Okay, it's Daniel. Thank you for joining our annual talk. This meeting will begin about three minutes later. Thank you for your patience. Thank you for joining our second quarter annual call. This meeting will be hosted by me and our CEO, Johnny Shen. As usual, this is the safe harbor disclaimer. Yeah, anything related to the forward-looking is just our subjective view. This meeting will be in English. If you need Chinese presentation slides, please go to the MOPS, [Foreign language], to download the Chinese version. Participants, you can write down your question through the Teams application. During the Q&A session, you can use the raise hand function, and we will call your name and unmute your microphone for the question. This video and the audio content of the meeting will upload to, again, MOPS, Market Observation Post System about one to two hours after the meeting. This meeting will kick off right now, starting from our CEO.
Good afternoon, ladies and gentlemen. I'm Johnny Shen, Chairman and CEO of Alchip Technologies. Thanks for joining our investor conference meeting. We appreciate the opportunity to share our Q2 result and provide the guidance for future business outlook. Let me skip this company page. I decide to do this maybe just once a year unless it's a big change. Let me do a Q2 summary. Our Q2 financial result came in slightly below our plan. Primarily, it's due to the delay of N3 tape-out from June to July. As a result, our revenue reached $297 million, net income was $43 million, and EPS is NTD 16.4. A detailed breakdown and analysis will be provided by Daniel in a later section. A few highlights for last quarter. We made significant progress on our most important account.
The top-down design was taped out in February, was successfully proven, and is on track for volume production starting from Q1 next year. Building this success demonstrates our proven design quality, strong schedule control, and effective cost saving. We have already started the next generation design activities. Auto business update. The prototyping and evaluation for our ADAS production chip has been completed successfully. Our end customer is highly satisfied with our result and already placed a substantial wave of orders. The chip remains on schedule for production and is expected to rank as one of our top three revenue generators starting from next year. Meanwhile, the next generation design is also kicked off. Gross margin. Yeah, unfortunately, our Q2 gross margin decreased a bit due to the NRE slip. For this year, we still have a good confidence to achieve notable percentage point increase compared to last year.
That was mainly driven by a favorable NRE mix and improved mass production margins. The last one, ecosystem. Starting from N2 and beyond, HPC design complexity increased significantly. Unlike most of our competitors who adopt a captive model and demand higher margins, Alchip is promoting an open ecosystem approach, inviting a wide range of partners to build comprehensive and efficient solutions while maintaining the lower margins. Our ecosystem has recently expanded substantially with more partners eager to collaborate. In coming months, we anticipate issuing more press releases and potentially announcing investment activities as we further strengthen this model. Future business outlook. As we mentioned on the previous earnings calls, our major mass production line is currently in transition phase. N7 design is approaching EOL, while the next generation is scheduled to ramp up starting from 2026.
With N3 production coming online and next generation design being secure, we anticipate a strong and sustainable outlook for these key accounts through 2029. As for other potential businesses, we continue to see strong demand for HPC and AI-related business opportunities in the North American region, leveraging the advanced technology like N3 and N2. We have already secured several designs and regained confidence in setting a record for NRE revenue this year. Beyond HPC segment, mass production for new gaming and smart speaker applications has also begun to ramp up. Geopolitical risk management. We continuously diversify our business beyond China into other regions. In Q2, only about 7% of revenue are originated from China. In terms of workforce, we have launched an aggressive hiring plan, strengthening our engineers' support outside of China, including Japan, Taiwan, Malaysia, and Vietnam.
Our Japan office can accommodate up to 200 employees now, and our Malaysia and Vietnam office now have over 60 engineers combined. We plan to grow our Southeast Asia workforce to 120 people by the end of this year. As for conclusion, yeah, we anticipate relatively flat revenue quarter over quarter till the end of this year. However, our percentage gross margin is expected to improve, driven by increased customer recognition, higher contribution from NRE revenue, and continued pricing support from key suppliers. Despite the revenue decline, our earning numbers remain competitive with last year's result. Looking ahead, we remain highly optimistic for the AI market from 2026 to 2029. We expect our growth to be in line with other leading industry suppliers, partners, and competitors. We are confident to outperform average market CAGR in high-growth HPC applications. Thank you very much.
Okay, as for the second quarter numbers, we already announced the monthly revenue in the three months in the second quarter. I will do a summary here. The total revenue for the second quarter is $297.4 million, which is a 6.7% quarter-on-quarter decline and a 29.4% year-on-year decline. As for the operating income, for the second quarter this year, we record $37.6 million operating income, which is a 17.3% quarter-on-quarter decline and a 26.6% year-on-year decline. However, because of the higher non-operating income and the relatively a little bit lower tax ratio, the net income in the second quarter is $42.9 million, compared to last quarter and last year, which is a 3.3% quarter-on-quarter decline and a 12.8% year-on-year decline. This net income of $42.9 million translated into EPS of NTD 16.4.
