United Microelectronics Corporation (UMC) Q3 2026 Earnings Report, Transcript and Summary

Welcome everyone to UMC's 2026 third quarter earnings conference call. All lines have been placed on mute to prevent background noise. After the presentation, there will be a question and answer session. Please follow the instructions given at the time if you would like to ask a question. For your information, this conference call is now being broadcast live over the internet. Webcast replay will be available within two hours after the conference has finished. Please visit our website www.umc.com under the Investors Relations Investors Events section. Now, I would like to introduce Mr. Michael Lin, Head of Investor Relations at UMC. Mr. Lin, please begin.

Thank you, and welcome to UMC's conference call for the third quarter of 2026. I am joined by Mr. Jason Wang, CEO of UMC, and Mr. Chitung Liu, the CFO of UMC. In a moment, we will hear our CFO present the third quarter financial results, followed by our CEO's key message to address UMC's focus and fourth quarter 2026 guidance. Once our CEO and CFO complete their remarks, there will be a Q&A session. UMC's quarterly financial reports are available at our website www.umc.com under the Investors Financials 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 risks 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 the ROC security 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. Chitung Liu, to discuss UMC's third quarter 2026 financial results.

Thank you, Michael. I would like to go through the third quarter 2026 investor conference presentation material, which can be downloaded for view in real time from our website. Starting on page four, the third quarter of 2026, consolidated revenue was TWD 74.24 billion, with gross margin rate 34%. Net income attributable to the shareholder of the parent was TWD 16.43 billion, and the earnings per ordinary share were TWD 1.32. Capacity utilization rate in the third quarter reached 92%, compared to 85% in the previous quarter. Overall, EPS for the quarter was TWD 1.32. On page five, revenue grew 8% sequentially to reach TWD 74.2 billion. As we stated earlier, gross margin rate also improved from 32.5% in the previous quarter to 34% in this quarter for TWD 25.2 billion. EPADS is $0.206. For the year-over-year comparison on page six, revenue grew 16.1% to TWD 204 billion. Gross profit also grew 30.8% year-over-year to TWD 65.3 billion or 32% of the gross margin rate. With the help of the non-operating income, overall net income attributable to the shareholder of the parent grew 136.5% year-over-year to reach TWD 74.86 billion for the first nine months of the year. EPS for the first three quarters reached TWD 6 in NT dollar terms, or $0.945 per ADS in U.S. dollar term. On page seven, our balance sheet at the end of third quarter, total equity is TWD 470 billion, with cash account for about TWD 142 billion on September 30th. For EPS, sorry. For blended ASP in the third quarter was relatively stable, remained firm compared to the previous quarter on page eight. For revenue breakdown on page nine, there's very little change for quarter two compared to quarter three of 2026 for revenue breakdown by geography. On page 10, our revenue breakdown by customer type also remained virtually unchanged. For IDM, it was about 14% of the total revenue, and fabless account for the others. For revenue breakdown by application consumer stay unchanged at around 32%, when communication and computer also stay quarter-over-quarter unchanged. On page 12, our 22 nm and 28 nm revenue still account for more than one-third, reach about 35% of the total pie chart, and 40 nm is around 16%, and 65 nm is about 19%. Our quarterly capacity continue to increase slightly, mainly coming from our Fab 12i facility in Singapore. For quarter four of 2026, we expect the capacity for overall UMC reach about 1.33 million 12 in-equivalent, 1.33 million for 12 in-equivalent capacity by Q4 2026. Capacity compared to the last quarter's estimate still remain unchanged around $2 billion. The above is the summary of UMC result for third quarter 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.

Thank you, Chitung. Good evening, everyone. Here, I would like to share UMC's third quarter results. In the third quarter, we saw a healthy demand across all market segments, driving a 7.3% growth in wafer shipments to more than 1.2 million 12 in-equivalent wafers, a record high. Our overall fab utilization rate reached 92%, with several of our 12 in fabs operating at full capacity, while 8-in operations also saw a strong recovery. By technology node, our 22/28 nm business had another record-breaking quarter, with 22 nm revenue now accounting for 18.3% of wafer sales. As the customers continue to tape out new products, including display drivers, automotive radar, and Wi-Fi ICs. In new business area, active customer engagements for advanced packaging and silicon photonics remain robust. We expect this engagement to translate into accretive growth over the coming quarters as we ramp existing projects and support new product tape-outs. Looking ahead to the fourth quarter, we expect the positive demand momentum to sustain, supporting stable shipments and utilization rates. Revenue from our specialty technology portfolio will continue to grow sequentially, driven in part by the ramp-ups of our advanced packaging and silicon photonics businesses. We have seen a significant increase in demand for these new business areas, and we are in the process of converting some existing capacity to address near-term unfulfilled demand. In the longer run, our two-track expansion plan in Singapore and Tainan will be crucial to ensure UMC's capacity growth can keep pace with our customers' needs. We are pleased to announce our board of directors today approved an additional capital expenditure budget of approximately $3.2 billion, which will enable us to accelerate our expansion schedule as we continue to align with customers on their product roadmap and requirements. Now let's move on to fourth quarter 2026 guidance. Our wafer shipment will increase by low single digits. ASP in U.S. dollar will remain firm. Gross margin will be comparable to the preceding quarter. Capacity utilization rate will be in the low 90% range. Our 2026 cash-based CapEx budget will be at $2 billion. That concludes my comments. Thank you all for your attention. Now we are ready for questions.

