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

Okay. Thank you for you for waiting for our First Quarter '26 Earnings Call. We are subject to start the meeting at 2:30 p.m. It's about 2 minutes later. So just be patient for 2 minutes. We will start the meeting accordingly. Thank you. Okay. It's time. Thank you for waiting. Dear analysts, portfolio managers and all the participants, we welcome to join our first quarter '26 earnings call. And I'm Daniel, CFO of Alchip Technologies, and our CEO, Johnny Shen, will also be hosting this conference call. And thank you for your patience. We will start. Okay. First page is, again, safe harbor disclaimer. And for this meeting, we will conduct this earnings call in Chinese. But me and Johnny, we are -- English, and Johnny and I are Chinese speakers. So if you want to ask questions in Chinese, please feel free to do so. And the Chinese version of presentation slides, I believe right now is on the MOPS already. So you can download it from MOPS. For the Q&A session, you can write down your questions, through Teams message board, and if you want to ask question orally, you can use the raise hand function and we will call your name for your question. This video and audio content of the meeting will upload to MOPS probably after 2 hours after -- 2 hours after the meeting. So probably sometime because the file is too big, you have to wait for a little bit longer. So this session is for the message from our CEO, Johnny Shen. Johnny, please?

All right. Good afternoon, ladies and gentlemen. I'm Johnny Shen, Chairman and CEO of Alchip Technologies. Once again, thank you for joining our investor conference meeting today. We truly appreciate the opportunity to share our Q1 financial results and provide an update for our business outlook. For Q1 summary, our first quarter revenue came in slightly below plan. Revenue was lower than last quarter due to the limited production revenue, while our NRE remained very strong. As a result, total revenue reached USD 132 million, with net income of USD 45.1 million and EPS TWD 17.55. A detailed financial breakdown and analysis will be presented by Daniel in a later section. There are a few highlights worth mentioning in Q1. First, our gross margin reached 50%, driven by strong NRE demand and higher production margin. This demonstrates even during the production transition period, company can remain highly profitable through strong NRE contribution. Second, our N3 design for #1 customer is now ready for production. The shipment expected to begin in June time frame. And we successfully secured required capacity across all suppliers, including wafer including wafer, CoWoS partnering, substrate, cooling, tester, et cetera, to support on time delivery. At the same time, the next generation design also progressing smoothly and remain on track to tape out by this year. Third, our automotive business enter volume production last quarter. The volume expect to ramp in Q2 and remain stable through the end of the year. We expect automotive project become our #1 revenue contribution in the first half, and our second largest contributor for the whole year. In parallel, the next-generation automotive chips currently under design remain on track to tape out by Q3 this year. Quick update for geopolitical risk management. We continuously to diversify our business and design resource beyond China to mitigate geopolitical risk. In China, we are mainly focused automotive, robotics and consumer-related products using N3 technology and above. In terms of engineering resource, we have launched a very aggressive hiring plan focused in the region outside China. Now our Japan office can accommodate up to 250 employees, while our Malaysia and Vietnam office can host more than 120 engineers combined. As a result, the majority of our design resources are now located outside China. As for the future business outlook, starting from Q2, particularly from the June time frame, we expect to regain momentum in production business. In terms of the revenue trend, we expect Q2 revenue to be higher than Q1, followed by a much stronger ramp in Q3 and Q4, both of which we expect it to be significantly higher than Q2. We expect this revenue momentum will continue through the remainder of this year and extended into late next year, followed by the mass production of the next-generation chip. In conclusion, we truly believe the most challenging period for the company has already passed. We have confidence to perform a significant revenue growth even compared to our peak year in 2024. Looking ahead, we remain very optimistic about long-term outlook for AI market. For this year through 2029, we expect our growth momentum to remain in line with the latest industry leaders and competitors. We are also confident in our ability to outperform overall market CAGR in this high-growth HPC and AI market. Thank you.

