Alchip Technologies Ltd

TWSE:3661

Earnings Call Analysis

Alchip Reports Record Q2 Performance Amid Strong AI Demand

In Q2 2024, Alchip achieved record revenue of $421 million, a 90% increase year-over-year, driven predominantly by the AI HPC sector, contributing 91% of total revenue. The net income reached $49.3 million, reflecting strong demand, especially from North American clients. Despite lower gross margins of 19% due to high production revenue, the company remains optimistic, projecting over 20% revenue growth for 2025, heavily backed by AI computing and advanced technology migrations. Additionally, a significant part of future growth will stem from automotive applications, expected to be key revenue drivers starting Q4 2025.

Strong Financial Performance in Q2

Alchip Technologies has reported stellar financial results for the second quarter of the year, showcasing a revenue of $421 million, representing a remarkable 80.2% increase year-on-year. Alongside this revenue surge, the company has achieved a net income of $49.3 million and an earnings per share (EPS) of TWD 20.03. This marks the highest performance in the company's history, reflecting robust operational execution.

AI and HPC as Key Revenue Drivers

A significant portion of Alchip's revenue, exceeding 90%, comes from the High-Performance Computing (HPC) and Artificial Intelligence (AI) sectors. The company is actively involved in more than 95% of projects related to advanced nodes, primarily focusing on AI ASIC services. This heavy investment in the AI market is expected to continue delivering substantial returns, as demand for AI-related applications remains optimistic and poised for growth.

Production Challenges and Gross Margin Pressure

Despite achieving record high revenues and profits, Alchip's gross margin was pressured, resting at 19%. This was attributed to a higher ratio of production revenue, which typically yields lower margins compared to Non-Recurring Engineering (NRE) revenue. The leadership has expressed concern over gross margin improvements, indicating efforts are underway to enhance margins while managing production capacity efficiently.

Diversity in Customer Base and Market Focus

The company plans to diversify its customer base beyond its main customer, as evidenced by approximately 78% of revenue derived from North America in Q2. This marks a consistent rise from previous quarters, indicating a strategic pivot toward a more diverse and geographically balanced customer portfolio. Furthermore, Alchip is entering the automotive market, building designs that align with the technological requirements of this sector to fuel future revenue streams.

Future Growth Outlook and Revenue Guidance

Looking ahead, Alchip anticipates strong demand for mass production in AI and data center segments, forecasting substantial revenue growth. For 2024, the company targets a revenue growth rate of over 20% annually. It highlighted that as AI computing demand remains strong, they expect 5-nanometer AI accelerator shipments to become a significant contributor to revenues, likely the largest contributor by 2025.

Navigating Geopolitical Risks and Production Expansion

Alchip is also keenly focused on managing geopolitical risks, successfully diversifying its business outside of China, with less than 11% of revenue originating from there in the first half of the year. This strategic move reflects the company’s commitment to resilience amidst potential market disruptions. Additionally, they are ramping up production capabilities in Southeast Asian markets to enhance service flexibility and cost-effectiveness.

Executive Confidence and Shareholder Commitment

In a move reflecting confidence in the company's future, Alchip's management has increased insider shareholding by over 30% this year. This decision is seen as a commitment to aligning management interests with those of shareholders. The leadership emphasizes a robust business outlook, asserting that 2024 is projected to be another record-breaking year for the company.

Earnings Call Transcript 2024-Q2

[Daniel Wang, executive]

[Audio Gap] second quarter '24 earnings call meeting and this meeting will be hosted by me, Daniel Wang, the CFO of Alchip, and our CEO, Johnny Shen. The meeting will start as usual with the message Johnny wants to deliver to the investors and followed by our second quarter results, and the analysis for our second quarter numbers. And of course, the future outlook for the second half this year and 2025.

After that, we will go into the Q&A session. We will control the time. So the whole meeting will probably be around 1.5 hours, that's the maximum. So thank you for you all for joining our second quarter Institutional Investors Meeting.

Okay. Again, it's a safe harbor disclaimer, as usual. Okay. The meeting -- this meeting will be in English. And Johnny and I are both Chinese. So if you want to ask questions in Mandarin, you can do so. We are more than happy to answer with Mandarin. [Operator Instructions]

Next page. Yes, let's start with Johnny?

[Johnny Shen, executive]

Good afternoon, ladies and gentlemen. I'm Johnny Shen, President and CEO of Alchip Technologies. Once again, thank you for joining our investor conference meeting. We appreciate the opportunity to share our Q2 results and provide guidance for future business outlook. Yes, let me quickly recap our current company status. The company is founded in 2003 and IPO in 2014. Our current market cap approximately USD 7 billion. Since we found the company, we have been successfully tape-out more than 600 design and with 60 FinFET related applications and also more than 17 CoWoS related tape-out.

Last year, revenue is $978 million. This year, I think we show -- obviously, we show a very significant growth again, 90% of the revenue coming from HPC, AI HPC area. We are TSMC 3DFabric Alliance member and also TSMC VCA one of VCA members. Currently, our capacity is to tape-out 20 to 30 a year. Our market focus is HPC, AI, networking and also automotive.

