Operator: Welcome, everyone, to UMC's 2026 Second Quarter Earnings Conference Call. For your information, this conference call is now being broadcasted live over the Internet. Webcast replay will be available within 2 hours after the conference is finished. Please visit our website, www.umc.com, under the Investor Relations Investors Event section. Now, I would like to introduce Mr. Michael Lin, Head of Investor Relations at UMC. And Mr. Lin, please begin.
Jinhong Lin: Thank you, and welcome to UMC's conference call for the second quarter of 2026. I'm joined by Mr. Jason Wang, the CEO of UMC; and Mr. Chi-Tung Liu, the CFO of UMC. In a moment, we will hear our CFO present the second quarter financial results, followed by our CEO's key message to address UMC's focus and third quarter 2026 guidance. Once our CEO and CFO complete their remarks, there will be a Q&A section. UMC's quarterly financial reports are available at our website, www.umc.com, under the Investors Financial section. During this conference, we may make forward-looking statements based on management's current expectations and beliefs. These forward-looking statements are subject to a number of risks and uncertainties that could cause actual results to differ materially, including the risk that may be beyond the company's control. For a more detailed description of these risks and uncertainties, please refer to our recent and subsequent filings with the SEC and ROC securities authorities. During this conference, you may view our financial presentation material, which is being broadcast live through the Internet. Now, I would like to introduce UMC's CFO, Mr. Chi-Tung Liu to discuss UMC's second quarter 2026 financial results.
Chi-Tung Liu: Thank you, Michael. I'd like to go through the 2Q '26 investor conference presentation material, which can be downloaded or view in real time from our website. Starting on Page 4. Second quarter of 2026, Consolidated revenue was TWD 68.73 billion with gross margin at 32.5%. Net income attributable to the shareholders of the parent was TWD 42.26 billion, and earnings per ordinary share was TWD 3.39. Utilization rate in quarter 2 climbed to 85% from 79% in the previous quarter, and total wafer shipment in the second quarter reached 1.13 million, 12-inch wafer equivalent. On Page 5, we will start the sequential comparison. Revenue grew 12.6% quarter-over-quarter to reach TWD 68.7 billion. Gross margin rate increased by almost over a 3 percentage basis points to 32.5% or TWD 22.3 billion. And because of the recent stock market performance, our investment and dividend income together reached TWD 30 billion in the third quarter under the nonoperating income and expenses, which helped our net income to reached TWD 42.2 billion. And for the net income attributable to the shareholder of the parent is TWD 42.26 billion or an EPS of TWD 3.39 per share in the second quarter. On Page 6, for the first 6 months of the year, the annual comparison for the first half, our revenue grew 11.3% year-over-year to TWD 129.77 billion in the first 6 months of 2026. Gross margin rate also grew by over 3 percentage basis points to 30.9% or TWD 40.1 billion in the first 6 months of 2026. For the net nonoperating income, similar for what happened in the second quarter, for the first half of the total nonoperating income reached TWD 35.6 billion, which leads our -- net income to reach TWD 58.4 billion in the first 6 months of the year. EPS was TWD 4.68 in the first half of 2026. On Page 7, cash on hand is around TWD 124.7 billion, with total equity reached TWD 443.9 billion at the end of second quarter of 2026. On Page 8, our blended ASP increased by a low single-digit percentage in the second quarter of 2026. For revenue breakdown on Page 9, Asia remained our largest revenue pool, around 66% of total revenue, and North America reached about 22%. On Page 10, IDM didn't really change much. This quarter is around 15% versus 14% in the previous quarter. For Sales breakdown by application on Page 11, and there's almost no change for the revenue among 3 major segments. So revenue breakdown by technology on Page 12. Our total revenue under 40nm still around 52%, with 22nm/28nm is becoming our largest revenue pool, represent 37% of the total revenue. Our quarterly capacity has shown some increase in our Singapore site, Fab 12i for the second quarter, and there will be a more meaningful increase in the coming quarter to reach 192,000 12-inch wafer capacity for our Singapore site. On Page 14, our annual CapEx budget has raised or increased to $2 billion from the previous number of $1.5 billion, which we will elaborate in more detail later during the conference call. So this above is the summary of UMC results for second quarter of 2026. More details are available in the report, which has been posted on our website. I will now turn the call over to CEO of UMC, Mr. Jason Wang.