Yes. Thank you. Ladies and gentlemen, we will now begin the question-and-answer session. If you have a question for any of today's speakers, please press star key and number one on your telephone keypad, and you will enter the queue. After you are announced, please ask your question. If you find that your question has been answered before it is your turn to speak, please press star key and number two to cancel the question. Thank you. Now, first question will be coming from Sunny Lin, UBS. Go ahead, please.

Good afternoon. Thank you very much for taking my questions. Several questions from my side. Number one, now that UMC will finish 2026 on a high note, any early view for 2027, based on current visibility?

Yes. For 2027 early view, we expect the growth momentum established in 2026 will continue into the next year, supported by the healthy demand across our core business and growing opportunity in differentiated technology and high-value applications. If we look at this, in 2026, our growth has been driven primarily by a high wafer shipment, improving utilization and customer share gain. Looking ahead to 2027, we expect the further growth, although with utilization already above 90% and several technologies operating at or near full capacity, our shipment growth will increasingly be constrained by available manufacturing capacity. As demand AI remains a key growth driver, creating demand across power management, FPGA, and connectivity application. Meanwhile, silicon photonics and advanced packaging are emerging as a new growth engine for UMC, addressing the next generation optics, connectivity, power delivery, and memory bandwidth requirement for AI infrastructure. Beyond AI, we continue to see opportunity from customer market share gain and increasing foundry share across premier smartphone, automotive, industrial, and other specialty applications, despite an uneven recovery in some end markets. With customer demand in selected technology already exceeding our available capacity, we see opportunity to capture greater value from our existing manufacturing footprint to improve pricing and richer product or technology mix in 2027, alongside continued shipment growth. While 2026 growth was driven primarily by volume and utilization, we expect 2027 to become increasing value driven, with higher value per wafer and differentiated technology, and contribute more to UMC revenue growth and long-term earning potential.

Thank you very much. Very helpful and encouraging. On the pricing side, how should we think about the ASP environment for UMC, especially going to 2027? How are the negotiation with key clients progressing for the coming few quarters?

We are engaging with customers right now for our 2027 pricing, and we're seeing a significant shift in pricing environment heading into 2027, supported by strengthening the customer demand and increasingly tight capacity in selected technologies. We see clear opportunity for pricing improvement across both our 8-in and 12-in business, while maintaining our discipline and value-based pricing approach. Importantly, we expect 2027 to mark a turning point in our blended ASP trajectory, moving from an environment of ASP erosion toward a meaningful positive uplift beginning in the first quarter. While we anticipate a 10% blended ASP increase year over year, this is more than an incremental pricing improvement. It reflects both a more constructive pricing environment and our increasing ability to capture values through a richer product and technology mix. Our record high 22 nm revenue contribution and continued specialty technology momentum also provides a strong foundation, while silicon photonics and advanced packaging create additional higher value growth opportunity at scale.

Thank you very much. Also, maybe a quick follow-up. You mentioned 10% year-over-year improvement. Is that initial view for your blended ASP improvement going to Q1?

No, that will be on a year-over-year basis.

So 10% will be your target through 2027, not just for Q1?

Yeah. As a whole. A year-over-year. As in 2026 year compared to 2027, year-over-year. Yeah.

We will provide Q1 ASP guidance for the next quarter's earnings call.

Got it. Maybe on the CapEx. You just mentioned an additional $3 billion budget on top of the $5 billion budget announced a quarter ago. Now, could you share a bit more color on how that translates into a capacity expansion in the coming few years? On silicon photonics and advanced packaging, how are you seeing the market demand overall in coming years?