The reason for this lower EPS is partly because of the FX change and the strong NT dollar against the US dollar. Next page. For our application breakdown in the second quarter, you can see the HPC remains the top contributor to our overall sales. 82% of our revenue contributed by the HPC-related applications, mainly the AI. You can notice that 14% of the revenue in the second quarter came from the consumer part. This one refers mainly to the revenue from our gaming-related application. Next page. For the process node breakdown, we are still proud of we are the, I would say, the technology leader in terms of the process node technology in this industry. In the second quarter, we had 81% revenue came from 7 nm and 5 nm, and 5% came from 3 nm and 2 nm, which means 86% from 7 nm or more advanced process node.
We see the trend to continue, and we see the revenue contribution from 3 nm and 2 nm to increase sequentially in the following quarters, especially in 2026. For the breakdown by region, as our CEO mentioned, we try our best to manage the geopolitical risk. You can see right now, currently, we have 79% revenue coming from the North American region in the second quarter. The revenue from Asia Pacific, more specifically from China, has been lowered to single digits in the second quarter this year. For the second quarter business review, mostly covered by our CEO, I will say, first of all, the revenue decline in the second quarter, the main reason is the 7 nm accelerator shipment ended actually in the first half this year. Starting from the second half, we may not have too much revenue from this project.
The current major revenue contributor, the 5 nm AI accelerator to the North American IDM customer, does not show up some strengths to the supplier. I guess many people understand what is the situation. In combination, we have to say that for this year, from second quarter to fourth quarter, the revenue performance could be, I would say, lackluster compared to what we performed last year. However, for the profit, we managed our profit, I would say, to a certain level. The second quarter, 25% gross margin, this was actually lower than our expectation. The reason is because, as Johnny mentioned, a major 3 nm tape-out milestone slipped from late second quarter to early third quarter, which creates quite a significant difference for the NRE revenue in the second quarter.
For the overall gross operating margin, operating income, or the net income, as I mentioned, we consider we manage it in a certain level compared to the revenue decline. The lower operating expense is due mainly to the expense control, coupled with some effect from the lowering ESOP-related expenses. Another reason is the higher non-operating income, which was due mainly to the interest income. For the outlook in the second half this year, we see NRE revenue to post meaningful growth. I think there are many, many reasons. First of all, the most important 3 nm accelerator project, we tape out that in the early quarter. We expect multiple other 3 nm projects to tape out in the second half this year. We still are viewing that the process node migration from 5 nm to 3 nm will be the main reason behind our NRE growth.
In addition to the revenue side, we also expect meaningful gross margin improvement in the second half this year. We also expect we will have 2 nm AI-related NRE to kick in in the second half this year. For overall revenue, we expect the 5 nm AI chip shipment to North American IDM customer to taper out in the second half this year. For the NRE, the overall NRE pipeline remains very promising. We have multiple very good projects related to the AI or HPC applications coming in from the North American region. We also expect the ADAS and the crypto-related production to kick in in the late second half this year, which could be meaningful contributions to our revenue starting from late 2025. I guess this concludes our second half operation outlook. We will go into the Q&A session. Thank you. If you have questions, please raise your hand. Gokul, please.
Yeah, hi. Thanks, Daniel. Thank you very much for your remarks. Thank you, Johnny, as well. First question on the N3 major project. Congratulations on finally achieving the tape-out. Could you talk a little bit about your supply situation now that first cut kind of numbers are probably already available for you in terms of both front-end and back-end wafers? Any changes to the size of the project also compared to your expectations, maybe two to three months back? Lastly, if we look at this customer, they seem to be having some changes in how they procure wafers. At least in this 5 nm generation, they seem to be procuring quite a bit of wafers directly from TSMC. Do you think that this practice continues in 3 nm, or do you think we will go back to the 7 nm procurement process where they procure more completely through you? That's my first question. Thank you.
Okay. Thank you, Gokul. I guess you all noticed that, first of all, we said our 3 nm project taped out in February this year. We also mentioned that there is a very important 3 nm project taped out in early July. I will say, although there are multiple tape-outs, the production schedule remains unchanged. We keep on guiding that for the chip. The production revenue will likely kick in in late first quarter next year. The schedule remains intact. We will have the 3 nm AI-related production revenue to kick in on schedule. As for the working model, we are still working with our customers for finalizing the model, the business model, the cooperation model. No matter how, we do see really, really significant shimmer or revenue growth from these 3 nm AI projects.
Yeah, I think Daniel covered it very well. Different companies, different suppliers have different expectations. Like our competitors, maybe they are focused on the gross margin a lot. That's why the model changes a bit. For us, as long as we can maintain, we can have some increase on the margin side. I think we'll be more open for the revenue side. I think different... Just like Daniel mentioned, we are working with the customer very closely, trying to meet both parties' requirement. As a result, I think we all have a consensus. Starting from next year, we will show a very reasonable growth for this account, I think, from next year and beyond.
Thanks, Johnny. Just to clarify on this one, do you feel that turnkey margins will be declining over a period of time for this project, or do you think that we can still earn the turnkey margins that we were used to earning for this project in 7 nm?
I will say it this way, because this year we do have higher gross margin production revenue as the major contributor. If you compare the blended gross margin between 2025 or 2026, I would say I don't know. If you compare the gross margin in 2024 and the 2026, currently, we have pretty good confidence that the 2026 blended gross margin probably could beat or at least be equal to our blended gross margin in 2024.
