Good afternoon, analysts, fund managers, and ladies and gentlemen. Welcome to Alchip Technologies' third quarter 2026 earning call. We are going to present you our third quarter results and the message from our CEO, and the analysis for our third quarter numbers. Thank you for your patience. We will start immediately. This page is regular safe harbor disclaimer, nothing special. After our presentation, when we get into the Q&A session, please use the raise hand function through the Teams. We can call your name, and you can ask your questions. This video and the audio content of the meeting will upload to [Non-English content] about one to two hours after the finish of our meeting. So, if you cannot download the content from the website, please be patient. Sometimes it takes time to upload the whole content.
We would like to start with the message from our CEO, Johnny Shen. Johnny, please.
Good afternoon, ladies and gentlemen. I'm Johnny Shen, Chairman and CEO of Alchip Technologies. Thank you for joining the investor conference call today. We appreciate the opportunity to share our Q3 financial result and provide an update on our business outlook. First of all, let me give you the summary of our Q3. Our third quarter result were in line with our plan, with all key financial metric improved significantly quarter-over-quarter, driven by a substantially higher production revenue. The total revenue reached $ 819 million, up 238% quarter-over-quarter, and 267% year-over-year. The net income reached $ 102 million, exceeding $ 100 million for the first time in a single quarter. EPS hit a record high with a TWD 37.68, which is surpassing most of our Taiwan-listed industry peer.
CFO Daniel will present more detailed financial breakdown and analysis in the next section. Okay, for our third quarter numbers. Q3 business highlight and challenges. There are several highlight from last quarter I would like to share. First, our N3 accelerator entered the high-volume production smoothly last quarter. The chips performed very well, meeting all test criteria with a reasonable yield. We expect the current generation will start in production for approximately two year before the next generation takes over. We expect the revenue for this project will be slightly higher in Q4 than Q3, and substantially higher in 2027 than in 2026. Meanwhile, the next generation has entered the final implementation phase. We plan to begin a series of tape-out this quarter to meet the planned production schedule. The project remain on track, entering prototyping with a low-volume production targeted for Q4 next year.
One most significant achievement last quarter was securing another major design win. The contract has been signed, the project has been kicked off. It is on track for high-volume production in 2027. We expect this project to generate approximately $1 billion revenue with a reasonable margin, marking our second-largest revenue contributor starting from next year. Beyond our data center business, our other production businesses performing reasonably well, with a strong demand despite the suppliers' constraint. Our automotive ADAS project in China continues to see the strong demand, and customers' next-generation design remain on track for tape-out in Q4 this year. Other application, including CPU, gaming, consumer, collectively generate $100 million plus revenue last quarter. I would like to also share two major challenge we faced last quarter. The first one is a supplier constraint. Several production project was affected by shortage in wafer, substrate, or package capacity.
Without this constraint, we believe we could easily generate another $100 million revenue last quarter. The second challenge we have is our gross margin. Supply constraint affect other high-margin production business, combined with delay in recognizing NRE revenue from a few project. As a result, our N3 AI accelerator account for unexpected larger share of Q3 revenue. The revenue mix drove our gross margin sharply lower to below 20%. Given this effect, we believe the Q3 gross margin was close to a worst-case scenario for our business mix. For future business outlook, like the industry leader and some of our competitors, we remain very optimistic about the long-term outlook for AI market, particularly in ASIC segment. We believe our competitive position is stronger than most of our peer, as we continuously to serve as a sole design partner for our end customer.
In addition, our customer maintain one of the industry largest level of CapEx. It create a substantial SAM serviceable available market for Alchip. For Q4, we anticipate another record-breaking quarter, driven by higher production revenue and major design tape-out activities. For 2027, we expect a triple-digit revenue growth, supported by significant higher mass production allocation, increased contribution from existing customer, and the major win we have secured this quarter. Beginning of 2028, we expect to enter a new phase of growth as our next-generation accelerator project ramp into high-volume production. Conclusion. We have regained our growth momentum, are well-positioned to capture larger share of the rapidly growing ASIC market. The industry continues to move toward the COT model, a shift we have anticipated for many years. We believe it is a secular trend playing directly to Alchip's strengths, and we expect to secure more significant win in the near future.
With this momentum, we are confident in our ability to deliver a sustainable growth and create a long-term value for our shareholders.
Thank you very much. For the third quarter, quarterly income statement, it is pretty straightforward. The revenue for the third quarter is $819.3 million, which is 238.9% quarter-on-quarter growth and 267% year-on-year growth. The operating income for the third quarter is $113.9 million, which is 118.7% quarter-on-quarter growth and 194.1% year-on-year growth. For the third quarters, the net income is $102 million, translating into EPS of TWD 37.7 . This is the breakdown for the third quarter. You can see because of our accelerators shipment ramping up in the third quarter, our revenue contribution from HPC increased to 87% of our total revenue in the third quarter.