Okay. Let me keep everybody a quick update for Q2. For Q2, we are pleased to announce another record-breaking quarter with revenue reaching $421 million, net income, $49.3 million and EPS standing at TWD 20.03. Combining with the Q1 number, our EPS reached 35.86%. All these numbers marked historical high for the company. However, our production revenue exceed our plan resulting higher income but also causing lower percentage gross margin. The detailed breakdown in comparison will be presented by CFO, Daniel in a later section.

For AI market, allow me to elaborate a bit on the current AI market through comprehensive market analysis and endorsement from both users and suppliers. It's clear potential AI and data center-related market remains highly optimistic. TSMC's recent earning call and technology symposium further reinforced this settlement, with AI highlighted as a primary segment experienced significant growth momentum. In contrast, other market segment are either flat, showing minimal growth or even facing declines. Alchip proudly standing as a premier AI ASIC service provider with more than 17 CoWoS tapeout track record, more than 90% of our revenue are derived from AI-related business. As a leading pure play ASIC player, we remain committed never producing products to compete with our customer. We have a confidence in our ability, deliver more cost-effective service to major CSP and unicorn startup IC companies.

As for the share performance, our share performance seeing significant decline since March, dropping from over 4,500 to even below 2,000 at one point. This decrease occurred despite our consistent record-breaking performance and also our ability to meet the guidance we provide during the earnings call. Management has fully embraced this fact. We are working tirelessly to meet heightened the expectation, the effort, including diversify our customer concentration, improving gross margin percentage and also addressing a rumor for losing key business.

Additionally, our key management staff have decided to further invest our own company resulting in a more than 30% increase in insider shareholding this year. For future business outlook, the demand for mass production in AI and data center segment remains robust. We anticipate substantial revenue growth for the entire year. Looking forward for next year, we have already received a significant amount of demand. We are confident in to further grow our company as long as we can carefully address and resolve capacity-related challenges.

In terms of the new business opportunity, we have secured several design wins from AI startup company in North American region using latest technology like M5, M4 or even improve. Additionally, we have to closely cooperate with major CSP company providing customer silicon, helping them diversify their business dependency. In China, we have to make a good progress in CPU and automotive application. Both the CPU and ADAS design are successfully taped out using TSMC M5 and M58 technology. In terms of production outlook, we expect automotive application will become one of our key revenue drivers starting from Q4 next year.

Geopolitical risk management, yes, we have successfully diversified our business concentration beyond China to other regions. In the first half, less than 11% of revenue are originated from China region, but our China business direction remain unchanged. We continuously believe in and support business in China as long as they are financially healthy and complying with all the rules and regulations.

In terms of workforce, we initiated very aggressively -- a very aggressive hiring plan for engineering in different locations, such like Japan, Taiwan, Malaysia and Vietnam. Currently, we have more than 100 engineers working in Japan. Our Malaysia office already set more than 20 employees. The workforce in Southeast Asia plan to be more than 60 people by the end of this year. This strategy expand designed to be more flexible and cost-effective to meet our customer dynamic requirements in global market.

As a conclusion, we have confidence our business in an excellent stage. We anticipate strong growth this year, 2024 is going to be another record-breaking year for Alchip. Thank you very much.

[Daniel Wang, executive]

Okay. This is the second quarter results. I believe you all -- will know our top line numbers. The total revenue for our second quarter is USD 421 million and the operating income is $51.1 million, which is 32.8% quarter increase and 80.2% year-on-year growth. And for the net income in second quarter, the net income was USD 49.2 million, representing 26.3% quarter-on-quarter and 105.8% year-on-year growth. For the EPS in the second quarter is TWD 20.1.

So for our breakdown in the second quarter, you can see that HPC remains the dominant sector for our revenue. 91% of our total revenue in second quarter is exposed to the HPC area, especially in the AI and the rest 3 applications account for less than 10% of our total revenue in the second quarter. From a yearly perspective, HPC is still dominant, the breakdown that is 92% contribution to our 2024 sales.

Next page. And for the process node, probably speaking, we -- I believe we are still the #1 company in terms of the process node exposure for the advanced technology. In the second quarter, more than 90, I would say, more than 95% of our total revenue is exposed to 7-nanometer or even more advanced process node like 5-nanometer, 4-nanometer, 3-nanometer. And for the year of 2024, 95% of our revenue is exposed to the 7-nanometer or more of the advanced process node.

And for the regional exposure, you can see the revenue contribution from North American region keeps on increasing. For the second quarter, the revenue from the North American region increased to 78% from 77% in the first quarter. And the 2024 for the yearly comparison, you may also see since 2022, our revenue exposure to the North America region increased from 39% in 2022 to 63% last year and 78% this year. We expect the revenue contribution from the North America region will stay in the high ground in the next several years.