Jason Wang: Thank you, Chi-Tung. Good evening, everyone. Here, I would like to share UMC's second quarter results. In the second quarter, our wafer shipment increased 10.6% quarter-on-quarter driven by strong demand in communications and consumer segments, further improving utilization rate to 85%. Revenue from our 22/28nm business continues to set record high with 22nm revenue representing 17.5% of the second quarter sales. Earlier this month, we announced the company's first mass production delivery of a 12-inch photonics IC to a customer, a major milestone for UMC that demonstrates the company's high-volume silicon photonics manufacturing capabilities on 12-inch wafers. As we prepare to launch our silicon photonics platform available for general customer use in 2027. Looking ahead to the third quarter, we expect demand momentum to remain stable across the computer, communication and consumer segments with the shipment projected to increase by high single digits, driven by strong demand of power management IC, sensors and microcontrollers. Our 8-inch portfolio is also seeing a strong rebound and utilization is expected to improve significantly in the third quarter. With our 12-inch capacity already at a healthy utilization rate supporting core business, we must also prepare in advance to ensure UMC is well positioned to capture future opportunities driven by AI. To ensure we are ready to scale rapidly to support our customers, we announced today that our Board of Directors has approved a plan to expand cleanroom capacity at our Singapore P4 facility and to construct a new fab in Tainan, Taiwan. The plan will be executed in phases, enabling UMC to remain focused on capital discipline with the pace of facility deploying capacity to fulfill customer demand. As a result, 2026 capital expenditure budget will be revised upward to USD 2 billion. Now, let's move on to the third quarter 2026 guidance. Our wafer shipment will increase by high single digits. ASP in U.S. dollar will remain firm. Gross margin will be in the mid-30% range. Capacity utilization rate will be greater than 90%. To support the growing customer demand in silicon photonics advanced packaging, the 2026 CapEx will be raised to USD 2 billion from USD 1.5 billion. That concludes my comments. Thank you all for your attention. Now we are ready for questions.
Operator: Yes. Thank you. And ladies and gentlemen, we will now begin the question-and-answer session. Now we'll have our first question, Laura Chen from Citi.
Chia Yi Chen: Congratulations for the good result and outlook. It's great to see that UMC has a good progress on the silicon photonics and also see that power management IC improvement. So I assume that these are all prepared for the AI-related business. So I'm just wondering for the longer term, does UMC have any indication or target of your AI-related revenue? And can you also give us a breakdown of products that you are aiming for?
Jason Wang: Well, the -- first of all, our AI-related business is driven by the specialty semiconductor solution, supported a broad range of applications, including the power management, connectivity, FPGA as well as our growing advanced packaging and silicon photonics business. Those are the focus. This business will have already begun important -- become a contributor to our growth in 2026. The current revenue for 2026 is projected to close to approximately $300 million for this year. And looking ahead, in 3 years, we expect the AI exposure to exceed USD 1 billion.
Chia Yi Chen: Very helpful. My second question is about, like, the overall demand outlook. We know that Q3, we will see the deterioration rate improvement. But I think consumer electronics demand in general is still quite weak. So what's the management view about the sustainability into probably Q4? Or do you have any visibilities into maybe early next year?
Jason Wang: Well, right now, what we see from the market is the worldwide demand is improving with a broader and more sustainable momentum. But it remains more AI-led. AI is leading that and it's building over into memory, connectivity and power segment. Now, AI demand recovery is still mixed across different end device markets and supply reduction plus inventory normalization but moving to the market towards more of a balanced and predictable environment. So we do see the future outlook is increased in terms of visibility. For the AI-related demand, while they coming -- remain very strong, what may be changing is the XPU demand remains strong. Besides that, the bottleneck is not only on compute, but also on memory, connectivity, power management that I mentioned earlier. In the near term, we are seeing upside to our silicon photonics, power and FPGA-related products, particularly in our 40nm and 65nm technologies. For the non-AI, it's uneven. So we will not categorize the current environment as a full-blown broad-based recovery yet. In the near term, we do acknowledge that consumer segment, including the handset, PC and notebook will experience a year-over-year decline. However, for UMC, our wafer shipment will grow year-on-year on 22/28nm as well as our 8-inch business. In conclusion, our 2026 wafer shipment will increase both driven by our own foundry share gain as well as a customer share gain in both AI and non-AI market. The 8-inch loading will improve to mid-80% range, while mature 12-inch loading will also increase quarter-on-quarter on AI-related demand.