For the regional breakdown, again, because the ramping up shipment of our accelerator product to the North American customer, our revenue contribution from North American region increased to 84% in third quarter from 51% in second quarter. For the process technology, since the accelerator chip we are shipping currently is in the 3 nanometer process technology, the revenue contribution from 3 nanometer and the 2 nanometer in third quarter increased from 47% to 80% of our total revenue. For the third quarter business review, first of all, I understand that the capital market today is a pretty suffering day. But for our financials in third quarter, for the revenue and the profit, those are both record high for our company. The record-breaking quarterly revenue, the major reason behind is the ramping up shipment of our N3 accelerator product.
For the gross margin, I guess everybody noticed that the gross margin goes down to sub 20%. It is mainly because of the high percentage contribution from our production revenue, which means we have a little bit lower than expected NRE revenue contribution in third quarter.
But I want to emphasize, although the gross margin go down considerably, you can compare the gross margin in the first quarter, second quarter, and the third quarter. But because of our operating leverage, our operating margin go down mildly. We see when our sales scale keep on expanding, the leverage effect from our business model will getting higher and higher. The net profit and the EPS hits record high for our third quarter results. For the product itself, the N3 accelerator for now is shipping in full speed. The N3 accelerator this year will be the biggest revenue contributor to us and will be the biggest revenue contributor for us next year. For the next generation N2 accelerator project, this chip is close to taped out, and we are expecting to see revenue contribution from its mass production by the end of next year.
So, like Johnny mentioned earlier, because of the ongoing project and the new customer and the new project, we are expecting a very promising 2027 or even 2028 for our business and the revenue growth. Thank you. I guess we can get into the Q&A session now. For the first one, Charlie, please. You can unmute your microphone.
Hey, great. Thanks, Johnny and Daniel, and also congratulations for your big customer win. My first question is to clarify. Johnny, you mentioned triple-digit growth next year. Is that referring to your overall business or just the major customer's production revenue? Thank you.
I would say Johnny is referring to both. Actually, you may know that our total revenue kind of dictated by the N3 project this year and next year. So for next year, for the N3, for the revenue contribution from the N3 accelerator, we are expecting triple-digit growth. And of course, with the new customers and the revenue contribution from the new customer, we are expecting the overall revenue to grow even better than the pace of the N3 accelerator.
Oh, okay.
Charlie, I think Daniel says absolutely right. I think both.
Yeah
both business depends on which one is your viewing angle. I think on the current business, the growing momentum is very strong. I think its growth rate is also triple digit. Yeah, because of a new project coming up, I think our revenue, overall revenue, I think is in very good shape next year.
Okay. Thanks, Johnny. I think this is based on your first quarter, sort of the not the guidance, but kind of commentary. Your full year revenue probably end up with like $2.3 billion-$2.4 billion. If I predict that next year your total revenue can approach to $4.5 billion-$5 billion, do you think it is approximates estimates for next year's revenue?
I think that sorry, Charlie, we cannot talk about numbers in the earnings call, which is strictly prohibited by the TWSE. But I would say for this year, of course, $2.2 billion-$2.3 billion is your forecast. To us, we do not have a problem with it. But for next year, you mentioned the $4 billion, $5 billion. I think, based on our current outlook, it is achievable. I will not say I will guarantee you what is the revenue next year, but it is achievable.
Okay, understand. That is fair. Yeah, again, that is just our assumption and thanks for kind of providing some confidence and color. Second question is really about this second customer. Can you give us a small color? We know you have been working with some like big unicorn, some CSP customer, some networking customer, but none of those that we can sort of they can contribute like $1 billion this soon, within a year. Can you give us some background on this kind of second customer? Is that a CSP or AI lab? Is that a networking or AI accelerator type of project? Thank you.
Okay. So let me try to respond first. I think Daniel can add some color at the end. I think on the previous earning call, we already kind of implied, we are going to have another significant win. We have been told by this customer strictly cannot share anything. I think we have been cooking this for a long time, and I think right now, the reason we start to sharing because all the contract already signed, it is already done deal. But we have to follow the NDA. We cannot tell too much. But you can easily tell, within a year, has this kind of a revenue number, it got to be a very, very well-known customer. So I can say that I think the business is very promising and also sustainable. I think that will be a very significant win for us.
Daniel, you.
Okay. Actually, not too much thing from me. Just like Johnny mentioned, for the revenue contribution, the scale of the revenue contribution, this business must come from one of the big guys in the market. But please forgive me that we cannot disclose more information to you because of the strictly signed NDA. I think as long as the product in shipment, you will get a lot of the information from the industry.
Got it. So, when do we think start to see that revenue contribution from second customer? Which quarter? May we know?
I would say starting from somewhere in the second quarter next year.
Okay. Okay. How sustainable is this revenue? Meaning if you start the shipment second quarter next year, the annual run rates of this revenue could be like $1.5 billion-$2 billion in 2028. Is that the right way to think about the second customer's revenue contribution?