Okay. For the second quarter business review. I will say the higher-than-expected revenue in the second quarter was due mainly to the higher-than-expected, the AI ASIC shipment to our #1 customer. And the reason why we ship so many parts to them is first of all, our execution for the production was pretty good in the second quarter. And of course, there are various reasons on the demand side that the customer was pushing us to deliver more parts to them. And in the second quarter, our 5-nanometer AI accelerator to the North American IDM customer started to ramp up. The exact time is near the end of the second quarter. So currently, we already have the revenue contribution from this chip to the IDM customer.

Our second quarter profit is record high, but the gross margin, we are not satisfied with the gross margin improvement, although it improved marginally from the first quarter basis. The second quarter '24 gross margin was 19% improved limited improvement from the first quarter -- from the first quarter number. I would say the slightly lower that we expected gross margin in terms of the percentage was mainly because of the high percentage of our production revenue.

As you may know that the production revenue enjoys less gross margin than the NRE revenue. So as long as our production margin in high ground, the gross margin improvement will be limited. And the nonoperating income, the numbers in the second quarter for our nonoperating income is relatively similar to the first quarter number. So the composition of the nonoperating income mainly come from the interest income we earned in the second quarter.

Next page. And for the business outlook, as Johnny mentioned, the AI computing demand remains very strong. And we do believe that being related application and the demand will still be the long-term growth driver for Alchip. First of all, for the design demand, the AI chip design demand from North American region remained strong from both service providers and the star AI chip companies. And I believe the market, there is a voice in the market that's saying the ASSP the standard product or they say the NVIDIA chip, the hardware performance is so strong, it may slow down the pace for those cloud service providers to develop their own ASICs, in the industry, to be honest, we don't see the trend.

The North American cloud service providers remain aggressively in developing their own ASIC. That's a short thing in -- and I believe it will keep on going for the next couple of years, this trend won't change. And we expect the trend -- this trend will keep on supporting our NI growth, along with the process node migration. For this year and 2025, we see the demand for the process node migration from 7 to 5 to 3 and probably last year, the 2-nanometer revenue will be -- will become another relatively significant revenue composite of our total sales.

For the growth outlook for 7-nanometer AI ASIC to the current #1 customer will remain strong in the third quarter. And we expect the shipments to gradually slow down in the fourth quarter as you may know that this product is entering the end of its life cycle. So we see the shipment to go back starting from the fourth quarter this year. And the MRE pipeline remains promising, as I mentioned, the design demand from MRE -- design demand from the AI sector or from some other high-performance computing areas are still strong.

And for the other product, we expect the 5-nanometer AI accelerator shipments to ramp up quickly in the second half. And we do see this product. This product will become the #1 sales contributor in 2025 and with really, really good revenue contribution and the shipments.

Yes. I guess that's it for the company's presentation for our second quarter and would like to go into the Q&A session. So let's start with Haas from UBS.

[Haas Liu, analyst]

Can you hear me now?

[Daniel Wang, executive]

Yes.

[Haas Liu, analyst]

[indiscernible] what is your view on as your...

[Daniel Wang, executive]

Haas, your voice is breaking up. Can you close to the microphone or something because we cannot understand your -- we cannot recognize your voice?

[Haas Liu, analyst]

Okay. Can you hear me better now?

[Daniel Wang, executive]

Yes, yes.

[Haas Liu, analyst]

So I would like to first ask on the questions on for Cloud ASIC project. Start from Amazon business outlook, what is your view on the [Technical Difficulty] expectation for 2025 and looking beyond the current project phase out, what is the latest update on the progress for the new generation project of 3-nanometer and even beyond?

[Daniel Wang, executive]

Okay Haas, because your voice is up and down, up and down quite significantly. So I try to understand your question. And I have to say, please don't mention the names of the potential customers. Okay, for our biggest customer, I would say, the current project, we do see the shipment to go down, as I just said, to go down starting in the fourth quarter. In 2025, I will say, if you do the year-on-year comparison, the revenue or let's say the shipment drop will be significant. That's based on our experience. But how many -- how many parts we can ship in 2025 still depends on the customers' demand and the 5-nanometer chip, the status for the 5-nanometer chip ramping by the others.

So I can give you a firm number for that. But for sure, in 2025, this 7-nanometer AI chip will not be the top contributor to our total revenue for sure. And for the future projects, in the earnings call, I can only say we will work closely with our current #1 customer for the generation by generation projects. Besides that, I'm sorry that I cannot give you additional information about it.

[Haas Liu, analyst]

Okay. Yes. So just 2 quick follow-ups. I hope you can hear me better now. I think for the contribution for next year, is it still going to be like over 10% from this customer on 7-nanometer? And for the next generation chip sets or the follow-on projects you are working with your customers, for the contribution in the future, what is your current expectation on the cadence?

[Daniel Wang, executive]

Yes. There will be contribution, but I don't -- I want to be responsible for what I'm saying. So I really don't know what will be the percentage of the contribution from the 7-nanometer AI chip shipment to the current biggest customer. As I mentioned, it will not be very significant to our total revenue based on our experience.