Understood. That's very clear. Thank you. My second question is on the progress for the N2 project for the same customer. I think last time, you did mention, Johnny, that we may have some clarity by maybe Q3 or something like that in terms of vendor selection. Given that you're already indicating that some NRE is kicking in, I presume this is also for the same customer, a next-generation project. Could you talk a little bit about what stage you are in and how important are your partnerships that you've announced, especially the one with the Astera Labs, going to be for this particular customer? Thank you.
Okay, Gokul, I'll try. We cannot discuss the details for the projects you mentioned. At the current stage, we can only say that, first of all, as we mentioned, we will have revenue for the 2 nm project to kick in this year. Secondly, as Johnny mentioned, this industry, we need to work with multiple partners to win or to complete projects. We believe the collaboration with our partners for a given project will be a common thing in the future.
Okay. Is it that the vendor selection, at least for the compute, is pretty much done? Is that your current reading, or are you still kind of not certain yet the final vendor selection for the 2 nm project?
I cannot comment on this. As I mentioned, I keep on repeating that we will have 2 nm revenue to kick in this year.
Okay. All right. Thank you. I'll go back to the queue. Thanks.
Okay. Charlie, please.
Sure. Good afternoon, Johnny and Daniel. First of all, very excited about your long-term outlook, a very strong growth. We really appreciate your kind of open ecosystem strategy. I believe there's going to be a way to go. Back to some key questions from investors, right? First of all, you just mentioned that you literally kick off the 2 nm design. In your long-term outlook, right, you think that the major customers' contribution can continue till 2029. Wouldn't they kind of suggest you already win the project, or how should we read your comments?
Yeah, we prefer not to go into the 2 nm project into depth at the current stage. I guess for institutional investors or analysts like you who cover us for a long time, you can feel the difference between this quarter's earnings call and the last quarter's earnings call about our feeling and the attitude towards the 2 nm project.
Okay. I guess it's more optimistic and positive. I just want to make sure I heard you right about the opening remark. Thank you. Also, not just limited to this 2 nm project, right? I'm wondering whether you're doing some circular XPU-attached projects because one of your competitors keeps saying that they do XPU-attached stuff, right? Do you do any like HBM-based die or I/O die in the current projects or future projects?
Okay. Let me try to take this. First of all, a few goals we set up by the end of the beginning of this year. I think most of the two important goals I think already achieved. Of course, number three, we try to diversify our business concentration. Just like I mentioned before, there are many CSP-related opportunities. Before, we can think about not too many CSP considering using Alchip, but right now, I can confidently say whatever they have an opportunity, they would like to let us try. We do have, we do receive many business opportunities from CSP and direct compete with our current competitors. I think different companies have different strengths. I wouldn't say we have some significant progress to win, win anything, but I think we are on track to demonstrate ourselves and also on many accounts.
Every time when we propose the proposal, we see a lot of inquiry. Ultimately, the CSP will adopt the most suitable service provider who can ask maybe the more reasonable price and reduce their dependency on the supplier. Ultimately, they all want to go to the COD type of approach. I can say we get more and more opportunities, and hopefully, in the near future, we can announce a few wins.
Got you. Thank you. That is my second question. I want to touch base on or kind of further discuss with mention about your kind of open ecosystem versus the captive business model, right? Can you elaborate a little bit why this can benefit your customers, even your foundry partner? Secondly, it's a little bit from the concern side that you just mentioned about the COT. I remember for the COT model, it's very difficult for the sensors coming to add value or they charge a very, very low margin. Johnny, can you help us to understand your kind of open ecosystem philosophy or strategy? Thank you.
Okay. For most of our CSP customers, they don't intend to be so-called locked in, and they try to provide COT type of, try to get a COT type of approach using the third-party open solution. In that way, they can control their future production and also manage their budget. Our competitor, I think most of them, almost all of them, are promoting the same approach. Starting from IP all the way to the packaging, eventually COP, everything is a captive for their own solution because most of our competitors are also doing their own product in parallel. For us, Alchip, I think it's kind of completely open. We are a welcome third-party supplier to work with us, and IP, everything is interchangeable.
I think the similar situation is go back to like 30 years ago when during that time, IDM dominated the entire market and TSMC provided the open foundry solution versus a pure IDM. I think most of our customers try to maintain their flexibility in order to control their budgets. For us, I think nowadays, so many solution providers are approaching us. To be honest, it's very difficult for them to win alone. They intend to work with a company like us to provide a more comprehensive and total solution. They really like, most of the people, especially our key customers, really like this model. I think they can continuously enjoy certain savings.
Gotcha. Thank you. Yeah. I guess if I kind of connect the dots, you have a strong partnership with Synopsys, and then recently, Astera Labs. You mentioned about some partnership with a CPO, even some investments. It seems like there's an ecosystem IPO technology vendors are very happy to work with you, and you kind of charge very, very reasonable margins. Kind of very attractive to your customers. Is that the right way to think about this strategy?