I would say it is a billion-dollar business. I am sorry that we cannot provide you more details. I want to emphasize this. This is a beginning of a creation of a new business model for us. We are very optimistic on this kind of business model. Like Johnny mentioned earlier, we consider this business model is a sustainable and with good profitability business for us. As for the first project, I think the main purpose, of course, we can make pretty juicy profit from it, but the main purpose for this one, kicking off in the second quarter next year, we consider it is the build up, it is a beginning of building up a business model with this customer.
Hmm. Okay. Yeah, last one, I will be back to the queue. Is this customer first time to become Alchip's customer, or was a customer for other application before? Is that a totally new customer to Alchip in the history?
Yeah, you can say that. It's pretty much a new customer for us.
Oh, okay. Congrats. I will be back to the queue and ask some other questions later. Thank you. Thanks, gentlemen.
Thank you. Okay, then Gokul, please, you can unmute your speaker.
Yeah. Hi. Thanks, Johnny and Daniel, and congrats on your securing the second customer as well. First of all, on the first customer, I just want to clarify, you're expecting triple-digit growth for the N3 project itself for the first customer, right?
Yes.
I just want to clarify that comment.
Yes.
Within that first customer, one of your competitors also reported a couple of days back, they are involved in some way, and they are also talking about a fairly big contribution. Could you talk a little bit about what is the kind of allocation business that you are seeing? Because it seems like you have secured more wafers and CoWoS capacity from TSMC, but still feels like this customer is engaging in the wafer buy from other parties as well. How do you see that evolve over the next couple of years, given TSMC is also becoming a little bit stricter in terms of who they allocate wafers to, when wafer supply is tight?
Okay. Gokul, for the N3 accelerator business, I will say this way. Actually, we have been talked to you guys many times. We keep on saying there will be upside. I understand, because you follow Alchip for quite a long time. Obviously, triple digit is a better number than our previous guidance to the capital market.
That is because the upside happens, and the relationship between Alchip and the customer is getting closer and closer. And the customer's dependency on us is getting higher because we provide very good value for the current N2 project. We are receiving more orders from the customer.
That is the, I think that is the main reason.
Understood. How do you look at this competitive landscape, and do you feel like it will become more straightforward in the next generation N2-based project? Because obviously, in this generation, there are quite a few people who are involved in placing orders at TSMC using the same mask, right? Do you think that kind of becomes a little bit more clear-cut? Probably one of the reasons why the market does not give as much credit to your business as it should as well.
Okay. Yeah, thanks for your input. I think you are, Gokul, you are right. For this N3 business, as you know, beginning of this business, we did not communicate with the customer into very detail. As you know, we are winning this project back from our competitor. Yeah, during that time, our priority is winning. As you know, the revenue size is very big. During that time, we may not have enough funding, operating funding, to support this project. That is why a couple months ago, we have a GDR. We just have a GDR. Now, we can say most of the constraint already addressed, and we are ready to take more volume, yeah, for this generation in next year. For the N2 project, that will be totally different story. I think our position will be even stronger.
That is why I say starting from 2028, that will be another phase of the growing momentum. But we see 2027 is a reasonable growing rate, but starting from 2028, that will be totally different story.
Understood. That is very encouraging. Just a follow-up to Charlie's question on the second customer. You did talk about this is a creation of a new business model. Could you talk a little bit about what exactly that means? Is this still an accelerator project with a lot of backend design involvement, or is it mostly a front-end involvement? What is the extent of Alchip's involvement in this project compared to, let's say, your current big customer? Is it higher involvement, lower involvement? Any more details you could provide?
Gokul, we cannot disclose into details about the project. I can say this way, that definitely our angle to tap into, to win the project has always been our engineering. As many people may know that currently the engineering shortage in the industry is everywhere.
It creates rooms for various type of cooperation between supplier and customers, or even between rivals or industry peers.
I would say because Alchip has been reputed for our engineering quite a long time, if there is any opportunity that we can work up together with a company by providing our engineering value and get reasonable profit and revenue, we will do so. That's the reasons for the new business.
Understood. We should think of this as working with another fabless company at this point in time, right? I think
I cannot. I am sorry. We really cannot disclose too much at current stage. Trust me, trust me. As long as we are close to the kickoff of the shipment, we will deliver more information to the outsiders. But for now, to better secure the outlook and the future of the cooperation between us and the customer
we intend to not talk too much about this customer and project.
Right.
Yeah, I think we cannot say too much. All we can say is our fundamental ASIC business, the business will be sustainable, yeah, with a reasonable margin. Yeah.
Understood. Maybe one last question from me, Johnny. I think a more broad-based, because some of your prospective customers have been using AI extensively to shorten the chip design cycle.
including front-end design, RTL, even netlist and other areas.
Even some verification-related stuff as well. Could you talk a little bit about how you see this evolving from your business perspective? I think Alchip's involvement primarily is in the backend side. But customers are trying to use more and more AI to shorten the curve and design more chips. How does it impact or benefit Alchip's business model?