[Haas Liu, analyst]

Okay. Yes. And then switching gear to the GPU project. You are ramping for your U.S. IDM customer. What is the latest expectation on the ramp in the second half and also contribution in 2025. And if possible, could you let us know how the project profitability relative to the current cloud ASIC?

[Daniel Wang, executive]

I will say, in this year in 2025, the 5-nanometer AI accelerator, we try -- we currently -- we are signing our best to ramping it up because the demand, the forecast -- the forecast number for the future demand is quite significant. We are -- trust me, we are trying every bit. We are working very hard to push the shipment. And I expect for this project, 5 nanometer AI accelerator, the contribution this year to account for I would say, still single digit, but on the higher end of a single digit to our total revenue.

[Haas Liu, analyst]

Okay. That's this year, right? How about next year? Is it going to be pretty similar to your largest customer this year from the sales contribution percentage perspective?

[Daniel Wang, executive]

I would say next year will be very, very significant. But there are still some -- there are 2 upsides from the current forecast. First of all, for the overall numbers, I will say probably based on the current capacity support by our suppliers, the total revenue contribution may not led to the contribution for our current #1 product in 2024.

[Johnny Shen, executive]

Okay. Let me add some color to this. The demand -- the order we receive from this particular customer actually, this is very high. And we still have some complexity related challenge for the CoWoS-S. Just saying that if we can solve the CoWoS capacity-related issue and the total revenue we expect from these account is no less than the #1 customer contributed by this year and even with the upside.

[Haas Liu, analyst]

Okay. So the current status, just let me clarify it. The current status is that based on the cohort supply you can secure now that the contribution from this customer is going to be no less than the top customer contribution this year. Is that the correct understanding?

[Daniel Wang, executive]

It is correct.

[Haas Liu, analyst]

Okay. Okay. Yes. And a quick follow-up on this U.S. IDM customer will be related to the project sustainability. If the customer does not have a follow-on project beyond the projects you are going to ramp on 5-nanometer. Will you have any opportunity to work with its parent company? And how should we think about the business model beyond this project?

[Johnny Shen, executive]

Yes, to be honestly, we all know this customer encounters certain challenge from internal or external. I can tell, based on my understanding, the current -- the next generation will be slowed down a bit. But so far, I heard the few of the progress they make us working with other supplier, they already designed the packaging solution, they were continuously using CoWoS-S.

And as long as they are using TSMC CoWoS solution, we consider we have a much higher chance to get involved for this business. Yes. As you know, there's a certain competition between the package solution on these 2 companies. And based on our suppliers' corporates -- yes, as long as this business stick with the CoWoS solution and Alchip's involvement will be the preferred model. But in the other word, if this project seeing some delay I'm thinking about the positive side for longer lifetime cycle for the current project. Yes. So to answer your question, there are -- is that going to be the next generation or not, we don't know. But definitely, the schedule will be longer than we originally expected.

[Haas Liu, analyst]

Okay. So if they are going to continue to use the CoWoS-S or CoWoS related technologies, that would mean that the front-end manufacturing will still be outsourced. Is that the correct interpretation?

[Johnny Shen, executive]

That's what I understand, yes.

[Haas Liu, analyst]

Okay. Yes, that's very helpful. And then my second question will be related to the production business outlook in 2025. With the ups and downs you just mentioned for the projects you are helping your customers, could you let us know what the total front-end manufacturing wafer and also CoWoS capacity you are preparing for your customers next year on a relatively basis -- on a relative basis? Should we expect more wafer and also CoWoS consumption from your customers in aggregate based on the current demand forecast from your customers? Or will it be relatively stable in 2025?

[Daniel Wang, executive]

You should know that we cannot disclose this type of information to you.

[Haas Liu, analyst]

Okay. But I mean just from a year-on-year comparison, what is your expectation on the current supply you can get from your foundry partners or your supply chain partners. Yes, on a relative basis on a year-on-year versus this year.

[Daniel Wang, executive]

I don't consider it is a good comparison because they use the different process technology. And the CoWoS -- the CoWoS are also different. So I can only answer you that the TSMC is supporting its project, but the allocation, the capacity allocation to us, we are not satisfied with the capacity allocation. So we will still work with TSMC in order to get more. And I would say that's what I said, there are 2 upsides for this project.

First of all, we can get more CoWoS allocation from TSMC in the future days. And secondly, we are also prepared to -- if TSMC agrees and the customer agrees -- the packaging, we are thinking to have a third party to involve for the supply. So that's the current plan.

And Haas, can you squeeze [indiscernible] to the assets.

[Haas Liu, analyst]

I mean yes, just squeezing one more question. Yes. I mean for 2025 on your AGM, you mentioned that you target to grow pretty nicely in 2025, even if there's a business transition, so I was just wondering what is the realistic target that you would share with us that for our base case scenario or even the worst case scenario that everything is or things just get worse than your original expectation? That's #1.