Okay. Thank you, Johnny. And for next page, it's a very straightforward quarterly income statement of Alchip. So in the first quarter this year, the revenue came in at $132.4 million, which is a 13.3% quarter-on-quarter decline and a 58.5% year-on-year decline. For the operating income part, the operating income for the first quarter reached $43.2 million, which is a 10.5% quarter-on-quarter growth and a 4.8% year-on-year decline. With the additions of the non-operating income and the deduction of income tax, our net income in the first quarter reached USD 45.1 million, which is a 5.7% quarter-on-quarter decline and 1.6% year-on-year growth. And the EPS for the first quarter is TWD 17.55. And the next page is for the application breakdown. As always, our #1 revenue contributor in terms of the application is HPC. HPC in the first quarter accounted for 70% of our total revenue, and we are expecting these numbers to trend up in the following quarters this year. For this page, this is the process node mix. As you may see that the N3, N2, N5 and N7 combined revenue contribution accounts for more than 80% of our total revenue. I still believe that based on this product mix in terms of the process node technology, our chip is still the leader within the digital design industry. And for this page is the geographic breakdown. For the first quarter, it's a little bit different from the quarters last year. The North America used to be our #1 market. In the first quarter, the contribution from this region reduced to 23%. This is because we don't have too significant production revenue contribution from this region in the first quarter. And for the Asia Pacific, 47% of the revenue goes to this region. That is because of our automotive business. In the following quarters, I expect the breakdown of the -- I expect the regional breakdown will change quite significantly once the N3 accelerator production revenue kicks in. So for the first quarter business review, like Johnny mentioned, the revenue came slow, a little bit below our previous guidance and the plan. However, the profitability for the first quarter is quite promising. Like I just mentioned for the numbers, the revenue for the first quarter this year declined 58% year-on-year. However, we record year-on-year growth profitability. The first quarter margin surprises on the upside because of the high percentage of our NRE revenue. And we believe this kind of mix will slightly -- the high percentage of NRE will slightly go down in the second quarter as we are shooting to kick off the N3 accelerator shipment, and keep on going down in third quarter and the fourth quarter. So the gross margin trend for this year will be trending down because of the significant increase in revenue in the following quarters. For the NRE pipeline, we see the NRE pipeline remain very, very strong, especially for the North American market. The design demand from North American region is very strong, especially for the AI-related sectors such as the accelerator, CPU, networking projects due to the massive CapEx by those hyperscalers in the industry. And we do see the process node technology for AI-related chips are moving from N5/N3 to N3/N2. We are expecting this year, the majority of our AI-related project will be in the N3 technology nodes. And we do see multiple ASIC design opportunities from North American hyperscalers. Okay. I guess that's conclude the first quarter. And for the outlook this year, as we mentioned many times, we expect the growth momentum to pick up starting from the second quarter '26 and actually, the production momentum to start in late May or early June. And we expect a very strong quarter-on-quarter revenue and profit growth starting from third quarter this year and going forward. The main reason is the N3 AI accelerator shipment to our North American customer. And for the revenue distribution, we expect for this year, it is a very back-end loaded distribution for our revenue distribution. Again, that's because of the N3 AI chip shipment starting in June. For the NRE, the NRE revenue will be growing strongly. First of all, the overall demand for design is still good. And secondly, as Johnny mentioned, our N2 accelerator project is going smoothly, and we are shooting to tape out this project by the end of this year. And the related revenue will be another reason for our strong NRE revenue performance this year. And this concludes the outlook for this year, and we are entering the Q&A session. Thank you.

[Operator Instructions] Gokul, please.

Great results. So for the N3 accelerator projects, it definitely looks like the customer demand seems to be much higher now compared to what people were thinking maybe 3, 4 months back, given they have signed several gigawatt deals. How does Alchip benefit from this? Should we assume that you will benefit proportionately from this? Any updated size of the project during its lifetime given this demand upside? And lastly, any thoughts on your 3-nanometer and CoWoS wafer allocation given this upside in demand, given that we are hearing many different vendors are trying to enter into the wafer sourcing or the customer is trying to kind of find various sources for 3-nanometer wafer sourcing from various kind of fabless companies.