Yeah, I think that's a very good question. Nowadays, I think everybody try to use AI, put the AI into current methodology, either from customer side or from our side. Yeah, like I mentioned before, right now, when you're doing the implementation, if it's a front-end software-related work, I think AI play a very key role. Single customer, when they write the coding, it's not from scratch. It's all generated from the AI, yeah, as a starting point. For us, yeah, because the process technology, yeah, very new, like N2, A14, we do not have too much reference, yeah, to train our model. So, we can do our own training. During the early stage, AI can help us to save some iteration, beginning to do the filtering. One thing AI can provide is analyze the report. As you know, after we running the tool, it does tons of report.
Before, we need a senior engineer to do the analysis. Right now, I think AI is doing a much better job to filter the variation, to provide a key direction for the next move. I think during the early stage, before we need a lot of resource, then most of resource can be saved. But unfortunately, at the final implementation stage, AI cannot help too much. We still need another resource. Right now, when we are doing a project, before we need about 150 people to do one complete project. Now, we can say 150 people is a peak time. We can save many resource during the intermediate stage. I think AI can gradually help. I think for this generation and compared with the previous generation, I think AI, we can see the improvement are quite significant.
The most sexy thing for the AI is once you train one model, you can easily crawl, and it is a zero turnover rate from the AI.
Okay, got it.
Thank you very much. Thanks. I will go back to the queue.
Thank you. Okay, next one is Haas. Haas, please.
Yes. Hi. Thanks, Johnny and Daniel, for sharing the detail and taking my questions. I guess to start with, just on the very top-down perspective, given the recent political environment change, there could be more and more data centers deployment slowing down. I was just wondering if you could provide some of the clarity regarding if that could actually affect the orders coming from your customers. If they cannot deploy their data centers timely, they may not need as much of the chipset inventory as they did in the past few years. Would you think that could actually be a concern for 2027, or most of the supply and also demand is pretty much locked in at this stage? Into 2028, would you be thinking that it is kind of a risk to monitor? Just from the business model perspective, would you be able to provide more clarity on it?
Thank you.
Okay, Haas. First of all, for your first question, it is kind of complicated question that I understand someone is questioning about the demand of the overall AI infrastructure build-up. From our angle, from our point of view, we don't see any weakness from customers' demand. Honestly speaking with you, I guess most of you know that for the N3 accelerator, it is an allocation business. We may not see the whole picture. However, based on the conversation between us and the high-level people of customer, the demand will remain very strong, no matter it is 2027 or 2028 or 2029. Probably you may heard that customer may slowing down his order placements. It may not come in from the demand side. It may come from supply side, because within the semiconductor supply chain currently, there are shortage everywhere. PCBs, substrates, testers, all those kind of things.
Is it possible for customer to slow down the orders? Maybe yes, because of the shortage. But for the chip side, as long as TSMC can provide the wafer capacity, the customer has no doubt to take all the wafer capacity they can get. That's the current situation. As for the next generation, based on the conversation with customer, the scale will be definitely much, much bigger than the current generation. That's why Johnny just mentioned, into 2028, that will be another jump for our overall business and the revenue scale.
That's super helpful. Just a follow-up question on that is that, given that kind of a restriction on the high power data center deployment because of the political environment change, would you think that could actually provide more opportunity for the ASIC design service companies like you? Or even for you might consider to do a little bit of the agentic AI CPU projects going forward, given there could actually be a meaningful mix shift of the demand away from GPU to ASIC and also CPUs in the next couple of years with the requirement of the data center deployment is getting more stringent that the power-related policy could actually be I think the power-related policy would also be more restricted. Just from your perspective, are you considering beyond inference ASICs or training ASICs?
Are you also working with your customers on the early engagement for the potential CPU projects at this stage?
Mm-hmm. Yeah. So, Haas, I think we cannot dictate the market. We're just a ASIC provider. Our role is similar to TSMC. Whatever the customer decide to do, we help them to do the implementation and help them to bring the chip to the market, yeah, in a short period. So, currently, I don't see any change direction on our customers' end. Yeah, the demand still very strong. Schedule is still very critical for our customer. Yeah, if they intend to do smaller design or CPU design, of course, we will help them out. But in terms of design complexity, CPU and GPU are totally different. I also mentioned before, GPU is now 2.5D, eventually it will be 3.5D, like a CoWoS-related design. Currently, CPUs, they are conventional 2D, without HBM. It's a DDR-related implementation. So the entry barrier for CPU design entry barrier is much easier.
Then most of a customer, CSP customer, is already doing the CPU design by themself. I do not think we can provide too much value. Also, CPU roadmap generation over generation is quite similar. Customer may not need too much help from the ASIC provider. GPU is a totally different story. The GPU already become chiplet for reticle design. So many different IP, so many different topology. Now it is a pure, the copper eventually will go to optical. I do not think any customer, the CSP provider, has a bandwidth and has a resource to do the GPU. But eventually, when CPU go to 3D, I think that also open the door for the company like us. So, so far, CPU-related design pretty much is COT, and the service provider may get a production-only business instead of a design.
But for me, I think we have to get in touch of a design in order to sustain the business. Otherwise, I do not think that will be too much value provided to our end customer.