Chia Yi Chen: Okay. In that case, can I also just quickly check what's your view on the ASP trend into the second half?
Jason Wang: I mean, the -- our pricing strategy has always remained consistent. We are not trying to maximize the short-term pricing based on the market cycle. Instead, we focus on maintaining a sustainable business model through value-based pricing that reflects our differentiated technology, manufacturing capability and long-term partnership with customers. As the demand in industry conditions continue to improve, we are working with the customers to ensure pricing appropriately reflected that value while supporting continuous investment in technology and capacity.
Operator: Next one, Gokul Hariharan, JPMorgan.
Gokul Hariharan: So Jason, could you talk a little bit about your -- the capacity expansion plan at Singapore P4 as well as your plan to construct a new fab in Tainan? Like, what are the kind of milestones we should look at? And what are the kind of phases that you are looking to kind of phase this capacity in? I think how should we compare it to your previous plan about 4, 5 years back when you had this new phase in Tainan and then you also build out the Singapore fab? Is it similar scale or is going to be much smaller scale?
Jason Wang: Okay. Let me maybe start off what drives that, right? What drives that decision that we made today. When we map out the industry growth over the next 5 years, we see several important trends. Within the AI data center, while growth in compute and memory will remain high, we also see the connectivity and power also growing at a high CAGR, driven by the need for more bandwidth and more efficient power. The second is the automotive electrification trend continues. They are not just for EV, but for autonomy and infotainment applications. This third is when we look at this emerging applications such as the robotics, satellites, we know that again, very, very high growth in compute, memory, sensing, connectivity and power. These industry trends, combined with our entry into the advanced packaging such as logic and memory stacking as well as the silicon photonics, will accelerate growth within the UMC addressable market. Within our existing portfolio, we also see that advanced packaging is enabling our customers to think differently about new architectures and designs of their products. Since we believe we are ahead of our peers in advanced packaging, this is leading to share gain and many new opportunities. Therefore, we actually believe the TAM actually will grow significantly compared to the past. Consistent with this long-term outlook, in terms of how to do this, we're going to execute each expansion phases we will pursue is based on the market validation, this alignment and customer commitment, which will provide both long-term capacity assurance to our partners and loading protection to UMC. So that's truly the systems behind the decision as well as how we want to execute this. You also have a follow-on question in terms of scale?
Chi-Tung Liu: The details and the schedule.
Jason Wang: The milestone and the schedule, right? For right now, for the Singapore, we will invest in the cleanroom for our P4 facility and the 2 purchase to expanding our silicon photonics capacity. In Tainan, the construction of 12A, P7 and P8 will set up a robust foundation for the UMC to scale the advanced packaging alongside with the customers, long-term product road map, but that's only going to be more of the foundation. We're building the shelf for the 12A, P7 and P8. The cleanroom installation at this time is focused on the Singapore P4. And from the milestone line, we -- now the Board has approved it and then we're going to start engaging with the follow-on activities and then we will report back in terms of the install schedule once we get the confirmation from our [ installers ].
Gokul Hariharan: Got it. Understood. And just follow up on the Intel collaboration on the 12nm now that we are likely to start recognizing some revenues next year. How should we think about how meaningful these 12nm base revenues are going to be? And is it going to be like gross margin accretive given you have like a revenue sharing and a profit sharing kind of agreement with them? Could you talk a little bit more about how we should model this contribution going into next year?