Okay. Gokul, I will say for the -- obviously, the customers' demand is really, really strong. For now, for the real revenue contribution, I think there are 2 factors are very important. First of all, the wafer capacity -- how many wafers, extra wafers we can get or the customer can get is a very important factor to the upside on our plan. And secondly, what is the execution of the production. As you may know that the yield rate is quite critical to this kind of accelerators. As long as we can execute the manufacturing smoothly, there could be upside from the yield improvement. So definitely, the increased demand from our customers or the increased demand on our customers' customer is definitely a good thing to us. As for the various source, looking for various different vendors or sources by customer, I can assure you that for the design part, there is only one physical design vendor for the N3 and the future N2 projects. It is Alchip Technologies.

Let me also add some comment on this. Yes, I think it's not a secret. It's not the secret news. I think our customer really received the higher demand. But to be honestly, now it's already May time frame. If you talk about the upside for this year, I think it will be very limited because wafer turnaround time need about, including CoWoS need about 6 months plus. So even the upside will happen, but for sure, it's not going to be this year. For next year, yes, we are working with the customer very closely, try to get the upside. I think from the wafer, we all know N3 capacity is the most precious thing in the industry. So very difficult to get an additional one. Yes, of course, over the year, foundry will have some upside because of yield improvement, those kind of stuff, we try to capture those kind of gain. Again, we are working with the customer very closely to try to maximize the upside next year.

Got it. Understood. And on the N2 project, the follow-on generation, as you mentioned, you are kind of going to complete the tape-out by end of this year, so probably revenue sometime end of next year, very likely. Could we have some understanding of like the value capture for Alchip given this is a much more complicated design, is it like multiple times like what we are seeing in some of the other ASIC projects out there in the market, like let's say, the 2-nanometer version versus 3-nanometer version, your value capture could be multiple times of the 3-nanometer version. And secondly, there seems to be so many different interconnect technologies or interconnect solutions that are being used, I think NVLink, UALink and maybe some proprietary stuff as well. So what is -- how does Alchip get involved in each of these areas? Or is that not something that you're kind of involved in because you are largely focusing on the compute die and the overall back-end integration?

Okay. Gokul, for your question. For the N2 project, I would say for the NRE total contract value, definitely the N2 project is much, much higher than N3. And the same thing for the chip price. The chip price, as you already mentioned, it is a chipset and there will be different -- there will be multiple part of the chip, the compute die, the I/O die and the others. So the chip price will be much, much higher than the N3 chip. And for the so-called what is our role for the I/O die or the others, I would say we will provide the physical design for the compute and the integration of the whole chip. So for the I/O die part, we will also involve in part of the physical design as well.

Got it. Got it. Understood. So compute die and overall design is still -- you are the only partner?

Yes.

Understood. Understood. Okay. And maybe last question from me is on any -- I think previously, you have talked about CPU is probably not something that you want to kind of engage too much in given it's a lower margin, lower value-add project. Any change in thinking on that direction given CPU demand is now growing much faster than what anybody would have thought with the rise of Agentic AI? And any thoughts on any potential kind of CPU-related projects that you're working on with some of the hyperscaler customers?

Okay. First of all, I have to do some correction. We are not saying that we don't like CPU application. We prefer -- we will be very cautious about taking production-only project. It is not because of it is a CPU. It is because of it is a production-only project because we consider for the production-only project, the value added from our is very limited, and the gross margin for this kind of project is relatively low to those so-called PD1 or PD2 project, which we provide the design value to the chip. So in the future, if there are CPU opportunities, definitely, we will compete.

Yes. As you know, our company has a long history to do this kind of arm-based server CPU. I think we still have a few customers there. But just like Daniel mentioned, if any particular customer come to us only for wafer capacity only for production service only business, we will put the upside. We need to provide the value in order to sustain the business. But nowadays, the situation is very -- changing very dynamic, the CPU design also getting more complicated. If the customer needs some design resource, need our support, we definitely will take it. But obviously, the CPU design compared to GPU is relatively easier, because in terms of size, all the GPU is radical now and CPU is much smaller. Nowadays, most of the companies go for the COT trend. The first project, most of them start to do with the CPU. They have a certain volume and entry barrier is relatively lower. But again, there are a few opportunity we are talking on right now. Just like Daniel mentioned, we are not against to take any CPU business, still very good business.