And the second quick follow-up, if I can squeeze in, is probably just on the production capacity because foundries have been announcing that they are going to more than double their CoWoS capacity. And as you just mentioned, that the third-party OSATs are also expanding their capacity as well. So I was just wondering, from a total amount perspective, for all of your projects, are you just getting more CoWoS supply from your partners? Or will it be relatively flat now for next year.

[Daniel Wang, executive]

Maybe that's your last question. Okay. For your question, I would say our view does not change. We are still very confident that we can grow our total revenue in 2025. And how much we can grow it still depends. Like I said, there are still some dynamics to swing our revenue next year. For example, the upside we can give for the CoWoS or for the supply capacity support.

And secondly, for our autonomous driving chip shipment, if we can do everything very well, the revenue contribution from this project to see our previous expectation in 2025. So again, I want to say that the view is not changing. We will deliver revenue growth in 2025. And again, in usual years, we target 100% plus every year. That's our revenue target. That's the target we ask our sales, our engineering to deliver every year. So we are still shooting for more than 20% revenue growth every year. Okay.

Robert, please. Robert from JPMorgan.

[Robert Hsu, analyst]

I just have 2 quick questions. First, on your [ USSPAI ] ASIC projects. I think during the opening remarks, you kind of talked about there's some market rumors. So can you talk about the spills on these market numbers by probably sharing what's your design progress so far? When do you expect the chip to tape-out within the mass production? And as a result, how should we think about revenue opportunities into 2026 for this projects?

[Daniel Wang, executive]

Okay. Again, in the earnings call, our saying will be for the project will be -- first of all, we will keep up working with our current #1 customer for the future generation project. And we are spending a lot of engineering resource project on the project right now. And I want to remind investors that -- if you look into the AI sector, especially accelerator ASIC sector until today -- until today only 2 major suppliers, the #1 is Broadcom, the #2 is Alchip. So I would say the fact speaks for itself that we are a true competitor within this sector.

I knew that there are many rumors probably from our U.S. competitors or from our Asian competitors. But I would say we're trying to focus. We try to deliver our engineering value to our customers. And because of this, we are very, very confident that 2026 will be a very positive year for us because there will be accelerator projects into production and the shipment volume will be very significant.

[Robert Hsu, analyst]

Okay. Yes. My second question will be on the IDM customer. I think you previously talked about your core supporting rate was 50%, 55%-ish over the past 3, 4 months. But today, I think you seem to be securing more cost capacity at this point in time. So how should we think about the supporting rate right now?

[Daniel Wang, executive]

Obviously, it is still relatively -- but obviously, we do feel there will be room for improvement. And even with these supporting rate, the revenue contribution is significant. So our guidance for these products contribution is based on the current capacity allocations, not the optimistic version.

[Robert Hsu, analyst]

Got it. Got it. Okay. My last question will be on more of the long-term picture. So can you talk a little bit about what your -- the projects that in your bidding pipeline, say, AI Accelerator versus a server CPU, which area do you think that you're likely to win the next CSP projects, say, in the next 6 to 12 months?

[Johnny Shen, executive]

Let me try to take this. As I mentioned before, winning CSP project indeed many, many -- need a lot of corporation. Yes, I can tell you right now, we are working closely to almost each of the CSP on their next generation, starting from N2 or A16, design will become very complicated, either minimal will be CoWoS-L type of chiplet or even the 3D. So we need to do a lot of cooperation, including test chip in order to demonstrate our solution. Just for your information, we are going to tape-out N2 test chip next year.

And for this chip, we started working with some target customers, put some of their IP inside and to test the performance. It's pretty much the first GAA FET. We are doing is a beyond FinFET now. So again, there's a certain progress, but it's a little bit too early to tell to confirm winning. But we are -- I can tell you that we have a very high close relationship working with the key IP partner working with each of CSP very closely. But at this moment, no confirm we need from any of their AI design yet.

Okay. And Charlie, please.

[Charlie Chan, analyst]

Again, congratulations to good results. I think people can understand that the EPS upside, there was some gross margin dilution, so great execution.

So I have several follow-up questions. Maybe start with the 2-nanometer. So I'm wondering because your industry peers also talking about they want to do a 2-nanometer test chip. So number one is that whether customers are paying for those test chip? And secondly, for the customers who are working closely. Do they also use other design service to do similar kind of practice, right, meaning 2-nanometer test chip with your competitors?

[Johnny Shen, executive]

Okay. So the test we are doing is try to catch TSMC's shuttle. TSMC has a 2-nanometer shuttle maybe twice a year. The next one will be September time frame. We are working with close IT partner and customer and few customers try to put some of their IP inside and also for our design-related the staff, try to do the comparison with N2 and N3. So LRE is not contributed by customer, but they know they are taking some credit from us. Yes, we are working with TSMC. TSMC is a quite supportive and we can tell there will be insignificant costs for us in terms of material, we continuously get good support from TSMC. Yes, with this, we have a good position to put the potential customer and also IP vendor together. We've been doing this for years, including N3 and N2 and also N5.