We are accelerating selected investments because several of our new growth business are moving from technology development into a commercial execution, which faster than our original capacity roadmap anticipated. We are seeing three things happen at the same time. The market is expanding as the AI infrastructure drives higher requirement for bandwidth, power delivery, memory, performance, and system integration. Our customer engagement are also progressing from qualification into production ramps and capacity planning. At the same time, our solution roadmap is moving forward with a product already in mass production and additional product completing tape out, customer validation, and new generation of product already under development. Our focus now is shifting from whether this opportunity will materialize, to how quickly can we expand capacity to meet the customer demand. Under our current expansion roadmap, we are targeting monthly capacity approximately 18,000 wafers for silicon photonics, 22,000 for DTC and silicon capacitor, and 20,000 for TFLN modulators. This next phase of the capacity will start its ramp in late 2028 and reach its final scale by 2030. This is not a broad-based capacity expansion. We are directing our capital toward customer-backed growth opportunities that will fuel UMC's next phase of growth. That will give you a feel of why we are accelerating and milestones that we plan for those CapEx. For your question on the silicon photonics and advanced packaging, we believe UMC is well-positioned to capture the next wave of AI infrastructure opportunities through our competitive silicon photonics performance and proven manufacturing execution and advanced packaging capability. In silicon photonics, our data indicate that our PIC technology outperform peers in key parameters such as propagation loss, supported by our manufacturing high production yield now. Our 12-in PIC platform already support a 1.6T optical application, while our 8-in TFLN platform has completed 1.6T system validation with an industry-leading yield. Building on this achievement, our roadmap extend toward 300G per lane, and ultimately 400G per lane solution for next generation 3.2T applications, supported by hybrid bonding, TSV, and chip-to-wafer integrations. For the advanced packaging, we are making progress in two key solution areas. EDTC for power delivery and customized memory stacking. Our discrete eDTC solution for mobile will enter mass production this quarter. For AI and server application, our substrate embedded DTC solution has completed tape-out and functional verified. With our customized memory stacking solution has passed initial customer verification to address the growing bandwidth requirement of AI inference. All those achievements that demonstrate UMC's ability to translate technology innovation into the commercial execution now. By combining the device performance, manufacturing expertise, and close customer cooperation, we are expanding our role in next generation AI infrastructure, helping the customer to address the three critical requirements, including faster optical connectivity, more efficient power delivery, and high memory bandwidth. That's why we believe that we are in a great position to capture those opportunities.

Thank you. It does sound like your capabilities are pretty competitive. For both silicon photonics and also advanced packaging, say for eDTC or silicon capacitors, some of your peers are also building more supply. Should we be concerned about maybe potential oversupply, say, if AI demand slow down at one point? How should we think about the oversupply risk?

Well, we look at this on two fold, right? One is the demand side. Well, first, we are seeing a very strong demand driven by the AI-related application for silicon photonics and advanced packaging. In silicon photonics, we have been working with a leading customer. We recently achieved significant milestone as a multiple hyperscaler have qualified our 1.6T optical transceiver platform. Our 12-in PIC platform support the 1.6T application, and the 8-in TFLN completed 1.6T validation and already demonstrated manufacturing yield, with the roadmap that is going to the next generation. With all this, the end of advanced packaging, our discrete solution, like you said, and our first interposer with discrete DTC solution and for the AI and server application has completed tape-out with a key customer. The focus of the customer discussion already moved to that, from validation to customer demand planning. Our current expansion plan are tied to those customer commitments, in terms of demand. How about the oversupply? Based on what we see today, we do not believe we are building ahead of the market. Demand for silicon photonics and related AI infrastructure are growing faster than qualified capacity is being add up. We continue to monitor our peers expansion and timing of the new industry supply closely too. Overcapacity is an industry question, not just a UMC question. We look at how quickly end demand is growing relative to the capacity being added by overall industry. Based on the customer demand and industry capacity plan we see today, we believe the market can support the expansion we are planning. Meanwhile, UMC has been building the capacity in phases to mitigate risks. At the same time, we remain cautious and disciplined in how we add capacity. Most importantly, our capacity expansion is guided by customer commitment, like I said earlier, and tied with the customer product roadmap, production ramp, and capacity requirement. Even with all this expansion, we expect customer demand to exceed our planned capacity according to the customer, underscoring the strength of the commercial opportunity ahead, and our decision to accelerate the selected investment.

Got it. Sounds good. Sorry, last question, if I may. With the call a quarter ago, you quantified the AI related sales to be about $300 million for 2026, and you have a target to triple or more than triple to over $1 billion by 2029. Now with the accelerated expansion for both silicon photonics and advanced packaging, how should we think about your AI revenue by 2029?