Yeah, pretty much you understand it correctly. We will see. You will see more partners coming in the near future.
Thank you. Thank you. I do have several questions, but I will be back to the queue. I will ask you later. Thank you.
Thank you.
Okay. Laura, please. Laura?
Hello? Hi, good afternoon. Can you hear me?
Yes.
Yeah, thank you for taking my questions. I appreciate, Johnny, your comment on your customer identification aside from the current cloud service provider customers. I'm also wondering that aside from the different business model and the approach, we know that previously Alchip is also one of the partners of NVIDIA in the NVLink Fusion. Can you share with us what's the current progress or any interest from your customers on that kind of cooperation?
Okay. Yeah, let me try to take this. For NVIDIA, I think that's a very special partner. To be honest, they are only focused on a few specific customers. Unlike Synopsys or other companies, they are willing to take all projects, including emerging accounts. For CSP, I still believe NVLink Fusion will be the shortcut for them. Before they can complete all the internal solutions by themselves, I think the easiest way is to adopt NVLink to do the scale-up and force you to connect to their switch tray. I think that's the easiest way to build out this instead of doing the huge investment on all the networking and rep-related investment. I think NVIDIA and us are thinking about that's the right way to do if a customer would like to leverage NVIDIA's existing network solution by using their own CPU or GPU. Sorry.
I think that's the easiest way to do. We do have, we propose this kind of solution to many customers. I think so far, so good. Hopefully, in the near future, we have some good progress from that.
Thank you. Do you think that will be a great opportunity when you engage with other new customers, or that could be the chance your existing customer may also adapt that?
Oh, Laura, I would say there are four opportunities. This means we have a better arsenal. Eventually, customers make the decision. If we have a better arsenal, they can choose from our weapons to fulfill their needs. I think to work with NVIDIA, or let's say to be a member of the NVLink, or to be in other alliances with other companies are all positive things to us and, of course, to our customers.
Okay. Thank you. Secondarily, I think it's also related to your business model. We know that in the past, you usually will prepare the math for your customers and kind of depreciate or amortize over the product cycle. We understand that, like Johnny and Daniel, you just mentioned that you are still working with your customers and looking for finalizing your business model. Since their current existing projects, they kind of do the direct cooperation with the key foundry partners. Looking into the future project, like—
Laura, your acknowledgment is incorrect. We do the 3 nm project in the same way with what we did for the 7 nm project.
Okay, yeah, that's very clear. I appreciate that. Since the math expense could be quite high going forward, just wondering, I think if there's any potential fundraising, we would see it in the future?
No. First of all, only a few customers we give terms for the math. Usually, for the majority of the customers, we will collect all the NRE money or probably 90% of the NRE money prior to tape out.
Okay, yeah, that's.
Yeah, our current funding is, I think, good enough to cover all the NRE. Unless the mass production volume peaks to a certain level, very high level, then we are considered about the fundraising. Other than that, I think our fund is more than enough to support tape outs.
Okay. That's very helpful and very clear. Thank you very much.
Thank you. Okay, Bruce, please.
Bruce, you can unmute.
Hello. Can you hear me?
Yes.
Okay.
Yes, go ahead.
Yes, my question is about the addressable market for the ASIC for your business. As we might know, there are a lot of ASIC projects that have been delayed or the tape-out schedule has been delayed here and there. The difficulty for your business is that you have to work with your customers. If your customers have some difficulties, then there's no project, there's no tape-out, there's no business, which happened to your competitor here and there. It turns out to me that your business model requires your customer to be very good to deliver whatever they can achieve, but not good enough to deliver the whole product.
It looks like all the cloud service providers, the design capabilities are very, very different. A good and bad thing is that you have a good customer who has stronger design capability over other cloud service providers, which you continue to win the business and your schedules continue to be on track. It also means that other cloud service providers have difficulties to design a competitive product or are good enough to become your customer. I mean, if they can fill their, if they are not good enough to design their computer or whatever, they cannot deliver the whole ASIC 10, as they want, or they have to be competitive with NVIDIA solutions. It sounds like the sweet spot for your business is somehow a bit tricky, depending on the capability of your customer to a certain way.
Can you somehow tell us how you access the addressable market for the ASIC business?
Okay. Let me try to take this question. Your understanding, I think, is right to a certain degree. The success for tape-out and on schedule customer dependency, we do have a lot of dependency on the customer side. When we are doing the project, we are the partner. Everybody has a responsibility to control the project. Of course, nowadays, I don't think there will be any customer who has no experience and go to N2 and go to N3, N2 directly. The customer definitely has a certain experience and track record from their previous company also. Our job, the reason we can work in, we can control schedule much better than our competitors is because every time we define the project step by step, step by step, customer needs to meet A before going to B, need to B go to C, we are pretty much in control.
When they engage us, we are driving this kind of effort. When customers somewhere need to be improved, sometimes we work with the customer closely, assign more resources, or up to a certain stage, if we feel schedule is facing some trouble, we will escalate to their upper management people. It's not so easy to explain our design service into detail. The reason we can do much better than our competitor, I always emphasize this, is the DN A. Our customer, most of our current competitors, they do not have a service sense. Their main experience is coming from doing their own product and facing with their internal engineer. The reason they can do it quite successfully is because they've been working together for a long time, generation over generation. Nowadays, they start to face the new engineer, which is the customer side.