[Charlie Chan, analyst]

Yes. So I would assume customers are also trying out other design service or IP providers for 2-nanometer. So that's just you, right?

[Johnny Shen, executive]

Maybe.

[Daniel Wang, executive]

We don't lie what we don't know, yes.

[Charlie Chan, analyst]

Okay. And speaking about TSMC, I'm wondering whether you sense that TSMC may raise their wafer price to all the customers? And what would that mean to your margin or revenue in 2025?

[Daniel Wang, executive]

I would say we haven't received the official notice yet, but in the industry, it seems everybody is thinking, TSMC is going to raise the price. But for TSMC for the pricing, it is complicated. It is not okay, we raise every cost to buy 10%. It is a really complicated mechanism for the pricing of TSMC. So we are just waiting for the notice. And to be honest, we don't consider it is -- it will be an impact if the TSMC changes its pricing because for the majority of our revenue, for the production, we are running the cost plus margin, as we mentioned many, many times. So we can pass through the increased cost of TSMC to the customer for sure.

[Johnny Shen, executive]

Yes. I think Daniel is absolutely right. And TSMC has increased price before. But after we do some analysis, I think every time TSMC do some pricing adjustment, there will be benefit to us. But to be straight, I don't think TSMC will raise the wafer price for every single application. Maybe AI, HPC, I think, related definitely for N3 and N2 is the target to raise the price, but for consumer or auto maybe remain the same because I don't worry about our CSP customer, but for consumer or auto customer, if the pricing increase, that will be -- they were facing a certain challenge. For CSP, I think to be honest, we will totally work on TSMC to adjust the price.

[Charlie Chan, analyst]

I see. And lastly, I mean, there was an announcement about your -- one of your key customers do their private placements at Amazon. So it was after your last earnings call, right? So even today, some overseas investors are wondering why I want to take these investments? Is it so small and probably will create some potential intrinsic conflicts, right, because they are also engaging with other CSP customers. So can you elaborate your rationale publicly on this taking the private placement from Amazon?

[Daniel Wang, executive]

Okay. Okay. Basically, there were agreements between both parties that we cannot say too much about the reasons behind the private placement. And of course, it is hard to believe that Amazon wants to make some money from the private placement, it is not true because of the math. And I will say, I -- for the company's position, we consider it is an encouragement by our customers. So beyond that, yes, I would say we cannot talk too much about this private placement.

[Johnny Shen, executive]

Yes. Charlie, sorry, it's very sensitive. I think we can only tell I think the -- obviously, our relationship will sustain. So that's why they are continuously invest on us.

[Charlie Chan, analyst]

Okay. Yes. I guess we just want to make sure there will be no impact to your engagement with other customers? I guess as we wanted to.

[Daniel Wang, executive]

Very little investment, I think, is insignificant. Yes, if other company wants to invest a little, it's also welcome.

And Laura from Citibank, please.

[Chia Yi Chen, analyst]

Yes. My question is actually about your technology migration. And how would that impact your revenue and also the gross margin outlook. We know that up to now, the majority of your revenue coming from the 7-nanometers project. And into next year, it seems like the N5 will be the majorities, so -- but the wafer pricing for N5 is significantly higher than N7. So even though management didn't really provide a concrete guidance for the next year, but based on the ASP premium that already be quite significant. So I'm just wondering that when we move into maybe 2026, assuming that N3 will also ramping up, so can you give us more pictures about the wafer process node migration of your major revenue? And also, how would that impact your gross margin as well?

[Daniel Wang, executive]

Okay. First of all, I would say probably maybe the next earnings call, I will separate it -- I will do the change for the categorization of the process node breakdown to separate the 7-nanometer, 5-nanometer and more advanced process node. And for this year, 2024, the 7-nanometer remains a main contributor, in 2025, the 5-nanometer product will become the major contributor. And for sure, 2026, the 3-nanometer project will be kind of main contributor.

But for the blended gross margin, I would say, the breakdown between production and the NRE revenue is the key designing factor to our blended gross margin, not in the process node. And then of course, when the process starts moving, we may grow our NIE revenue more effectively. So I would say for the next year, we expect the NIE growth will surpass the production growth over last year, supposedly the blended gross margin yearly will improve on this year's basis.

But for 2026, based on the information we received from customers, we -- I would say that the production revenue percentage could be very, very high because the shipment effect to vary a bit. So the blended gross margin, probably will be down because of that. So that's the direction for our blended gross margin. So for the process moving, I think it has higher correlation with high revenue growth instead of the blended gross margin.

[Chia Yi Chen, analyst]

Okay. That's clear. And also, can you share with us that Johnny already talked about your working on N3 and also N2's wafer shuttle as well. Can you give me or us more color about what type of the projects you are working on for N3 and also N2?