Jason Wang: Well, first, let me update the 12 current status. The -- overall, the 12nm cooperation project with Intel is advancing smoothly, and we anticipate the product tape-out will commence in 2027. So all the PDK will be ready in May 2026, and the customer design and design in and ready to tape-out in 2027. So 2027 was still at the early stage of ramping the product, I would say, probably at a pilot stage. The production will probably be more meaningful in 2028. So at the current status of the 12nm, in terms of the business model, there will be accreted to our current models -- current financials.
Gokul Hariharan: Okay. Maybe one last question on gross margins, Jason and Chi-Tung. So you are already about guiding for about 90% utilization and gross margin is kind of reaching mid-30s. Any thoughts about how you think gross margins could progress in this cycle because it feels like in the past cycles where utilization had reached above 90%, gross margins were probably at a higher level given you probably had a little bit more price leverage. So I just wanted to understand how you think about like where gross margins can reach given the demand seems to be pretty strong and utilization continue to improve. Can we go beyond 40% gross margins like we did in -- back in 2022 or that is a little bit of a challenge right now?
Chi-Tung Liu: So our utilization rate and operating income has increased versus last year, we will continue to improve our fundamental, and we do expect the new project like silicon photonic and advanced packaging will enhance our EBITDA margin, but the higher depreciation expenses will certainly have a swing impact on the gross margin. So I think with the announcement of the new fab in Tainan and also the new cleanroom in Singapore, no doubt, the depreciation expenses will increase as a result. So we are confident to deliver higher profit numbers and also improve -- enhance our EBITDA margins. But gross margin will also highly depend on the installed equipment and the depreciation expenses curve as a result.
Gokul Hariharan: Okay. Chi-Tung, I think previously, we were expecting maybe after this year, depreciation curve could kind of taper down. Is that still the case? Or should we expect that there is still some increase in depreciation next year as you bring on some of the new cleanroom and some new phases?
Chi-Tung Liu: Yes, the newclean room and the new shell in China certainly changed the curve. And now we are expecting the depreciation to increase by low teens for over the next 2 years at least.
Gokul Hariharan: So low teens each year over the next few years or low teens...
Chi-Tung Liu: Early it will be each year, at least, yes.
Operator: Next one, Charlie Chan, Morgan Stanley.
Charlie Chan: Jason, Chi-Tung, yes. So just some quick follow-up about a previous topic. So first of all, I know you kind of gaining share, but how about your customers' chip inventory, right, because the end market seems to be pretty challenging. So according to our analysis, it seems like your -- some of you have fabless revenue is kind of undergoing your wafer shipment. So I'm wondering whether you're concerned about the chip inventory accumulated at your customer side.
Jason Wang: Well, I mean, Charlie, the first -- I mean, we're always cautious about the inventory situation, right? I mean -- but not to the level that we need to be worried at this point. For the first quarter '26, the DOI level is actually rose slightly. They've gone up a little bit. What we see is really coming from, one, PC was strongest driver of inventory buildup as the AI infrastructure build cycle is still ongoing. In the PC -- I mean the HPC was the strongest driver. In the PC segment, early stocking and shipments occurred in response to rising memory prices, we are seeing that. The inventory and the DOI for the smartphone and consumer segment are rising simultaneously, indicating the end market remains weak. So there are some areas are weaker. And although the demand in automotive and industrial side is stable, their DOI remained higher than the historical average. For the Q2 '26, while we're tracking that the overall consumer spending remains weak, and we expect the semiconductor sales to stay strong in 2026, so which will drive the DOI by several days again. But certainly, we're not to the level that we need to be worried, but we are cautious and continue tracking the progress of that.
Charlie Chan: I see. Yes. So with that kind of impact your kind of negotiation with customers to passing through the cost because you're kind of increasing your investments, right? Do you think there's some dynamic change for you to reflect your value next year?
Jason Wang: Well, I mean, first of all, the market outlook for us is we remain optimistic because it's driven by both AI-related and non-AI-related demand. While those demand and industry continues to improve, the pricing environment become more constructive, so the discussion -- engaging discussion with customers is actually a constructive discussion, giving that visibility, and we expect the annual pricing trend to be better than we anticipated, and we're even expecting more meaningful pricing uplift in 2027.