Wow. Okay. First of all, you remember. Our AI related business spans a broad range of specialty semiconductor solutions, supports power management, connectivity, FPGAs, with our growing advanced packaging and silicon photonics. All those applications are becoming increasingly important contributors to our growth going forward. We are expanding our capacity like we reported, and investing in technology, manufacturing infrastructure to address all those growing demands. Our AI related revenue, it is close to $300 million in 2026, but looking ahead in three years, we expect our AI exposure will probably exceed a quarter of our revenue.

A quarter of UMC sales, maybe by 2029?

Yes.

Got it. Thank you very much. Sorry, let me get back to the queue.

Thank you.

Thank you. Next one, Laura Chen, Citi. Go ahead, please.

Yes. Hi, thank you for taking my questions. Good afternoon, gentlemen. Just a quick follow-up on the 10% increase on pricing. It is more like a blended ASP increasing or like-to-like increase into next year?

It is a blended, yeah.

Oh.

Yeah. It is more an incremental pricing improvement. It reflects both the pricing environment improved and our increase captured the other products, richer product, and a different product mix. Yes.

Okay, got it. Very clear. Also, we think we have a various different plan for AI related applications. Can you give us more color about that? How much the capacity we are planning for the silicon photonics, including in Taiwan and also mainly in Singapore? Also on the TFLN, we are ahead of our peers. Just wondering, can you also give us update about the capacity build up plan and in maybe two years, how big the capacities are we are looking for? Thank you.

Given the current milestone and the facility and tool schedules, we're doing all those expansion in phases, and we're going to continue updating you in terms of progress. Meanwhile, we're still working closely with our supplier. Hopefully, we can accelerate and expedite those installations. Meanwhile, as I mentioned earlier, if I look at the overall CapEx plan, which approved by our board, the capacity build up is approximately 18,000 for silicon photonics, 22,000 for DTC, which is silicon capacitor, and 20,000 for TFLN modulator. At this current stage, we project the initial install will happen in late 2028 and going into 2029 and reach the scale by 2030. For the location wise, we are focused on the silicon photonics in Singapore facility and the silicon capacitor in our Tainan facility, and the TFLN will be in our 8-in operation.

Sure. Thank you very much. Just also very quickly on the I node, probably we will talk more next year, but I think given the high utilization rate is already like 90%+, and also pricing into next year is more favorable. Are we seeing that any structural gross margin upside based on current visibility?

Certainly, the price increase likely to happen in the first quarter. We are also seeing some headwinds from the increase raw material cost, which we need to work hard to offset the impact. It's our view that the structural profitability remain intact. EBITDA margin, certainly, we will see in good shape, if we compare it to the previous cycle. But due to the stronger than expected CapEx, led by stronger customer demand. Depreciation will be higher and gross margin as a result, is unlikely to see near the previous high level. We will maintain a disciplined spending and try to capture the mega trend of the AI growth opportunity.

Okay. Thank you. Very helpful.

Thank you. Next one, Haas Liu, Bank of America. Go ahead, please.

Yes. Thanks, management team, for taking my question. I think just starting from price hike by 10% year-on-year, you clarified it is going to be on blended basis. Would you be able to share, on a like-for-like basis, how much is it going to be hiking, and how much of the depth of the price hike is related to the cost pass-through, and how much of that is because of the tighter supply-demand landscape?

Well, first of all, if we look at on a very high level, like I said, we expect 2027 will mark a turning point. The ASP trajectory will move in from historically erosion to uplift. We are talking to our customers and across the broad range of the applications. The conversation actually is more complicated. One is there are current contractual relationship, a portion of that. Secondly, a focus on different node and application and technology. So they are various. There is a mix of the pricing adjustment. Therefore, it is a lot easier to look at this from the blended basis. Instead going to every specific area because they are quite complex in terms of the mix. Yeah.

Because the reason why I am asking this question is because your utilization has been improving by 16% from fourth quarter last year to fourth quarter this year based on your guidance, but your pricing has been relatively stable. Just wondering why your customers suddenly are deciding to accept this price hike by 10% into 2027. If in the past couple of quarters, you have been striving to improve your utilization by 16%, but the blended pricing has been relatively more stable, but now you are seeing an inflection of that. Is it because of the product mix or cost pass-through that your customers decided to share the burden with you?