Okay. My second question is just really regarding when you talk about the N2 project that you are engaged with your customers, set to be ramping production from late next year. You said the value add of the design content is actually more as compared to your N3 project. Would you be able to elaborate more detail on that? Because I believe that a lot of the IPs are coming from the third party. So just from which angle should we think about that your value add is going to be even higher as compared with your N3? Because there is just a lot of discussion ongoing regarding a potential accelerated shift of the COT business model in the next few years. So we would like to understand your perspective on that. Thank you.
Okay. Haas, first of all, the design complexity between the N3 and the N2 are significant. For N3, the current N3 project, the N3 is a SoC design, 2.5D packaging. But for the N2, it is still 2.5D, but it is a chiplet and the GAA process technology. We are responsible for every part of the chip's physical design and combine them together. I think, if you talk about content, you cannot only consider the big category like the compute, like the IP, or like the controller. You have to considering the difficulties of the design. The involvement during the production, I think we will play much, much more important role compared to the previous generation. You just mentioned about the third-party IP. I do not think this design has any IP dependency. There is no industry peer ever provide IP to this project. We are using the common solution.
Got it. Very helpful. Thank you so much, Johnny and Daniel.
Okay. Next one is Jerry. Jerry, please.
Thanks. Can you hear me?
Yes.
Yes, please.
Yeah. Thank you for taking my question. First, on the fourth quarter, I think you previously mentioned that revenue should continue to see some growth for the mass production business. At the same time, you are also going to have the tape-out for the N2 project. So, how should we think about the margin outlook for the coming quarter?
The margin outlook for the coming quarter, I would say the third quarter and the fourth quarter will be pretty similar. So, for now, I would say we are still shooting to achieve more than 20% blended gross margin target for this year.
Okay. But on operating level, should we see more leverage for the fourth quarter?
For the operating level, how to say that? It still depends on what is your gross margin, right?
Okay.
I will say in the long run, I cannot give you any particular quarter. I will say in the long run, we are shooting to have our operating margin around in the high teens.
Okay.
I will say this way, as long as the production jump up to another level in next year and beyond, with such revenue split, I guess you can expect the operating margin to be in high teens.
Okay, got it. Very clear. I think you, Johnny, mentioned that N3 next year will continue to ship and then another very strong growth on more allocation. I just want to understand, the growth for next year, is it the more wafer you receive, is it because of the customer giving you higher allocation, or is that the overall wafer allocated to this project that is driving your much stronger growth into 2027?
I would say the first reason is the main reason, since the capacity allocation decided actually like three to four months ago. Of course, along the way, there could be some extra capacity allocated to this project, but what you first said is the main reason. We get more allocation because of our value increase.
Okay. Then back to this. If you are getting more wafer allocation from your customer, I remember for N3, the business model probably, the margin, 20 wafer you receive or you can secure over a certain limit, you will need to probably negotiate with the customer on the margin side. How should we think about the margin for this project into 2027 compared with 2026?
I would suggest you to see it pretty much the same. Like we mentioned, nowadays, our position within the supply chain is getting better.
Okay.
I would say the N2, the design for the N2 project is kind of a very critical point for us to talk with the customer, because the customer do see our value and our effort to help them implement the design and to catch up with the schedule. Because you know that nowadays, the scheduling is very, very critical for our customers to compete. I would say the customer is highly appreciate our help, and it results in better term or conditions for our cooperation.
Okay. Got it. I think that is very helpful. For the N2 from next year, in terms of business model, can you help us to understand how it different from N3?
I will say, because the production is not there yet.
Yep.
At least, first of all, for the design, there is only one physical design supplier for the N2 project. It is Alchip Technologies. Secondly, for the mass production, at least at current stage, we have been promised by the customer, there will be only two party placing the orders for N2, the customer and Alchip.
This means that two parties placing the wafer order to TSMC for N2?
Yes.
Okay. Got it. That is very clear. Lastly, the question on the new customer side. Can you elaborate a little bit? I think you discussed a lot already, but I just want to-
Yes
I understand that for this customer, is this a single project for 2027, or how should we think? Or maybe we should think that this customer could have a handful of project, then there could be more to come in the next couple of years. This is a more like sustainable business model rather than every year, every project, different CSP. You need to compete at the next generation, you got to compete with some others again. Thank you.
Yeah. Like I mentioned before, I don't think that will be the single event. Yeah, we expect more project from this customer. Yeah, we just mentioned about that's just a beginning of new business model. I think this model, for me, is going to be sustainable. Yeah, we expect more and more project from this account.
Okay, got it. Thank you. Maybe a last question, if I may. I think previously mentioned, I think Johnny Shen and also Daniel mentioned about component shortages. Is it for the N3 project or it's for other customer? How should we think about this into the fourth quarter of 2027? When can this be resolved? Thank you.
Okay. Yeah, I think mainly happen to the other customer. The current N3 accelerator also suffer a bit on the supply side, but we are solving accordingly. I think the supplier, I think, very focused on the CSP business. That's their priority. Unfortunately, for other market segment, yeah, no matter you are in automotive, consumer, or any other area, I think it start to suffer. Yeah. Not on the wafer side, but on the other supplier side, like a package, the substrate.