Charlie Chan: Okay. Okay. Great to hear. Yes. And can I follow up a little bit about advanced packaging business plan because I think it's a pretty into me that you want to extend your event advanced packaging cleanroom for two phases. My understanding is that for 2.5D, your previous focus was more about the interposer production, and you do have some 3D IC, but this is more for RF. First of all, are you going to do like the full stack of 2.5D, for example, CoWoS and for the 3D IC, are you going to do not just the RF, but also some AI accelerator kind of products. And lastly, we noticed that your Vice Chairman right now kind of move to Unimicron. So I'm not sure whether we can link these two development together. UMC very aggressive advanced packaging and your kind of partnership with the Unimicron in substrate will be even closer than before. Should we think in that way?
Jason Wang: I mean from our advanced packaging offering, I'll start off with that, you're absolutely right. We start with the interposer solution and follow with the RFSOI 3D IC, the chiplet solution. But it's actually -- offering is actually more than that now. The overall addressable market is projected to more than double by 2030 in our addressable market because it expanded from the 2.5D interposer with DTC, discrete DTC, 3D wafer to wafer stacking which that's what you're referring to the RFSOI and the memory to memory stacking as well. Our customer engagement is building up. with more than 10 active customers now and over 35 new products in discussions and expect to tape out in 2026 and early 2027, and we actually feel pretty optimistic about this advanced packaging space. We -- and that's why we start to deploy the facility preparation. Now, we have -- like you said, we have already entered production for the 3D wafer-of-wafer hybrid bonding, bridge die and discrete DTC, but they will follow by the wafer-to-wafer stacking as well. So we'll continue to broaden our advanced packaging offering, and -- but not to the CoWoS solution because it's not a platform solution. It's advanced packaging capability serving various different combinations of a solution, both from our existing offering to even the new offering, new integration options. So we see many different new exciting opportunities there, but not a CoWoS, okay? Now, in terms of the ecosystem, I mean, we're working with entire ecosystem in terms of the -- our solution to serve our customers. So it's not going to be limited to any one particular.
Charlie Chan: Okay. Any kind of a strategic angle or synergy or partnership between you and Unimicron going forward besides the kind of financial investment?
Jason Wang: At this point, it is mainly driven by the financial. It's one of our investor portfolio company, and we remain as an investor. And so not at this current point. But once there is an opportunity, right, in the ecosystem, we will definitely explore that with them.
Charlie Chan: Okay. And last one, if I may. Just the AI-related revenue. So Jason, did you say that your overall revenue from AI already exceeded the USD 1 billion. What was that, your comments?
Jason Wang: Yes, close to $300 million this year. And in 3 years, we will exceed $1 billion.
Charlie Chan: Okay. Then, how would you define the AI-related revenue? I believe all I can see silicon photonics, silicon interposers. But what's your definition about your AI revenue?
Jason Wang: Well, the solution associated with AI as end product, including the connectivity, power management, those are all category of AI [indiscernible].
Charlie Chan: Okay. So silicon photonic can be considered as the connectivity.
Jason Wang: Yes, as part of the connectivity solution, yes.
Operator: Next one, Sunny Lin, UBS.
Sunny Lin: Congrats on the very strong outlook. So I want to follow up on the new expansions. So may I double confirm that these 2 new expansion will be mainly for silicon photonics and advanced packaging, not for a typical foundry business. And Jason, you just mentioned that you do think advanced packaging overall the addressable market for UMC should more than double by 2030. Would you be able to share with us the addressable market that you are forecasting for UMC in the coming few years? And would it be fair to assume that maybe we see that the capacitor -- embedded capacitor may be the major one? Or how should we think about the contributions from the various projects that you are working on?
Jason Wang: Yes. I mean, first of all, for the announced the cleanroom preparation in our Singapore P4 facility is prepared for the silicon photonics capacity. And the Tainan facility of the P7 is prepared for the advanced packaging. And in terms of the advanced packaging addressable market growth, that includes the DTC, like you said, and the interposer with DTC, discrete DTC, the wafer-to-wafer stacking and the memory to memory stacking, a very broad offering of advanced packaging. Now the DTC does play a major part of that, and that's a very significant part of that. Since that we have more than 35 different products taking out, there's a high percentage of those are considered the DTC.