Sometimes the customers are very demanding and cause some problems for us. We've been doing this business for 20 years. Every time we can easily anticipate the potential delay, and we try to solve accordingly. Based on the current situation and track record, people working with us, and the end result is quite positive, that's why our repeating rate is very high. I don't know if it's an answer.
I was saying, yes, sorry.
I don't really worry about your competitiveness, to be honest. I think you're very good. That's why you continue to win training and all the future projects. I have no problem with that. I think you are better than your peers. The problem, I mean, to make the shared product go to the next level, we need other CSP to adapt it, right? We need more than AWS. We need a lot more. The problem I have right now is other CSP seem to have much weaker design capability, which they cannot deliver a good enough product. That's why, you know, there's no business for us, right? They have to be good. Otherwise, they cannot really deliver what we do to deliver a good, better product. That's the question I have, right?
Bruce, I try to answer your question. First of all, I don't consider it is a tricky thing. I consider it is a trend. The industry, especially for CSP, AI-related infrastructure, everyone is trying to, at least trying to develop their own ASIC. Of course, they have different levels of capability of doing so. Luckily, our customer is the more mature one in the ASIC area. However, we do believe that the others will gradually catch up. First of all, those other potential customers, they are rich. They don't afraid to spend money into building up their front-end design capability. Secondly, they have a broad customer base, and they will, they can calculate how much they can save by doing the COT model. To us, we consider there are a couple of trends in the industry.
First of all, eventually, every major cloud service provider, or let's say hyperscalers, they will have their own front-end design capability at a certain level. Secondly, we consider the cost saving will be a quite important consideration in the future or near future or even right now because the volume and the dollar value of the AI-related chip is getting very, very big. We do consider by doing ASIC, our customer can save a lot. The third one is we consider it is already a winner's game. The AI industry has been there and being hot for several years. You don't see too many newcomers in this supply chain. The reason is because, first of all, any given project, the shipment volume could be really, really significant. Those customers or potential customers won't rely on newcomers, which is considered no track record and no confidence level of Silicon success.
We do consider the whole trends kind of, I would say, favoring companies like us.
What the devil's advocate is that you don't think that Elon Musk is giving up, like, you know, the chip just by NVIDIA is the detour of the trend?
I don't consider that. Tesla, my personal opinion is you cannot take Tesla as a good comparison to those hyperscalers.
He's Elon Musk.
Okay. Yeah, for sure, he's an amazing person. Anyway, I really know what you mean. The entry barrier for the XPU, I think, is getting higher and higher. Customers, I think that's why they need more partners. They need more partners like us. They need more partners like other ecosystem suppliers. If any company tries to do everything by themselves, I think sooner or later they will face a big challenge. The reason TSMC is so successful is because they are so focused. The reason we are doing a good job is because we are focused. Let the expertise suppliers focus on their area. If you intend to do everything, then sooner or later we're in trouble. For our customers, potential customers, I think they should focus on architect. They should focus on their selling channel. They shouldn't focus too much beyond. I think I agree with you.
Customers need to improve in order to control the schedule, in order to make everything working well. On the architect side, I think they are improving quite a bit. Of course, the new thing doing the first design takes longer. When they are doing the second generation, they are improving quite a bit. If you imagine they are taking care of the front end and also thinking about the back end, recently, the back end area is the one that changed the most. Using the N3 as I go to N2, suddenly one tape-out becomes four, minimum three to four tape-outs. There are so many different packaging solutions, so many different architects, so many different technologies. If you expect customers to do everything by themselves, I think that's too much for them. That's almost impossible.
If they rely on partners like us, only focus on their front end architect with our help, I think we have a confidence to narrow the gap and meet their production, improve their production schedule.
I understand. Thank you. That's all my questions.
Thank you, Bruce.
Okay. Next one, Mr. Saxena.
Hey, yeah. Can you hear me?
Yes.
Okay. Thank you so much for taking my question. I just wanted to quickly ask on the Astera Labs and Alchip collaboration. I'm not sure exactly how collaborating with each other improves your chances versus individually going and pitching the U.S. hyperscaler, right? Can you talk us through whether going together increases your opportunity from just the compute die to more than just the compute die and training force? Can you win more together? What is the point of collaborating with them? Does it lead to higher dollar opportunity? Does it lead to higher probability of winning the project? Any clarity there would be helpful.
Okay. I would say, first of all, Astera Labs has been our customer for a long time. We have a really, really, really good relationship with Astera Labs as a customer, as a partner. By doing intensive communication, we consider our technologies are really, really complementary for providing the total solution for the end customer. That's the main reason why we have the strategic alliance together, not for any given specified technology or specified particular customer or project. We consider it is a forefront cooperation for providing the total solution. Yeah, just like Daniel mentioned about Astera Labs and us, it has a long-term relationship. Since they established the company, we have some relationship. I think everybody is so positive for the collaboration like us. First of all, it's purely complementary. We never compete with each other. We have so many common customers.