Okay And then Haas, please.

Daniel, congrats on the great results. I guess 2 questions from me. First one is probably just regarding the upcoming 3-nanometer AI accelerator ramp. How should we think about the linearity throughout second half this year? When should we expect it is going to hit the peak, at least the near-term peak in second half of this year? And related, I think you mentioned there's going to be some upside from the volume perspective. But I was just wondering if the COT business model is going to constrain your sales and also margins upside, even if there is additional volume demand from your customers?

Okay. The ramping up speed for the N3 project will be very fast. I would say you will feel the revenue to jump in June, and there will be another jump in the second -- in the third quarter. For the monthly, we expect the monthly revenue to reach the peak starting from the middle of third quarter. And as for your question for the upside, I would say that Johnny mentioned, for this year, the upside from extra orders from customers is a little bit limited because the turnaround time for this project is very long, 6 to 7 months. So you can imagine if we place the wafer orders today, the chip will come out in December. So for this year, I would say it's pretty much that from the orders perspective. And like I mentioned, there is still another factor to the revenue. It is the yield rate. So we will keep on improving the yield rate of the project and to see how much is the upside we can get for this project this year.

Your question regarding about when volumes start to increase, are we going to lose part of production margin or not? I think -- to be honestly, I think that's a problem most of people, major supplier competitor also facing the same situation when the project used to be $1 billion, now it's a multibillion $20 billion to $30 billion, can they maintain the same margin? The answer is very difficult. But for us, we all know that we are a very reasonable -- we charge customers a very reasonable price after the volume increase for multiple generations. And we are able to maintain the same profit margin. But I think it's going to be quite challenging for our competitor. I also consider that's a benefit for us. If the volume increase so much, sooner or later, people will go for the COT direction and try to find a more appropriate partners.

Got it. So based on your comment, would you be able to remind us about the contribution from this project this year and also next year? And it doesn't seem to be there's going to be like a ceiling even if the volume hit to a certain level next year, even if your customers' demand is stronger than expected, that you can still recognize a lot of revenue and also profits beyond the original scope that you signed or you have an agreement with your customers?

Yes, definitely, we are still working on extra orders from customers next year. And we do believe, based on the current situation, we will still enjoy pretty strong growth for the revenue contribution from this project.

Okay. Got it. And then my second question is just regarding the revenue or profit sharing for the upcoming 2-nanometer project. I guess since it is going to be on the chiplet structure, are you able to capture the whole content when you put everything together or the other IC design service partners doing some part of the design on different tiles will be able to share the revenue once the chipset enters into mass production. Yes. Just wondering how the business model is going to be working for the upcoming projects on 2-nanometer?

Yes, we shouldn't comment too much about customers' architect. But we honestly in the future, maybe starting from next generation, maybe it will be more provider, more partner getting involved. But current production, I think it's straightforward. We are pretty much handle everything, except the HBM, I think is consigned. The rest of them we are handling just like before, even go to the chiplet stage. I didn't see any additional partner get involved for the design we are doing now.

Sure, sure. And then I think just regarding your prepared remarks that you mentioned that you have visibility through 2029. Would you be able to share some of your financial target regarding what is -- what kind of the sales CAGR outlook through 2029, we think we should be thinking about for your company in the next few years and also your expectation for the overall addressable market in the next few years?

No, we are sorry, we cannot provide it. Yes. And first of all, it's the earnings call, it's a quarterly earnings call. We suppose not to give the growth guidance to -- in this meeting. And the reason why we are so confident for next 4 years is because, first of all, we secured the 3-nano -- not 3-nano, the N3 and N2 project. And the N2 project definitely will bring the revenue contribution to another scale than the N3. So that's the reason why we guided outsiders saying we are really, really confident for our growth in the next 4 years. As for the numbers, I'm afraid we cannot provide further information to you.