Sunny Lin: Got it. Very helpful. So would you be able to share with us any color about the size of the addressable market that you are looking at? And also it would be helpful for us to think about the trajectory for your revenue coming from advanced packaging. And so would it be fair to assume that your advanced packaging should see inflection point when your new supply starts to be on board? And given the lead time for expansion, would that be mainly maybe in like late '28, '29 time frame?
Jason Wang: Yes. I think given the current construction time frame, it's been stretched out. There's a lot of activity going on in the construction side. So from a timing standpoint, we do looking at somewhere greater than 20 months in terms of lead time. And so that will put us into '28 and '29 cycle. However, like the earlier question, we -- in terms of milestone, I will probably like to share them when I have more specific data because right now, we're getting roughly a 20 months of lead time on the construction. But since we're just approved today, and we're engaging that discussion, and we'll probably report back when we have more specific.
Sunny Lin: Got it. And maybe a question on your Singapore expansion for the P3. And so now given the stronger demand outlook, how should we think about the capacity that you are going to ramp by end of this year and also by end of 2027?
Jason Wang: Yes. Part of the approval budget CapEx today, that includes some of the capacity expansion in our P3 facility. And in the existing facility, we will be adding the BCD, which is power management solution as well as silicon photonics. And for the 12A in Taiwan, we will start putting some of the CMS, which we call customized memory stacking and the DTC solution in our 12A in Tainan. Meanwhile, we're also putting some test on the capacity setup in our 8-inch facility. So that will be underway now. That's all going to happen within our existing facility.
Sunny Lin: Got it. So sorry, just to add one question on Singapore P3. So is there a target for your capacities maybe, let's say, by end of this year and end of next year?
Jason Wang: Yes. I mean once we release this given the tooling time and install, we expect we will see some production ramp on this in, I would say, late '27, early '28.
Sunny Lin: Got it. Okay. Maybe last one on silicon photonics. So now given you have 2 solutions, one on 12-inch that you licensed from imec and the other one on 8-inch by TFLN. And so would you be able to share some color regarding the respective strength of two solutions? And how should the client choose? And based on your current development, which one do you think may drive more meaningful revenue contribution in coming years?
Jason Wang: Okay. Let me maybe start off. For silicon photonics, we are releasing the 12-inch solution. And that's which we believe is the -- that will be the best solution for our customers and how to differentiate while our other competitors do on 8-inch. The 12-inch will offer better process control, which will give us better performance. For example, the propagation loss, better yield, and we have demonstrated that on silicon wafers already to our customers. For the TFLN, we have the world's first TFLN modulator in production already, and we are working on the 40G per lane and for the 3.2T, that is based on the TFLN for the customer today. So we think that's actually the best solution for beyond the 400G. And combining the 2, there will be an integration option for those 2 and we can implement the TFLN with our silicon photonics, the PIC and the through our advanced packaging solution. And along with that, we can offer the optical IO, the OIO, which is interposer with the PIC through our advanced packaging. And we also can provide TFLN a component for the CPO solution. And so we think there's a lot of various combination of this and between the 2. And so we think that we have a unique position on that.
Operator: Next one, Haas Liu, Bank of America.
Haas Liu: Congrats on the great results. My first question is regarding your CapEx outlook in the next couple of years. You mentioned that AI revenue could be in 3 years could be at around USD 1 billion contribution. So how should we think about that your CapEx growth trajectory in the next few years to achieve that goal? And relatedly, I think how should we also think about your equipment investment as a percentage of your CapEx in the next 2 years? Because it seems that you meaningfully revised up your depreciation outlook for the next 2 years.
Chi-Tung Liu: For CapEx, today's Board meeting approved nearly USD 5 billion already. So that's the number we will work with for the next -- over the next 2 years or maybe 3 years. And as our CEO pointed out, this is going to be a phased construction or phase expansion strategy. So we will adjust accordingly based upon customer commitments and our customer wins. So we don't really have a full number, but to start with USD 5 billion for the next 2 to 3 years is probably the beginning. And in terms of depreciation, as I mentioned, earlier it will be low teens increase year-over-year for at least this year as well as 2027. And for 2028, it will depend on the CapEx number based upon the factors I just highlighted.