Okay. Jerry, let me say this way. For the accelerator business, I guess you understand the industry practice in Taiwan. For those accelerator projects, they enjoy quite privilege from those suppliers. I would say the severe impact will be happened for other smaller project, because nowadays, since the supply chain is in shortage, they will choose to support the big project or big customer. For the accelerator, I won't say there is no impact, but the impact is really, really minor, because for those accelerator business, they have been always the priority for not us, but all the suppliers.
Mm-hmm. Yeah, at this moment, I think even the smaller customer really appreciate our service. Yeah, they know if they negotiate with the supplier directly, pretty much they will get nothing. Yeah, because Alchip still have a much higher bargaining power. Yeah, at least they will have something. Yeah, may not meet their original expectation, but they will have something to put on their system. Yeah, we continuously working with the supplier, yeah, try to help this smaller customer.
Okay. When can this get better or be resolved?
It gradually getting better, but right now, I think situation is, yeah, it's very serious. It's very tense.
Okay. Right. Okay, got it. Thank you.
Okay. Next one is Laura from Citi. Laura, please, you can unmute your speaker. Laura, you are-
Hello. Yes, hi. Can you hear me?
Oh, yes.
Yeah, thank you. Yeah, good afternoon. Thanks for taking my questions. Just wondering how we should think about the growth trajectories into next two years? As Johnny Shen mentioned that you already secured the second-largest customers, and that will supposedly to kick off in Q2 next year. Next year, we will see better momentums, more allocation of our existing N3 project. I just want to kind of to have a more understanding, how should we think about our growth into the next two years? Should we think about that the second-largest customers to start contribution sometime second half next year, or that would be more like 2028?
Okay. Laura, as I mentioned, first of all, we cannot mention numbers in the earnings call. That's prohibited by the authority. Secondly, for the growth trajectory, I would say this way. I guess you aware our previous guidance about the N3 accelerator business, the revenue contribution this year, right? For next year, we guided that the revenue contribution from N3 could enjoy triple-digit year-on-year growth. Yeah, that's for the N3 part. In the same time, we will have a new customer who will become our second-largest customer next year, which means the revenue contribution will happen starting from next year, and more precisely, starting in somewhere in the second quarter next year. So, with the growth outlook for the N3 accelerator and the potential contribution from the new customer, we consider our next year's revenue growth could be quite promising.
From 2027 to 2028, it is another story, because starting from 2028, we are expecting very considerable revenue contribution from N2 accelerator, N2 project. Like Johnny Shen mentioned earlier, 2027 to 2028 is a different level.
Yeah, sure. Thank you. Also, I recall that earlier we also mentioned about that we have a CPU and also the networking business. Can we get more color about the progress and how should we think about that non-AI accelerator business potential? We also have some bottleneck in terms of the supply, like ABF or other components could impact the overall momentum on those non-accelerator part?
Laura, I will provide you first with the numbers-related parts of the question. For the CPU and networking project, so far so good. Everything is on schedule. I would say this way, they are good customer and those projects are good project to us. Good project means they can provide us more than TWD 100 million revenue contribution. However, nowadays, not any given CPU or networking project can be the revenue contribution can be comparable with the accelerator. You can imagine that if the accelerator business can reach TWD 3 billion, TWD 4 billion, TWD 6 billion a year, even the good project to us, like TWD 100 million-TWD 200 million a quarter, it is not significant financially. We will continue to support those customers, but honestly speaking, because of the demand and supply status in the industry, we will be more selective.
Like I mentioned, it is also the reason, one of the reason that we can create the new business model with this new customer next year.
Sure. I mean, so seems like we got the quite good opportunities in the longer term, including that we start to engage big projects for maybe 2028. I am just wondering that for the next projects or next second-largest customers, do we have enough resource to support to make it another big jump into 2028?
Yeah. Laura, I think that is a very good question. Nowadays, I think if you are choosing a right project, the project, the revenue contribution could be easily double, triple, quadruple than any other project. So choosing the project wisely is our number one priority. The resource shortage is one of the bigger problem for us. Right now, the latest technology using too much resources. We already pretty much booked. Currently, I didn't mention on my opening statement, but we are doing hiring very aggressively recently. The headcount right now, I think till the end of this year, we have a chance to hit the four digit. Then we can take much, much more project. Right now, I also glad you are asking a lot of other project question.
As I mentioned in the opening statement, other than accelerator business, all other projects still contribute more than TWD 100 million per quarter. Some of them, I think, are due to the substrate or packaging shortage. Otherwise, we can have another TWD 100 million. We can easily get another TWD 100 million. For sure, when we are taking the emerging account, we will be more selective. Some of them, we still consider they have a very good potential to hit a billion-dollar level. The resource constraint, we are solving accordingly, then we will choose the project wisely.
Sure. Thank you very much. Very helpful.
Mm-hmm. Okay, next one. Charlie, please.