Got it. What about just the market forecast, if you have anything in your mind you can share through 2029 from the CAGR perspective?

No, I guess that's your job.

Okay. Yes.

And Charlie, please, Morgan Stanley.

Also congratulations for a very strong quarterly results. So maybe not to get your so-called numeric guidance, right, but your major customer announced a 5 gigawatts with Anthropic just recently. So I'm not sure if you see so-called upward revision from the key customer in the coming years for your order. Or if you cannot disclose your revenue CAGR, I think your industry peer MediaTek shared their view about the TAM. It's industry information, right? So are you okay to comment on both of those. First of all, do you see upward revision recently? And secondly, if you have any view about AI ASIC TAM, no matter for 2028 or 2029, it would be great.

Okay. First of all, for the TAM numbers, I would say actually for this number, we -- our information is most likely from you. And to me that I think MediaTek guided $100 billion next year for the TAM. I think, first of all, it depends on the definition of the AI chip because some include those networking, some are not. But to us, I would say $100 billion for the market TAM is not that -- I would say it's pretty accurate because we see the number going up every day, every month. What we know is the demand -- the end demand to our customers and our customers' demand to us keeps on increasing, for the past at least 1 to 2 years. There is always upside request for how many wafers we can get from TSMC.

Okay. Johnny, do you have anything to add? And also based on the TAM, any -- your target, say for market share?

Right. Personally, I mentioned so many times, I'm very optimistic for this market. I'm a true AI believer. And yes, the market TAM, I think on TSMC event, right, they also mentioned about by 2030, people thinking about $1 trillion. But this time, TSMC symposium in the U.S., they revised to $1.5 trillion overall TAM size. And I think the ASIC among these TAM will play a significant role. But for the past 3 years, everybody think about ASIC. But in reality, Google, Amazon contribute majority, more than 90% of the ASIC revenue in that area. So through that, I think we have a confidence to overachieve the market CAGR. And in fact, we keep receiving some pressure and also additional demand from our end customer for 2 things: maximize the capacity, and we are using all kind of channel and influence to try to get more wafer as much as possible. In addition, we need to tape-out on time for the next generation. I think that's -- we can see customers is so aggressive. I think we should be able to capture this wave in the near future.

Great, Johnny. So since you mentioned about the tech symposium, I think on your website, you also have a press release. And I think you talked about you highlighted 2-nanometer ecosystem readiness, 3DIC integration and also advanced packaging leadership. But there were some participants also told us that you showcased your design with Ayar Labs, which is optical I/O die using TSMC COUPE technology, right? So may I know that it's going to be used for your 2-nanometer project already or the CPO or optical I/O dies for the next generation?

Yes, Charlie, I'm afraid that we cannot answer your question. It is related to our customers' design in the future.

I can also mention about -- I think on the previous few earnings call, I emphasized our ecosystem. The reason Alchip is so different compared to most of our competitors because we are completely neutral. We didn't make any IP, we didn't make any products. So in the other world, everybody, whatever the solution provider is willing to work with us very closely since we already have a track record, and we are neutral. So you will see more and more this kind of cooperation or even press release in the future. So I cannot -- it's not appropriate to make which one is going to hit volume. Are we going to have any production design soon? Yes, it's not appropriate to comment. But overall, I think you will see more and more well-known company is willing to work with us very closely.

Sure, sure. Yes, I think that's totally reasonable. I think it was great to see you can showcase your technology capability with partner. So last one, I will be back to the queue. I'm wondering about the competition, right? I think some is well discussed. I think I can call a name, for example, Meta, 2-nanometer projects. And there's also another one. I think it's a U.S. automotive for space or robotic customer, right? I think that is also very important for your foundry partner. So can you comment a little bit about those 2 project win chance? And also, how do you compare yourself to those short list, right, meaning GUC, MediaTek, even Qualcomm for those project competition?