Haas Liu: Got it. That's very clear. So in the next 2 years in 2027 and also 2028, your CapEx will be at least USD 5 billion for the new investments?
Chi-Tung Liu: Total will be -- today's Board approved USD 5 billion, which will be spent across 2026 and '27. And the '28 numbers will depend on the phase expansion.
Haas Liu: Got it. Very clear. And then I think just regarding that capacity expansion plan you are targeting and the AI revenue mix you are targeting that it seems like in 3 years based on our model that your AI revenue could reach 10% of your total sales versus low single digits this year. Would you be able to try to -- or do you have any view on which part of the applications could actually be the main drivers? You mentioned a couple of drivers, for example, like connectivity, silicon photonics and also power management IC. But would you be able to rank it in terms of the growth rate or from the revenue contribution perspective, which part of the application is going to be the key driver?
Jason Wang: Obviously, from the growth standpoint, because the silicon photonics and the advanced packaging is still at early stage. So from the compound annual growth rate standpoint, they are the fastest growing and the highest growth rate. And -- but we also believe on the existing -- our current existing solution will also grow. And so -- and it's driven by the -- also the AI and non-AI space. So those will also grow in a lower pace growth rate.
Haas Liu: Okay. Yes. And that's my follow-up question is just regarding your strategic positioning. We start ramping your capacity for silicon photonics and advanced packaging in 2028. I was just wondering compared to your peer solutions, which probably have already been in the market for a few years, what do you think your strength is compared to them? Is it from a technology road map perspective or your customer relationships? Or is it still coming from the demand spill over, which could actually -- you could also capture some of the fast expanding addressable market within that big pie.
Jason Wang: Well, I mean, I think on the silicon photonics, like I mentioned earlier, we are the first offering the 12-inch solution, and we believe that's actually a better process compared to the 8 inches on the market today. So it offers much better performance and the process control. So we think there's a huge differentiation there. So I think that's the technology process advantage. So for the TFLN [indiscernible], we also believe that's the best option for anything beyond the 400G in the market today, and we're going to be massive -- putting the mass production on that as well. So I think there's a big technology differentiation, not a spillover. I mean there's -- on the existing solution, I think there will be there will be multiple factors. And -- but in a much lower growth rate is because some of the spillover reason. But despite the existing technology also has our own driver that drives the growth. But coming back to the silicon photonics, which is the biggest, highest driver as well as the advanced packaging, I think we have a very good differentiation there.
Haas Liu: Okay. Yes. And I think just a quick follow-up before jumping back to the queue is that your gross margins had a pretty nice uplift in the second quarter. And I think the guidance for the third quarter near term is also pretty solid as well. Would you be able to quantify the factors supporting your gross margins? For example, like utilization, pricing and FX, any of these are positives or negatives? Could you just try to share with us in a more quantitative way?
Chi-Tung Liu: So the higher quarter 3 gross margin guidance is mainly attributed to the higher utilization rate. So loading was 85% in second quarter. Our guidance for the third quarter is 90% plus. And Yes, there will be multiple factors, including ASP product mix and utilization rate, foreign exchange rate and depreciation, et cetera, et cetera. So our focus is certainly to enhance our profitability. But as I mentioned earlier, for the next 2, 3 years, we will continue to improve, deliver better results for EBITDA margin and gross margin will come along with the depreciation curve.
Haas Liu: Yes, that sounds great. And I think just one more is probably just on your power IC exposure. Could you share with us how much of your revenue is coming from these products, no matter it is for which end markets? And I think second thing is probably just on the utilization by inch and also 12-inch. Last time, you mentioned that 12-inch was still slightly higher than 8-inch. But what's your view right now for same quarter and third quarter?
Jason Wang: I mean the 12-inch is still above corporate average. We believe the -- we look at the -- we expect the Q3, the loading will be greater than 90% and 12-inch is greater than corporate average and the 8-inch is below. And I think I actually mentioned that earlier, I think the 8-inch will reach 85%.