[Daniel Wang, executive]

Most of the -- you can say almost 100% as you see and majority of that is the AI-related project because you -- you may know that. For now, AI application, the application usually applies the most leading edge process node because it requires very high-speed computing and interface communication. So usually, when you hear about the most leading-edge process node, the 2.5D/3D packaging or even the newer technologies such as steel, the system wafer, the 3D IC, I would say, those are related to the AI applications.

[Johnny Shen, executive]

Okay. Let me try to add some color to that. For this year and also the year before we consider N7 is our typical application. We treat N5 as advanced, so between N16 as mainstream. But starting from next year, N5 will become our typical process. There will be many N3 technology as advanced technology, but of course, there will be a certain N7 as well.

In terms of N2, I don't think that any design will be in production next year. The best case will be in design stage. And starting from next year, we can say the production will be N5 as a primary. The design start will be most of the leading edge, most of the customer will target N3.

[Chia Yi Chen, analyst]

Okay. That's very clear. My final question is on the ARM-based CPU. We know a lot of your peers also discuss about the ARM-based CPU. But we know that -- we know that probably it's not really your first priority at the moment, but just wondering what's your plans on that type of business opportunity.

[Daniel Wang, executive]

Okay. I would say our top priority for sure is the accelerator because as I mentioned, the accelerator usually have to apply for the most leading edge process node and the design will be relatively -- will be much more complicated than the CPU -- ARM-based CPU. So accelerator type of project is our top priority. And for ARM CPU, we consider it as a way of penetrating into new customers. But to be honestly, because the designs that complicated and usually, those projects don't apply complicated packaging and testing technologies.

So you may understand that many of the project is what we call PD3 or CK3, whatever you call it, the production-only project. For the production-only project, you may know that in the past, we try not to take the production-only project because we think we cannot deliver our value to the customers. So for this type of value-added projects, we will try to compete if the project comes from the cloud service providers in North America. However, we have our minimum requirement for the value which also means the profit margin.

[Johnny Shen, executive]

Okay. Let me add some color to that as well. Yes, for ARM-based CPU to replace X86 in the server side, I think one -- I think the -- our major customer, I think, doing a great job. I think a lot of service providers try to borrow this trend. But in terms of design, CPU design is much -- GPU is much, much more complicated than CPU, yes, for current generation. As you know, CPU does not require HBM, does not require the CoWoS. And they are pretty much with the current generation, no matter it's N5 or N3, will be homogeneous kind of approach.

But we're thinking about the next generation will become our opportunity. As you know, the CPU need to use a lot of SRAM, much, much more than the CPU, but SRAM scaling in fact is saturated. So in order to make the chip more cost effective, SRAM will be no longer go to the most leading-edge technology they will stay at N3. And compute die will continue to go for the N2. So they have to use a heavy bound to connect everything together, and that will be chiplet-related approach. So those kinds of design become very complicated.

Yes, we are thinking about [indiscernible] will be our opportunity to penetrate on the next generation because the design entry barrier will be much higher and the customer may no longer only need PD3 or [ CK3 ] kind of approach. But right now, I think the market is to product, if it is only production-only kind of business.

Okay, [indiscernible], please?

[Unknown Analyst]

Yes. I just have a quick follow-up. How should we think about the future nonops income or interest income, given we now have a very excessive cash on hand -- and when do we expect the cash level to decline again in the future, probably in the next-generation turnkey rent or consigned for wafer, HBM for customer. I want to know your thoughts.

[Daniel Wang, executive]

Yes, we do have a good cash position. However, we expect even with such cash position, the 2026 will be quite challenging. As we mentioned, based on the information currently we received, the production scale for the 3-nanometer project will be very, very significant. It may need plenty of cash to support the project. So that's the current situation.

Okay. Robert?

[Robert Hsu, analyst]

I just have a few follow-ups. On the US CSP projects. How should we think about the gross -- the production gross margins compared to your N7 projects? Because when I look at the HBM content, I think it should be much higher. And also, you also got the competition heating up for this one. So how should we think about the gross margin?

[Daniel Wang, executive]

You are saying the project to the IDM customer.

[Robert Hsu, analyst]

No, US CSP N3.

[Johnny Shen, executive]

[ US CSP N3 ] project. Okay. Yes, let me try to take this question. How to improve the gross margin, I think obviously, there will be 2 ways. First of all, customers consider your contribution, appreciate your capability are willing to pay more. The other one is working with the supplier. Our current supplier, I think we can tell it's a fair to a certain degree. Once your volume increase, you will get related support. So obviously, for the past year, we are growing quite significant. Yes, we are -- we continue -- we start to get some support, significant support from a major supplier, including foundry, packaging and testing. It's impossible when you are revenue at $500 million and grow to $1.5 billion, you're still paying the same price.

So that's why -- for the future, we have much higher confidence to improve our gross margin. Yes, right now, for the design already done or desire facing EOL is very difficult to get additional support, yes. But for the new business, there is a lot of room for us to negotiate with the supplier.

