on reduce.

Yes. I think that's what I'm trying to get into. Because if you look at next year, we are seeing a lot of projects kicking off. So does that mean that for '24, we're going to see lower NRE percentage versus higher turnkey revenue contribution? Would that be the case?

That's possible. That's possible. But 2024 NRE number will be higher than 2023.

I see, I see. So that comes in my second question will be on your NRE growth. How should we look at it? And given the R&D capacity that we have, are we looking to further expand our capacity overseas? I do mean outside of China. What's your plan on this one?

Yes. Our target is always that, at least a minimum of 30% as an internal target. And we try to staff more engineers outside of China. But keep in mind, not all the customer request design outside of China. But right now, we only have 3 to 4 have supposedly asked this kind of question. Yes. So our Malaysia eventually [indiscernible] office established and doing a staffing. We have a very aggressive hiring plan in that region.

Okay. I see there are 2 to 3 people on the line for questions. I would say because of the time constraint, I will just take questions from those people. And the next one, the next one, Charlie, you have extra question?

Yes. Just wanted to help to clarify the CoWos comments, right? So I guess the key message from management, you said, first of all, your allocation support from TSMC is increasing, number one. And also based on those industry information, TSMC's CoWos capacity is also expanding. So I guess just based on our observation, right, maybe there was some technology transition in July or August at TSMC, maybe like CoWos or those type of transition. So I just wanted to make sure that investors are getting the right message. Is it just a [ shortage cut ] and your business will ramp up in September and onwards. Is that the right interpretation on the CoWos comments?

Yes, Charlie, it is right. It is one of the reasons. And another reason is because, as I mentioned in the presentation -- earlier presentation, the second -- if you compare second quarter and the third quarter, in the second quarter, we had the delayed shipments from the first quarter. So you see quarter-on-quarter decline. To us, we really don't consider the declines really, really significant if we follow the very original plan.

Right, right. Yes, understood. That's super helpful, yes. So we can leave it until the next time, but if we have like one minute, would you mind go through a little bit China business because I feel like even under the kind of current U.S. export control, as long as we comply with the EAR rule, there are still amazing opportunities there. Can you give us some color about the opportunity next year from China?

Okay. Let's say this way. We are -- we -- our philosophy is to keep the usual position. As long as we comply all the rules and the regulation, we don't go out to support China customer. But in order to avoid the geopolitical risk, I will say that we try not to touch projects with high sensitivity, very high computing power or some customers that has complicated holding structure. Those are what we are trying to avoid. But at the same time, we're also trying to win more projects from some insensitive segments like automotive. We consider automotive is not that significant to the geopolitical issue.

Daniel said right. In China, for sure, we are focused on automotive. For HPC or AI in high compute power requests and leading-edge technology, I think when we get RFQ, I think the stage, we will conduct a 3-way NDA with a major supplier. If we sign it all together, everybody seen here that is the right customer to do. Yes, for sure, we will take the -- as long as it's leading-edge technology and we consider it's a good project, I think we will take even if it's from China.

Okay. One last thing I want to add is for the geopolitical issue, we try our very -- we try very hard. And I guess, at the same time, some suppliers in the supply chain is doing the same thing. So all in all, let me give you a number. For the second quarter, the China revenue only accounts for a single digit of our total revenue. However, let me say this way. We are not trying to intentionally to avoid projects from China customer. But the overall situation in China IC industry is a little bit difficult, I guess. Every one of you knew that the overall macro environment in China is facing some difficulties. So that's the current situation.

Got it. Congratulations for very great execution.

Thank you. Thank you, Charlie. And next one, Tina Cheng -- Chia. No, no, sorry. Sorry.

That's me, sorry. Just very quick one because no one is asking. I'm talking about the CoWos capacity because the fabless customer actually asking the OSAT to add the CoWos capacity. My question is, is your -- will be very proprietary to the TSMC? Will you be able to work with OSAT if the OSAT also can do the CoWos? Yes, just have your views on this, yes.

I think currently TSMC does the CoWos by itself. And of course, we are trying very hard to get as many CoWos allocation [ as you go ] in many different ways. We leverage our personal relationship. We leverage our position to represent multiple, multiple customers. We represent our position right now for TSMC's CoWos customer portfolio. And in the same time, our customers will work with us together for the CoWos capacity allocation. But when you say the OSAT, I'm not so...

Yes. Let me try to take that. Because of CoWos, I think many customers, especially big customers consider how to use other alternative solutions. To be straight, some of our customers are also thinking about this approach. But when we talk to TSMC, basically TSMC is still trying to offer. They are working very hard to increase their CoWos capacity. So both right now, CoWos-S is facing capacity issue. They have proposed CoWos-R. Some of [ these ] use CoWos-L. And for the -- for customer -- specific customers, if they want to use the OSAT solution, from TSMC's standpoint of view, they have to -- first of all, they cannot support the capacity. And second, the bumping, the CoW, Chip-on-Wafer size in TSMC for the OS portion, they may thinking about using our OSAT's complicated [ ASE ]. Even TSMC, they have a plan to do this. Few of our customers, I think, already started to do the OS part up in the -- using the OSAT solution. But so far, most of older CoWos production design is everything is still coming from TSMC. But starting from next year, I'm not surprised some of the OS TSMC may counter or our customer may use other OSAT solution.