Haas Liu: Okay. So your power IC exposure is how much percentage of your sales now?
Jason Wang: I mean we categorize that is the part of our specialty offering. And the specialty offering today is representing 50% of our revenue today.
Operator: Next one, Katherine Yu, Goldman Sachs. Katherine Yu So my first is on -- I would like to know what's our strategy on more advanced nodes going forward because we're now working with Intel on 12nm and where does that go from here? Or is it fair to think that we could enter more advanced nodes, say, 7nm and below? And if so, what would be this model look like? And maybe I'll put it this way, that what are the key factors that we need to see before we're committed to expanding beyond 12nm?
Jason Wang: I think the simple answer to that is we have to first deliver the 12nm. We have to prove the business model as well as deliver the 12nm. And so the overall, the 12nm collaboration project is going smoothly. And so I think the 12nm need to be the solid foundation for us to explore the next generation. But meanwhile, the 12nm representing more than just the pure logic today, we actually are already expanding that to the high-voltage from the 12nm, which is 14 [indiscernible]. And there are more in terms of derivative and specialty technologies in discussion right now. So there's a lot of activity and lots of work that we have to get done for the 12nm today. But yes, I mean, the simple way to look at it is we have to execute the 12nm and then we will explore beyond. Katherine Yu All right. So is it fair to say that beyond maybe enter into 7nm and beyond might be after 2028? Well, I mean, if there is a discussion, we'll probably have more clear milestone.
Chi-Tung Liu: Yes. This is really need to be a mutual beneficiary collaboration and the current focus is on 12nm only. Again, UMC is always open to find the best solution to have a low asset type of migration. But again, without a successful 12nm, it's going to be difficult. Katherine Yu Got it. So my second question is on how would you characterize the current cycle now versus the chip shortage cycle in 2021? I think the last time was a more broad-based supply-driven super cycle with utilization over 100% with pretty aggressive ASP increases almost every quarter. And this time, the recovery looks more narrow to more concentrated on AI. So do you agree that the nature of the demand has actually fundamentally changed? And the key question I want to ask is that how should we think about your margin trajectory going forward and the pricing power this time versus the last up cycle? Do you think it's possible for your gross margin to surpass the level -- the peak level that we saw in 2022 in the coming years?
Jason Wang: Well, maybe start off with the outlook. I mean, we -- like you said, we are seeing the AI-related segment remain as the primary growth driver for the entire industry. And with the continuous commercial deployment of edge AI application, demand for chips or in general purpose server is also expecting to rise, but that's also related to AI. In contrast, the non-AI demand is still mixed across different end markets, end device market. While the overall semiconductor industry is projected to grow higher to maybe low 20% range this year compared to earlier year, increase of low 20%, we think mainly driving by the AI. So yes, it is different than the previous. It's not a broad-based recovery. And I will probably have Chi-Tung talk about the gross margin. But before that, and I have the same answer on the previous about the ASP is we do see the demand and the industry conditions is improving. So the pricing environment has become more constructive. So we think the pricing trend will be better.
Chi-Tung Liu: So as for gross margin, we really don't compare ourselves to the historical data. I mean we certainly try very hard to deliver higher profit in absolute dollar terms back to our shareholders. And as I mentioned, because of the new fab ramp, both in Singapore and in Tainan, so the depreciation expenses and how we amortize them will have a big impact on the near-term gross margin. So I think we are a lot more confident to say our EBITDA margin will show steady growth over the next cycle or next few expansion phases. But the gross margin will be largely dependent upon how we book the depreciation.
Operator: And ladies and gentlemen, we thank you for all your questions, and that concludes today's Q&A session. I'll turn things over to UMC Head of IR for closing remarks. Thank you.
Jinhong Lin: Thank you for attending this conference today. We appreciate your questions. As always, if you have any additional follow-up questions, please feel free to contact ir@umc.com. Have a good day.
Operator: Thank you. And ladies and gentlemen, that concludes our conference for 2Q '26. Thank you for all your participation in UMC's conference. There will be a webcast replay within 2 hours, and please visit www.umc.com under the Investors Events section. You may now disconnect. Thank you again. Goodbye.