Okay. Yes, I think -- yes, it's true, the competition is getting more intense. But if you think about the whole picture, the design opportunity is getting more and more. And before each of the hyperscaler only have one solution. Now they are thinking about the primary alternative or second source. Eventually, each of design will have a huge volume. They are going to have multiple source. I think this trend is unavoidable. But if you look at all the competition, maybe just 5 fingers. I don't think there will be any newcomer or the hyperscalers barely to use any newcomers. So I think at the end, we are only competing with maybe 4 of them. Of course, each of them are very respectful competitor. The size are even much bigger than us. But I think we have our niche. We have our own DNA If you talk about the pure ASIC provider, I consider we are still #1. A lot of newcomers, they are doing product in parallel with doing ASIC. By natural, there's -- first of all, there will be some conflict interest out there, plus there will be a different DNA because doing the service and doing the product are totally different, totally different. So when the market opportunity getting more and more, and if we still -- if we can deliver the design on time with the quality like we used to do, I think we can -- for sure, we can capture more projects. But I don't intend to win in all. But for sure, we are going to be continuously to be a very important player in this industry.

Okay. Next one, Jeffrey, Macquarie, please. Jeff?

Can you hear me okay?

Yes.

Yes.

Okay. Great. I guess maybe to follow up on Charlie's question. Over the past couple of quarters, you really said you're focused mainly on your current customers, current generation that you're ramping up now and next generation one. And then this time, you also add that multiple ASIC design opportunities for North American CSPs. So I guess, are you a lot more confident on getting other major orders in the next year or so now versus a few months ago? Anything has changed there?

Okay. As always, we really don't want to release something like, oh, we have 70% confidence to win or 60% confidence to win because NRE winning everything, lose is nothing. I will say we do have opportunity to compete for multiple projects from those North American hyperscalers. As long as we have good results, we will find a proper time to deliver the message to the investors.

Okay.

You're right. For the past years, our current customer will be our primary focus. We need to make sure that design tape-out on time and winning their next generation. And -- but in that way, in addition to this customer, we do have so many wins in the North American region for the emerging account. So if you count number of tape-out and number of design win, compare Alchip with any vendor in the industry, I can tell we are one of the best. In terms of number of tape-out, in terms of number of project winning, I think we will be one of the #1 for sure. But if you talk about the hyperscaler significant winning, I have to admit we have a chance to win. We are in the progress to win another. But right now, I think our current customer, I think, play the most important role. We don't have other hyperscaler -- major hyperscaler in our portfolio yet -- I'm not talking about the consumer for the AI, HPC accelerator, we are -- right now, we only have one important account. And fortunately, this is one of the most important account in the industry.

Great. Understood. And one last quick question. You talked about the top 1 and top 2 customers this year. Who do you think emerges as top 3 customer by next year?

Top 3 customers, I think, is very difficult to estimate. Maybe I think the top 1, #1 will remain #1 for a while. Whether we have another account can overachieve our #2, I think it's also possible.

And if that happens like the networking company or something else, you say, within the next year?

Yes, a few significant -- few good account we already won. We hope they can hit volume starting from next year.

Okay. Next one is Gokul, please. I'm sorry, Laura, please. Laura, please.

I just have a quick ones. We know that Alchip has a very close relationship with TSMC and also work very closely on the CoWoS ecosystem. But we also noted that various different customers may also looking for other alternatives on the OSAT or a different approach on the advanced packaging. So just wondering that from your experience or your expertise on the back end and also the packaging side, will that be any kind of a challenging when you're working with other non-TSMC's packaging partner? Would that have any impact on your project progress going forward?