With collaboration between us, if we can shorten the customer's design schedule on the SoC side or on the system side, I think it's the best to do. Yeah, we have more and more engagement. I know when they have an inquiry, they also mention about our company a lot. I think this is a very good partnership. We expect to have a few significant wins in the near future.
Got it. Thanks. Maybe just to clarify two things. One is, could you just clarify why was there a slip of the NRE revenue from Q2 to early Q3? Secondly, could you just specify, I think Daniel mentioned that you are going to see N2 NRE revenues end of this year. I think I missed it. Can you just clarify whether this is for the large U.S. customer or can you just comment on what project this N2 NRE is for?
We don't comment on specific customers. I'll repeat again, what we can say is we will have revenue for 2 nm projects to kick in by the end of this year. That's the farthest we can talk about here. We don't comment on customer or project level for the 2 nm.
N3 did it.
Oh, okay. N3 did it. Actually, it is a little bit harsh. The last week of June to the first week of July, it's only a couple of days. It happens for engineering. It happens. Not a particular reason behind that. Yeah, we're tracking their schedule by June. Unfortunately, I think last minute, they have some minor changes because the project slipped to a few days and tape out to July, I think. Yeah, again, the relationship between us and this account is very good. We have a high confidence to continuously work with this company.
Got it. Just to press home the point, I know Daniel mentioned this already, but this does not lead to any delay in the mass production schedule, which is end of Q1, right? March next year, there's no delay that you can think of. Is there a possibility it gets shifted out, or do you think it's pretty much set in stone?
No delay.
Okay.
No delay.
Thank you. Thank you.
Okay. Gokul, please.
Thank you. I'll just take one question. When it comes to N2, Johnny, could you talk a little bit about readiness of third-party IP? Do you think the 2 nm AI ASICs will require [448 Gb SerDes], or we can stay at [224 Gb SerDes] itself? If you're going to 448 Gb, do we kind of need to worry about PAM-4 versus PAM-6? There seem to be a lot of technology questions out there. Secondly, on HBM, do you think that we can still stick with the merchant HBM, or do you need to also seek out custom HBM-based die for those 2 nm ASIC projects? I just wanted to understand what is your evaluation of the third-party technology readiness for each of these building blocks?
Okay. Gokul, I cannot comment on customers' IP and also the performance, but I can say for this account, the IP, we have no showstopper and no ending dependency on IP solution. So far, so good for the current generation and also for next generation. We don't need to rely on any competitor's IP or other stuff. I think everything will be controlled independently. For the HBM, same thing. There's no controller or the custom HBM controller dependency. Everything is so far so good. We may need some bridge die, but everything will be designed and controlled between us and our end customer.
Understood. Got it. That's very clear. Thank you.
Thank you, Gokul.
Charlie, please.
Thanks for taking my follow-up question. Just to clarify, Johnny just mentioned the bridge die is a so-called base die that, for our knowledge, is that the same thing with the HBM-based die?
The HBM-based die for this generation, most of our customers are still adopting the generic solution from the HBM vendor itself. Connecting between HBM to the main die, they're still using the UCIe type of approach with HBM5. We decide the HBM bridge chip. I think that's also common practice.
Okay.
Connect the compute to the HBM. Within the HBM die, they're using the standard generic controller.
Oh, I see. Between the HBM-based die and compute die, there's one kind of interface called the bridge die. Is that the one way to think?
The bridge die, yeah, the main just the UCIe interface, yeah, with HBM5.
Oh, okay. Is that part of the I/O die, or is it separate design?
It depends on the, I think different customers have a different approach. We do have a customer intent to put everything in the I/O die. It depends on the orientation. Somebody, if you have a high volume, they can be more flexible by separating the memory interface and PCIe interface. The I/O chiplet is a UCIe compared to the PCI Express. The memory HBM is the memory bridge, is the UCIe with HBM5 to the HBM memory module.
Oh, I see. Okay. Thank you. To Goku's question, are you planning to do some HBM-based die customization? I think there's a one of ten for some other design service providers.
I think the different customers have a different requirement. We do have some customers want us to do this, and some customers doing the bridge only. Some customers discuss with us on the controller as well.
Oh, okay. Gotcha. Thank you. Yeah, thank you. My next question is about your Li Auto business because we keep hearing there's a kind of a second-generation chip also at the 5 nm. Can you expand the potential business opportunity there? Also, outside of Li Auto, what are other China EV customers or even outside of EV, right? What are kind of new projects you are approaching?
Okay. Actually, the current 5 nm ADAS project to Li Auto is running smoothly, except the asset that happened last year because of the entity list thing. Right now, everything is clear. Yes, we are going to do the next generation chip, which will be in 3 nm, 5 nm. Most likely, I would say, depends on the schedule. Usually, the automotive kind of chip, the schedule will be longer. We are in the finalizing stage with the customer for the next generation thing.