[Robert Hsu, analyst]

Okay. My second follow-up would be the 2025 numbers. So I think we talked about that the IDM customer N5 project seems to have some upside if we can get more supply -- but is there any likelihood that we could see earlier contribution for less N3 projects with the US CSP.

[Daniel Wang, executive]

Of course, we want it. The customer want it. Everybody wants the project to go into production quicker. However, there are still engineering things needs to be done. So we try, we will try. But I would say the chance is not that high.

[Robert Hsu, analyst]

Okay. My final question will be on the capital allocation. I think when the projects -- the N3 project enters production in '26, it seems to have -- need to have fund or source more funding to -- I mean, to amplify your cash position, we will do that to additional share issuance again or you will see other ways to fund the working capital?

[Daniel Wang, executive]

Well, we don't have the trend so far. And as mentioned, we have good cash position, and we will try to handle the cash flow and the working capital issue internally. First of all, we will try to get better terms with our customers.

Secondly, we will talk with the suppliers in order to get more credit and better terms with suppliers. So for now, we don't have a plan.

[Johnny Shen, executive]

Robert, I think it's a very good question. It's too early to make a decision because there are so many dependency, the credit line from the major supplier, interest rate and also our share price. Everything is we need to put into consideration up to that stage, we will try to get the best solution from the company.

[Daniel Wang, executive]

Okay. I don't see anymore hand raising -- Mike yes, please. And I would say because the reading has been low. So it will be the last question since. Mike, please...

Okay. There could be some problem. And we have -- okay, then there is a question on the message board.

Hi, Johnny, Daniel, based on your current CoWoS allocation, what your fulfillment rate for the -- for the second half '24. And for the 7-nanometer project shipments decline last year, is that due to demand performance reasons or customer move in-house. On competition, please share the growth profit margin trend for projects that you see. And if there is a minimum threshold for Alchip on project margin?

Okay. Let me give you the answer question by question. Based on the current CoWoS allocation, what's your fulfillment rate for the second half? I would say, for the second half, we are -- actually it's ramping up phase. The production and shipment scale is ramping up in the second half. So for these -- for this space, there is no so-called fulfillment rate, actually we can get for support from the suppliers.

For the 7-nanometer project, the shipment -- for the 7-nanometer project shipment decline last year, is that due to demand performance reasons. Now actually, the 7-nanometer chip has a very long life cycle. Maybe you can remember the shipment started in 2021. So it's natural, it's like the project is entering the end of the life cycle. And there will be new ships to replace the 7 nanometer one.

And the third question is on the competition phase. Please share gross margin trend. I would say the [indiscernible] and production is quite separated. For [indiscernible] usually, we can still get reasonable, I would say, 40% plus gross margin depends on the project. It depends on the process node and depends on the scale of the design. So it varies a lot. And for the production side, usually, for the smaller production scale, we can get 15% to 20% or even higher and from the U.S. customers.

And from China customers, we can usually get even higher. But for some large -- for some large, with very large production scale, the gross margin could be still from 10% to 15% -- so I would say, currently, the rules, the packets in the industry, we still follow to no particular change or special cases for our business.

Okay. Mike [indiscernible] from BOA, probably, they have some malfunction with the equipment, so he write down these questions for the next-generation product by U.S. IDM customers, is only involved in the packaging design, then should we expect a lower growth margin versus the current generation.

Mike, I don't really understand the meaning of the next generation. I assume you are saying the project beyond the 5-nanometer one. I would say the -- Johnny saying because based on our knowledge, the project will be still in TSMC, no mater the wafer or the packaging. So we have pretty good chance to still offer our services, of course, including the turnkey services to this project.

[Johnny Shen, executive]

Yes. Mike, this design will be delayed. It's not closer to tape-out year. So whether our involvement, right now, I think it's -- we cannot disclose at this moment, it will be too sensitive, but we are continuously working with this particular customer. And we -- on the previous N3 test chip, yes, we are already -- 3 party already working together, including these customers of TSMC and us and whether we have a chance to get involved or the design that we don't know yet. Because that company is in the transition period, we expect they may need more support from us because of the -- yes, again, that particular customer facing a certain challenge at this moment.

[Daniel Wang, executive]

Okay. There is another question from Mike and I will say it's the last question for this earnings call. And this question is -- we have witnessed a very quick response by Alchip in the previous incident of sanction, patent, and the other major China customers. So the question is, if there is no design win finally for the advanced node projects you are working on with CSPs. Can you also react quickly in results according or should this based on the premise of very good design demand activity.

Mike, your question is a little bit weird. We are working as a design service provider. And we are very confident that we have the world-class engineering and the technology to serve the potential customers. So of course, we want to win projects from other CSPs in North America. So it's really hypothetical that you assume it is an incident for us not winning other CSP projects. We will keep on trying. We will keep on trying for every good opportunity in the North American market. That's our answer.

Okay. Okay. I guess that concludes our second quarter earnings call, and thank you for your participation today. And have a nice day. Thank you.

[Johnny Shen, executive]

Thank you very much. Thank you for supporting us. Thank you.