Thanks for taking my follow-up questions. I have a few. First of all, I think last quarter, we sort of discussed your potential role in the Google TPU. I think your competitor was very open to talk about that potential opportunity. May we know there is an additional kind of opportunity to your second customer, and what could be your role in the future Google TPU? Thank you.
Okay. I do not want to refer to some specific project. I would say Google has always been the target for us to win. At current stage, I would say this way, we are very, very optimistic and have very high confidence to win this custom. That is what I can say for now.
Okay. Is that part of the design or you will be the kind of coordinator similar to MediaTek's role in TPU v8 and v9?
Charlie, you are assuming we won already. I will say this way.
I hope so.
This is model for this customer. For customers like this, we will try every possible type of cooperation with this customer, but I cannot identify what it would be now for you. I am sorry for that. I think at this moment, unless, just like echo Laura's question, we have to use our resource wisely. If any design activity cannot generate any revenue, we wouldn't take. Also, like I mentioned on the previous earnings call, I think sooner or later, winning a Google, you won't feel any special. Everybody will have a chance to serve Google, because Google, I think, has so many different project, different skill. I think there's only a handful of service provider for them to pick.
Okay. Thank you. It seems like your so-called the second customer is not Google. Back to this so-called second customer. I think follow up.
No.
Oh, okay. Okay. Maybe that would be my own inference. But back to Gokul's question, right? You do not seem to rule out your sort of collaboration, even partnership with your sort of previous rivals. May we know, are you working with like a foreign partner or a Taiwanese partner for this second customer project?
A connection, and there's-
Oh
no other party involved. Just us and customer.
Oh, okay. Okay. I was thinking how to guess who they would be. So it is not like a startup company you kind of pursue previously?
No, no, no. It's well-known, the company. Everybody know that very well.
Okay. Is that their first generation ASIC or like second, third, fourth generation of ASIC?
Forgive us, we really cannot disclose too much. I guess the customer probably is listening to our earnings call.
Okay. Okay.
That allow us to be a little bit conservative at this moment. Like I mentioned, as long as the project is in shipping, we can talk a little bit more about the project, and I believe, on that time, there will be a lot of information from the industry.
Okay. Thank you. My last one is a bit, try to push it a bit hard that it seems like there could be also new opportunity from Google, as you just mentioned. Can I assume that after doubling your revenue in 2027, you can double your revenue again in 2028?
Yeah, for 2028,
Or more
we have a much, much higher expectation on that year. It's not like Charlie, let me say this way. In 2028, based on the current information we received
I would say growth is not Revenue growth is not an issue anymore. The issue is what is our balance sheet?
That can-
Okay.
Yeah, how much the revenue our balance sheet can take.
Oh, okay. I would take it as I get doubling again in 2028 is not impossible. You are aiming even higher growth, but balance sheet is a good problem to have.
Yes.
You said that. Okay. Okay, thank you. Yeah, I am not sure if you finalized your second business term with the big customer on the fourth generation. I am not sure if Jerry asked that question, but if not, maybe get some more color about your negotiation with big customer on the next generation. What will be the kind of the change?
Yeah, it's not done yet, but we can say so far so good. It's much better than previous one, but it's not completely done yet. Yeah, the design is not taped out yet. Right.
Okay. Remember for the current generation, there's been some very annoying competition. Some your competitors don't do the design, but they can get the production rights for their project. I think one of the analysts also asked TSMC management during their earning call that, would that be the case in the future? What will be your observation for not just the fourth generation project, but also your other project? Do we need to concern again that your industry peers to get a production rights if they can get a TSMC wafer allocation?
First of all, Charlie, I will say this way. TSMC's C.C. Wei also mentioned in his earnings call, he doesn't like the secondary wafer market. Which is a very good news to us, because we always believe by contributing our design capability, we suppose and we deserve to enjoy the benefits from the production. So if the secondary market can be eased to a certain degree, of course, definitely it is a good news to us. Like I mentioned earlier, at least for now, the customer said there will be only two entities placing orders to TSMC for N2. They are customer itself and Alchip Technologies. So, that's why we are here saying 2028 is another level. First of all, we do consider the overall scale for the N2 generation will be much, much bigger than N3.
Secondly, we do consider the supply, which means the wafer placing situation for N2, from our perspective, will be much better than the N3 project.
Mm-hmm. Okay. Yeah. Those are great to know. That will be all from my part. Thanks so much for your answers. Thank you.
Thank you. Gokul, please. And Haas.
Yeah, hi.
Sorry for a moment. Haas, you are the last one to ask question because we have constraint for the time. Gokul, please.
Sounds good. Thank you. Thanks for taking my follow-up. First of all, on the N2 project, based on what you see right now, could you help us understand, are there other companies involved in the I/O die? Or, like we discussed last quarter, even the I/O die involvement is largely only Alchip?
No. Yeah, I think-
Because there seems to be a lot of confusion on different I/O combinations.
No, at this moment, I think, no. I don't see any other involvement. It's only Alchip. No matter it's I/O die or other die, because it's all done by Alchip. Gokul-
Okay. So, yeah.