Design activity is already kicked off. I think, yes, the design technology changed a bit. Yeah, because last year, just like Daniel mentioned, there's an incident happened. They're kind of afraid there will be further restrictions in the near future. They consider who is using the most leading-edge technology probably will be the highest chance to get started. Yeah, they look at their current business. They think maybe consider the current generation is good enough. By enhancing some features, I think the next generation, we have so much leverage from the previous generation because the same technology we've been, we can control. They consider we can do the takeoff much quicker, which is true, and go on.
The answer to your the other EV car potential customer in China, because the successful producer design, in terms of EO, in terms of turnaround time, we can say we've become very famous in the auto industry in China. Many alternatives, we've been working with some of them. The challenge part is the auto business is unlike a CSP. Customer is so sensitive for cost. A lot of vendors are not even making money. I think the margin-related discussion and sometimes the risk-sharing discussion, I think we are working very closely to our customer because the risk level in terms of auto grade is higher. In the usual case, we need a bigger buffer to overcome that. Unfortunately, some of the car makers do not have enough budget. Those kinds of debating will go on. In terms of opportunity, there's many opportunities in China.
Got you. Thank you. Lastly, just two very quick clarifications. One is that you said the next year's gross margin level could be similar to 2024 or 2025. Can you clarify?
No, no, no. For 2025 and 2026, the contributor is different, right?
Right, right.
For 2024, I consider it is a better comparison.
Right. I see.
Because for 2024, the 7 nm AI accelerator is a major contributor. For 2025, it's different. For 2026, next year, I will say because of some reasons, we are currently more optimistic towards the blended growth model. First of all, we will keep on growing our NRE revenue. Secondly, we consider the blended one could be at least equal or better than what we did in 2024, blended.
I see, I see. Otherwise, it would be a very, very strong gross margin. Lastly, Johnny, at the beginning, you mentioned that you hope your coming five years' CAGR can be in line with supplier or industry peers. What would that be? Is that more like a TSMC is a 45% CAGR or even higher than that?
Okay. We've been warned by the government official [TSC] many times. Overall, I think I can only say whoever announced the number doesn't scare me. I think we have confidence.
Okay, okay. Good to know. Thank you.
Because of the time we take, the last two questions. First is from Sebastian.
Can you hear me, Johnny?
Wow. Hi, Sebastian. Long time no see.
Yes. Hi, Johnny. Hi, Daniel. Can you hear me?
Yes. Go ahead.
Oh, okay. There'll be some background noises in my office room because it looks like the typhoon, the wind is pretty big. I have two questions. The first one is on the second half, the guidance for this year, that it seems like the revenue will go on to continue to be weak because we're still in the product transition period. The gross margin will be higher in the second half because of the NRE contribution. Net-net, what do you think about the bottom line, the net profit as we compare the second half versus the first half? Would it be at a similar level?
I would say, again, Sebastian, we've been warned not to release the numbers guidance, no matter it is revenue or it is profit. If you compare the first quarter and the second quarter, you can see the net profit is only declined by single digits. I would say the same mentality, you can extend it to the whole year.
Got it. Let me phrase that another way. Not forcing you to give a quantitative guidance, but I phrase it another way, maybe easier for you to answer, is that do you think the gross margin improvement in the second half, because of the mix, will be roughly enough to offset the revenue decline?
Not 100%, but pretty good. I think obviously our second half of production revenue will be further reduced, but NRE revenue will increase quite a bit compared to the first half. You can do the calculation and imagine our gross margin will be looked like.
Got it. Got it. So basically, the net profit outcome would probably not be too far from what we've seen in the past two quarters. Okay. My second question is, I think, Johnny, you mentioned that for diversification into the other cloud service provider customers, we are actually, there's plenty of opportunities right now, and we are working with them with multiple opportunities right now. You're saying that they potentially announce some wins in the near future. My question is that can you share with us your definition of the near future in terms of the timeline? I think is it near future, is like within 12 months, or is like a 24, 36 month? Yeah.
Okay. We understand you want to know the schedule. It is not, we are unwilling to provide it. It is not depending on us. We keep on saying what we can say to the market is we consider we are in a good position to win. Honestly, to win hyperscalers in the North American market is hard. You can see not too many suppliers can win projects, especially in the major chip projects in the industry. Our belief is as long as we keep on trying, and right now, compared to two years, three years ago, we have the opportunity to compete. Two or three years ago, we are not invited to join the competition. As long as we can compete, keep on competing, we believe, given our engineering capability, our excellent pricing, and our flexibility of the design and the turnaround time, we eventually have a chance to win.
I think the situation is like that. Looking forward in the future, when CSP people have the opportunity, they will not hesitate for us, let us try. It doesn't hurt for them at all because we are proven vendors. You're asking me when can we win and also when can they hit volume. I think that question is very difficult for me to ask. Just doing this way, just if we rewind for two years, even if we win another CSP account, the revenue doesn't change too much. The top two CSP produce 95% of the ASIC revenue. That's a reality. Unless we can win that one, otherwise, I think the revenue doesn't change too much.
Yeah, that will probably extend or increase your confidence. I think, Johnny, you say that you're pretty confident through 2029. If you win another one or two, then you probably.