Just a simple one. So if next year, some of the stage of the OSAT capacity coming from the OSAT partner, will this loose up the capacity for you? Or is more or less TSMC is still the single determining factor of your capacities?

So far based on our RF production projection, I think TSMC are still the single supplier for our CoWos design, yes. We have 5 or more customer request CoWos. Unfortunately, the only first 2 are big customers. I think we get a pretty much good support from TSMC so far. So for the smaller customers, I don't think they have too much position to go for the OSAT.

Okay. Next one is Chris. Chris, please? Please. Hi. Chris, are you there? You need to unmute. Go ahead. Chris, we cannot hear you. Okay, if that's the case...

I've been muted. Hello?

Go ahead. We can you hear now.

Okay. I'm not Chris, but I don't know why my name shows as Chris. Okay. My question is related to the growth potential of CoWos, yes. So do we see a substantial growth in terms of project numbers related to CoWos in 2024?

No, I would say yes. But it really depends. It really depends on the customer and it depends on how the election goes. But for the AI type of product, usually the design will apply for CoWos.

Whatever the design request, high-speed memory interface using like HP and cathode design, they have to be CoWos.

Yes. The reason I ask is that for bottleneck being solved, can be, okay, increase the capacity or the other way around, right, because the project number decreased. So obviously, there's still a huge growth potential on the edge PC area. Am I correct?

You are correct.

Okay. And the next question, you sort of answered it. So I'm curious whether this application related to CoWoS has been expanded into other areas or it's still just within this HPC area?

Again, like Johnny mentioned, if the project requires high-speed intercommunication between the processor and the HP end, usually it applies CoWoS. It's not limited to the application. Okay. Jeffrey? Jeffrey? While we're waiting for Jeffrey, there are some questions in the message board. The first one is for the CoWos. Alchip use your own vendor code and for tomorrow, the code name will be customer's code. Actually, it is in Chinese. [Foreign Language] Okay? How much was the bad debt in the second quarter? I just -- we already answered it. And the next one is what's your goal in the target NRE percentage in 2023 full year and 2024? We don't have a target for 2024. We will try very hard to grow our NRE. And naturally, because of the process now migration and the increasing demand from the high-end application HPC design, we do have committed to keep on -- to further grow our NRE revenue. In 2023, I knew we gave guidance for you in the first quarter's earnings call meeting. But for now, I would say the demand for mass production is really, really high. So we will try to be in the high 20s percentage for NRE revenue to [ out of ] revenue. The next question is, hi, management. Third quarter CoWos capacity should be slightly up quarter-on-quarter. Actually, we already answered the question. People are curious about why the capacity expansion in TSMC while we have relatively lower production revenue in third quarter. Yes, I think we already answered the question. And another question is AI server chip are expected to grow by 30% CAGR, and the penetration of AC chip in this market is expected to double from 12% to 24% over the next 5 years, translating to an ASIC growth opportunity of 70% CAGR. Is this representative of Alchip's growth potential going forward? For your question, I would say we are just [ ICD ] company. I knew all the numbers, but for us, our position -- our answer is, first of all, our position is very good. Not too many companies can do the most leading-edge technological design and has the track record to prove it to the potential customer we are capable of doing large-scale design. So we don't want to go into the numeric growth rate for the next couple of years. I would say we just try our best to achieve the maximum growth we can do. And okay, I will say the last question from the message board. Do you have guidance for OpEx this year not counting the bad debt in the second quarter. Actually, we do have it. We guided that the full year operating expense will be about $83 million to $86 million for the whole year. I think we are still on track with this operating expense plan. And okay, somebody asked one last question. Quarter 1 earnings guided 20% year-on-year revenue growth for the 2024. I don't understand your question.

Q1?

I think -- can people hear me?

Please go ahead.

Yes. I think I read it in the Q1 earnings transcript that originally Q1 -- during Q1, you guided like USD 800 million revenue for 2023, right? And then you said...

Oh, I see, I see.

You would be targeting 20% year-on-year growth from your 2024 onwards. I understand that we talk about 2024 revenue guidance, you said probably not as strong as this year. But you're quite confident. I'm just trying to understand, is it going to be more optimistic than what you guided in Q1? Or what's going to look like?

It would be more optimistic since I said that the current shipping -- the biggest revenue contributor this year, we will have even more next year. So to us, we are, of course, more optimistic compared with the expectation one quarter ago.

Even based on a higher base for this year, right? So because before you were comparing to the $800 million number, now you're comparing to a $900 million to $950 million number. Even based on the higher 2023 revenue number, you're still more optimistic on the year-on-year growth going forward to next year. That's correct?

Yes, yes.

Of course. Of course.

Yes, Okay. I guess that's it for the Q&A session. We're already running out of time. And thank you, and thank you for your interest and participation to our second quarter earnings call. Thank you. Thank you very much.

Thank you.