No. Actually, first of all, for now, the wafer is the bottleneck, not the packaging, not the CoWoS. And secondly, I would say, for the N3 project, TSMC support the capacity pretty well. TSMC has a very high commitment to this project. So no matter for the wafer or for the CoWoS capacity, we were -- we are satisfied with the support by TSMC. However, if there is other vendors who can provide the so-called 2.5D like packaging, we don't have a problem with them, okay? So our ground rule is straightforward. I think everybody has a huge dependency on TSMC and we wouldn't provide -- we wouldn't propose any non-TSMC solution to our customer, if TSMC can provide related capacity. But if our customer due to capacity shortage thinking for an alternative solution, then we will -- based on the past principle, that will be noticed TSMC because we don't have a capacity, then we are thinking for alternative solution, and then we can working with other vendor closely. Yes. In fact, we -- on the previous generation, we're also working with that particular vendor to support customer project before. So I think it's -- in terms of technical, I think there's no limitation for us to find another package vendor.

Sure. So we can actually basically expect for the -- maybe the next 2, 3 years, the majority of your packaging was still at TSMC. But other than TSMC, if any other alternatives, you may also try in case of any capacity shortage.

That's correct.

Okay. The next one -- I guess Gokul raised hands ahead of Lucas. Okay. So sorry, Gokul, let's have Lucas ask question first. Lucas, please. Okay. I guess -- okay. Lucas's question is here. The management previously shared some progress on networking application. Could you provide more details regarding the current status and the specific application of your networking ASIC projects? Furthermore, what's your outlook for the networking ASIC market? And what kind of revenue contribution do you expect? Okay. For the networking projects, everything goes smoothly. We are doing multiple projects with multiple networking customers in North America, one of them already approaching the production phase. And the others, we are -- the other customer, they are trying to do multiple projects with us. And the N3 project is in design phase, and we are going to kick off the N2 project very soon. So those networking projects are actually data center related. I cannot disclose the details for each of the projects. But I would say for the networking project with the North American customer for the N2 and N3, we do expect they can contribute us meaningful revenue in the following years. The meaningful means to us, it means projects with revenue exceeding like $100 million annually. That's meaningful -- that's the definition for meaningful revenue. Okay. I guess this answers your question. Lucas, I'll go back to your second question later. So Gokul, please.

So Johnny, I think you mentioned the new hyperscaler customers coming in that you are engaging with. Can you talk a little bit about the nature of the engagement? Are they more like COT based largely like PD2 or beyond? Or they are also mostly PD0 or PD1 kind of Netlist level or spec-in kind of projects. Could you talk a little bit about what is the nature of these engagements? And your view on like how this COT trend is accelerating and the capability of some of these hyperscaler customers that you're engaging with on their ability to execute COT projects themselves?

Okay. Yes. I think to be very straight to you, I think I can say at this moment, all major hyperscalers, they have a huge intention to go to COT sooner or later. So we are discussing in so many different phases. The most simple one is our preferred model is if they have architecture-related capability, we can help them to do the back-end implementation and then doing the packaging testing. I think that's our normal business model. And some customers go even further, they say, I can handle part of a physical design. And I take the confidential block, you take the critical block, we can work together. I think this kind of flexibility we also have. And someone even say, do you want to do the production only kind of business, even though it's not our preference, but I think we also take. And -- but going further, if some of them need some I/O chiplet-related design, they don't have IP, they don't have a resource. And we find -- we are working with them to find the right IP and implement the I/O chiplet as a KGD to help them out. I think that kind of business model, we also start to establish. But if the customers do not have a capability to do anything, they need us to do the design architect design, provide the rack and those kind of stuff. I think those kind of business right now, we try not to do by ourselves, but we also have a partner that's also public announcement. We can -- if they are willing to use NVIDIA as a solution, we are NVLink partner, we can provide this kind of networking solution through partners. So there's so many different models. Like I mentioned before, now customer has so many different design opportunity. Each of the design has a multiple version and they need some -- they really need some help. And sooner or later, their preferred model is to go to pure COT. And then we can -- I still believe that -- going that direction will be favored the company like us.

Got it. That is very clear. Then secondly, I think for your automotive customer project, I think it's already ramped up. How are we thinking about the size of this business for the current generation and maybe for the follow-on generation as well, given there has been quite a bit of ups and downs in the China EV market. Are you seeing any changes in terms of the size of the demand? Or like some of the other automotive companies are also using some of these chips for nonautomotive use cases, including drone or robotics. Are you starting to see some of those things happening with this customer as well?