Okay. Gokul, let me say this. For now, you can imagine the schedule is very tight for this project.
For now, engineers from both company, their priority, their goal is to tape out the chip on time, tape out the first chip, and then to think about, will there be other version of the chip by changing the I/O die? Like we discussed before, the customer mentioned the chip will support UALink and NVLink Fusion. But no matter how, you have to get the first chip
running first, right? That's the common goal for both company right now. After that, there could be other I/O dies. But we can say, first of all, we will still provide our physical design. And secondly, to do the I/O die, it requires the connection with the compute. So we will know, and we will be the candidates
for other I/O die, if there are. Gokul, if you recall, and I think, last year and couple years ago, a lot of people say they doing the design involvement for the current generations, N3, but at the end, people already know Alchip is the only vendor to do the design. It is the only one mask doing the production. Sooner or later, you will find out N2 is the same story.
Understood. Thank you very much. Maybe one last question on the second customer. You mentioned that this is going to be directly dealing with the end customer itself, right? So there is no intermediary or any kind of vendor here. Could I confirm the IP also is still using the same model that you have, like third-party IP, or are you licensing IP from some other existing kind of fabless player or something like that?
Gokul, we cannot disclose that. Once we disclose that, you will soon guessing out what is the company. So, I would say, you guys are too good for this. I will say this way. It is quite simple to us that we are always selling our core competitiveness, which is our engineering. You can consider it-
Yes
this way. In the industry, there might be a lot of projects require physical design support, but cannot get it from those companies who are capable, but reluctant to provide the engineering resource to you, to those projects. And we can become a very good candidate as long as the project brings pretty good company revenue. So that is the-
Okay. So you-
whole origination of this kind of business model.
So we could say that this is like a co-owned kind of project, right? Not really like fully owned by one party or the other.
For ASIC business model is unlike the other merchant product. I think us and customer co-own the project, always the case.
We cannot use the mask to serve other customer. We can only produce the project for the customer who gives us the service, who gives us the business. You can say it's a co-own model. We can never sell their product.
Yeah
to any other customer. Yeah.
Sure. Okay. Yeah. Thank you very much.
Thank you. No problem. Haas, you are the last one.
Yeah. Thank you. Sure. Thank you so much for taking my questions. Just two quick follow-ups. I think first one is still on second customer. Would you be able to just provide more clarity the type of the customer, whether it is like CSPs or like frontier-
No
AI labs?
No, no. I can just tell you no.
Okay. Yeah.
we try our best to provide information to you guys, but please forgive us that at current stage, it's not proper for us to deliver more information to you. It may endanger the project.
Oh, okay.
The business is-
Or even-
Yeah.
Yeah. Just the region of the customer, would you be able to provide it, whether it is like U.S. or Chinese or, yeah, anything will be helpful.
Yeah, it's not Chinese customer. It's a U.S. customer, you can say.
Okay. Just for the purpose of the chiplet, we could assume it is also the AI accelerator ASIC. Will it be mainly for training or inference tasks?
Yeah. Sorry, we cannot provide it. Yeah.
Okay.
Yeah, I think nowadays, has such kind of volume and revenue, it has to be-
It has to be AI.
Yeah.
Hmm. Okay. Well, yeah, that's very helpful. Second question is just on your long-term operating margins target close to 20%. In third quarter, you reported gross margin 18%, operating margins 13%, 14%. Would you be able to just elaborate what is the time horizon that we should think about that your operating margins could improve back to 20%? Second thing is that, is it related to your second customer that it could actually be margins accretive, so that gives you the confidence that in the long term, your operating margins could actually be close to 20%?
Okay. Haas, I would say, first of all, it's an easy calculation. For example, if you assuming the revenue achieving like TWD 4 billion, which means 20% gross margin, which means if we can spend less than TWD 200 million for operating expense, the operating margin is 15%, right?
That's right.
Yes. That is our target because we are seeing that the revenue scale can reach or even be higher than this number in next couple of years. Our operating expense, which if you check the history, we have been pretty conservative and responsible. I do not see when you said that the high teens goal operating margin is unachievable.
Okay, got it. Then just shifting gear a little bit back to like second customer, I think just one more follow-up question on this. When you said second quarter next year, the revenue contribution will be kicking off from the customer, is it more from the NRE angle or it is from the mass production perspective?
No, production revenue.
Okay. So we can assume the chiplet is already in the pipeline of being manufactured in the supply chain?
We have been took this project for a while.
Okay.
That is the information we can provide to you. Starting from somewhere second quarter next year, the contribution will be linear. It is not a one time in second quarter. The revenue will continue starting from the second quarter to third quarter, fourth quarter.
Got it. Super helpful. Thanks so much, Johnny and Daniel.
Thank you. Okay. I guess it's conclude our earnings call today. Thank you so much for your participation. I'm sorry, probably because somebody already tell me that-
Technical problem
technical problem when you trying to clicking the link. Sorry for that. Thank you again.
All right. Thank you very much.
Thank you very much.