That would be great.
Would be confident through 2040, something.
Yes, yes. That's what we are shooting for.
Okay. I have a follow-up on this one. If I look at the, I think we announced, I think there are a few questions during this call on Astera Labs. I think there's a public announcement you and Alchip and Astera Labs made. In my understanding, I know that you cannot talk about it, but just my own understanding is that the potential opportunity or collaboration you guys have right now on your common customers are based on the UA Link. That will probably continue to be UA Link in the next one to two generations. I think both your names, both your two names, Alchip Technologies and Astera Labs, also show up in NV Link Fusion ecosystem partners.
I'm just curious about, is that a coincidence or is it actually, I'm just trying to understand that the collaboration between Alchip Technologies and Astera Labs are not just limited, not just about the UA Link in the near future, but also there's something that potential already happening in NV Link.
Yes, exactly. Our collaboration is not limited to UA Link. It's a forefront cooperation with Astera Labs. That's for your question. To us, we do consider UA Link or NV Link. They are all excellent technologies. Like I said, to us, we are happy to alliance with Astera Labs, and we are also very happy to alliance with NVIDIA. In the end, the customer decides what they want, and we can provide both of them for our customer to do the choice.
Got it. Okay. Just a last follow-up for me, I promise, is that I think it seems to me that you already, Alchip already have most of the things your technologies or partners require for winning the current 3 nm project, which is already taken down, so that you already want it. Also, the upcoming second 2 nm revenue recognition before the end of this year. My question is that you also talk about that you're going to make another fuse of a press release in the next few months on partners' collaboration. I'm just curious about that. Those upcoming partnership announcements, will that be for the existing customers, existing projects, or those are actually for the projects that we don't know yet or you haven't?
Sebastian, it is not related to any given project. Like Johnny mentioned, we are creating, not we are creating, all of the members are creating an ecosystem, a platform for customers to do the project or to do the choice. For the future technology, especially in 3 nm or 2 nm, there are many, many new technologies, such as [CoWoS], such as the CPO. We are working with expertise in different areas together in order to fulfill the customer's needs. That's the main purpose for our alliance with various companies. I think the principle is very straightforward. There's no exclusive for any partners. We are pretty much open. They are well welcome to co-work with us to form a total solution. Customers have more choices to choose. I think nowadays, any specific small vendor tries to win any N2 and A16 project, I think it would be very difficult.
They need an ASIC partner like us. We need a solution partner like that. That's why I formed the ecosystem. As long as there's no confliction and not exclusive, we welcome to add more to join us, provide a total solution.
We can conclude that whatever kind of announcement you're going to have on that kind of partnership collaboration, we can interpret that as the leading indicator or initial signs or initial evidence that you are getting new projects beyond the current biggest customers.
I think a similar situation will be working with Synopsys to win many, but not specifically. I think later on, we are not forming the partnership only for a specific project. We talk about the general trend and also the future trend.
Got it. Okay. That's very clear. Thank you, Johnny and Daniel.
All right. Thank you, Sebastian.
Okay. The last question, Bruce.
If you have one.
Can you hear me? Hello, can you hear me?
Oh, yes.
Yes.
Hello. Okay. Very quick one. For the reason for the ASIC project, right, for the CSPs, the ASIC business right now is highly concentrated in accelerators. I mean, I think there were some projects for CPU, you know, blah, blah, blah. What happened to those CPU projects? Why is that so concentrated in the accelerator right now? Do you also see the CSP will also do the networking chips by themselves, you know, I mean, beyond the current accelerator projects?
Okay. Depends on which customer. For the important customer, we mentioned that because they already have the GPU solution, they also have a CPU and networking. Probably they will continuously do all of them. For the newcomer, I think because the volume and size are totally irrelevant, the GPU can, AI can give them the most saving. I think if I were there, if a capability is not the issue, I will focus on this. The CPU and networking in terms of size are much smaller. I think the different companies, I see different approaches. They all have a priority. In terms of cost saving, doing the GPU first, they can save the most.
Okay. Do you have a rule of thumb, like, you know, like say you need a 1 million unit as a minimum, you know, volume to kick off the project? Do you have?
Yeah, I think for us, we are pretty much open. We don't have a volume requirement to our customer in order to kick off a project. We're choosing customers, I think, from many angles. Some of them, of course, they have a track record for the high volume production. For sure, that's the highest priority for us to take. Some customers, even the startup, they don't have any track record, but their solution we're thinking about is quite attractive. It's quite unique. We can consider that even though the production heat rate will not be too high, we can consider it's a very good R&D project for us. Our company, unlike our tier one competitor, they have a billion or a few hundred million budget to do their own test chip. For us, taking the right customer, working with them, we can consider that's an R&D project.
They pay NRE for us. Even without production, it's quite justified. We don't ask them to commit any project, any volume. I think that's almost impossible for us.
I see. Thank you.
Okay, I guess this concludes our second quarter earnings call. Thank you for your participation. We'll see each other in a few months.
Yeah, in a few months. Okay, thank you.
Thank you very much. Thank you for your support. Thank you.