Okay. Actually, your question is very similar to the question from Lucas, the message board. I'll say this way, for now, we are doing -- we are doing projects with Auto for their first project and the second project. And the first one is in production already, and we are doing the second project. And as for the other automakers, for now, we are targeting 1 to 2 automakers in China. It's not proper for me to speak out the names of them here. But we do consider that some of those China automotive makers are trying to build up their own ASIC, not only for their cars, but also for the future applications such as the robotics.

Yes. And also one highlight was to mention for this kind of business. Yes, we all know the total number of cars, the market is predictable. But nowadays, because of competition, because of functionality and people -- all the car industry is planning to put multiple chips into one vehicle, the current generation, minimum of 2, then for the high-end one, they even think about the 4. So I think the -- and similar situation happened to our customer. I think starting from the next generation this year, we will see multiple chips being imported per vehicle. I think that increased the volume quite a bit.

Okay. I guess we will take the last question from Charlie to conclude this earnings call. Charlie, please.

Great. So very quick follow-up. One is a follow-up on question about 2-nanometer generation. It was about ASP revenue size, but I'm wondering about the gross margin or operating margin profile versus 3-nanometer. Is that something you can talk about?

I would say the gross margin will be pretty similar with what we have currently.

Okay.

Yes. We also negotiate with the customer very closely. I know the revenue getting bigger, but exact capacity are also getting higher. But yes, to be honestly, we are the most reasonable service provider in the industry. We need to -- yes, so customers also understand that.

Okay. And next one is a follow-up to Lucas's question about U.S. networking project. But I remember you have 1 or 2 accelerator start-up customers as well, right? Can we get a sense about the progress and can either of those be the jackpot for your revenue next year?

Okay. Charlie, I guess I can understand what the project you are mentioning to. I would say for the 3-nanometer project, it's going to kick off shortly. And for the 3-nanometer one, we expect the production revenue to kick in most likely next year, the first half next year. But for the 2-nanometer, the N2 project, we are going to kick it off very also shortly. And for the production revenue contribution from this N2 project, most likely 2028, I would say. For the scale, honestly, we don't -- I don't have a picture for you.

Okay. And last one, probably more a bit long term or kind of strategic question to Johnny. So in the recent years, you hired several high-profile senior management, including 2 years ago, your CTO. In the recent months, you hired your Chief Business Officer. I think he was from NVIDIA, right? So just out of curiosity, how do you convince them to join your team? What do you expect them to deliver for you, especially it is a CBO, right, since he comes from NVIDIA. Can we enter some partnership with NVIDIA in the future?

Yes, I think, this is a -- NVIDIA is our good partner, but I think the different company has a different strategy. I think with this new CBO on board, I think we really boost our company's relationship and reputation to the next level. I think -- to be honestly, I'm also very surprised. He well, very shocked and happy he's joining us. Last week -- two years ago, I just attended a very important event from TSMC North America VIP dinner. Literally, every single one congratulate me to hire the right people. But anyway, so we consider when we talk to hyperscaler people from the technical side, from the track record side, there's no issue. Every time when we talk, the potential customer always extend the meeting period. They really like our solution. But in terms of upper management relationship, we are kind of behind compared to our competitor in the U.S. So in the future, we will continuously to invest in U.S. and hire more related people on the business side, on technical side and in order to fulfill the gap and also in order to prepare the solution ahead of the competitors.

Got it. Yes. So with those senior customer relationship, we look forward to your future major project win.

Yes. We are all very excited that people with Freddy, his name is Freddy Engineer, with Freddy's cachet can join our company.

Right. Yes. If you have a chance to get all his background, it's all data center or hyperscaler related on their previous company in NVIDIA and also even more previous company in Xilinx.

I guess because of the time limitation, we will end our first quarter earnings call now. And if you have further questions, I guess you can mail to me or just call me. It's really easy for you to have access to our company. Thank you very much, and thank you for joining our first quarter '26 earnings call. Thank you. Thank you very much.

Thank you very much.