Well, both. The product mix is a part of it, and part of it is really reflecting the inflationary cost increase. As Chitung mentioned earlier, we are seeing some of the rising cost headwinds and also additional depreciation associated with our investment. While we are anticipating that, the overall customer demand are increasing and increasingly tight capacity on selected technology. So combination of those resulting in the conversation with the customer. I think the customer understand the overall industry tightness in terms of capacity and also understand the rising cost. So in a variable space. That is why I think this is naturally a conversation that we will have at this current point with the customer. We usually do this on a yearly basis. Again, we started this conversation in Q3, and with the alignment going into Q4, we are at a very late stage of those conversations. So we have a fairly confidence with the guidance that we just provided. Yeah.

That is very helpful. Second question is probably just on your current expectation for the customer and market demand in 2027. Because the supply chain seems to be suggesting that the unit demand across smartphones, TVs, and PCs could be down from the unit demand perspective. If that negatively affect your utilization, keeping at around low 90 percentage levels instead of further hiking to close to full levels, will your customers be able to request for some price discounts or pushback on the price hike through 2027? How would you deal with it?

First of all, let me explain about the utilization outlook. We are expecting the capacity utilization to remain at a very healthy level. This year, we improved from 85% in Q2 to above 90% in the second half of 2026 already, with several of our 8-in and 12-in fabs operating at a full loading at this current point. We are seeing the AI-driven upside demand, such as the power management and FPGA, which we cannot fulfill the demand today. The backlog PO is piling up. We will continue to optimize the production efficiency by debottlenecking the key steps and transferring the technology across different fabs to maximize the output. But operationally, there are minor mismatch between our demand and capacity configuration. So when you see less than 100%, that doesn't mean the fab is not fully loaded. It's really more of the mismatch between the demand and capacity configuration. We are working diligently to address this capacity configuration mismatch. We are upgrading our manufacturing capability to support our long-term product mix optimization as well. For instance, we are converting one of our 8-in fabs to expanding TFLN capacity. While we converting that, the loading will be lower. But doesn't mean that the facility is not being occupied and worked on. While we're strengthening our silicon photonics role and positioning UMC's future growth, all those are critical for our future growth. We are managing that through this activity. That said, the 90-some percent, it doesn't translate to a vacancy in terms of capacity. It's more of a mismatch situation. Yeah.

Okay. Yeah. Just a quick question before jumping back to the queue is, it is very encouraging to see that you approved additional CapEx to support your new businesses, including silicon photonics, TFLN, and silicon capacitor. But, would you be able to provide the timeframe on your spending through the next few years? Because last time you guided 2027 CapEx could be up to like $3 billion. But now, with the incremental CapEx, would you be able to just provide an update on that? Relatedly is, how should we think about the economies of scale of your new businesses, including TFLN, silicon capacitor, as well as silicon photonics, that by 2029, when these capacities start to ramp to a certain scale, would those business be, the gross margins for those business, would they be comparable to the corporate average or still going to be below because of the higher depreciation burden? Thanks.

Well, first of all, we can provide you the updated when the update is more clear to us. Like you know, we have board's approval, and now we are closely working with the supplier to make sure that we can accelerate and expedite the schedule. Once those information are concluded, we will be able to give you more concrete outlook in terms of milestone. Meanwhile, we are phasing in these different tech capacity in a different timeframe. But since we are still in motion of that association, we are probably not at the right time to provide you that. But on a high level, we feel very comfortable with all those demand, and only things we worry about now is we can really get them fast enough to our facility. Let us work on that, and when going into Q1 next year, we should give you a more clear picture with the next phase. Now, in terms of each of the economics scale and the benefits to our business. The silicon photonics will ramp in 2027, while the advanced packaging in 2028 because of facility reason. I think all those will be accretive to our current model because right now they are above our current corporate average. They are higher than our corporate average compared to existing solutions.

Okay. By the time these new business start to ramp in 2028-2029, those business should be carrying higher than corporate average margins, even with higher depreciation.

That is what we have modeled right now at this point, yeah.

Under the same circumstance for the same ramp, if you compare it to, say previously 22/28 nm, of course, the curve looks better.

Got it. That is very helpful. Thank you so much.

Thank you. As a reminder, please press star key and number one on your keypad if you would like to ask the question. Thank you. Next we will have Charlie Chan, Morgan Stanley, for questions. Go ahead, please.

Thank you. Hi, Jason, Chitung. Again, congratulations for a very strong execution result. So my first question is probably to Jason that, I had previously asked you about some concern about the non-AI part of demand offsetting the AI side, right? But realistically, do you see any kind of cuts recently from non-AI segments, and do you have the preliminary assumption for next year's PC, smartphone segment demand and your observation on those end markets and inventory? Thank you.

Well, right now we expect a positive demand momentum to sustain from 2026 into 2027, Q4 in 2027. On a high level, we are seeing the momentum will continue to sustain. And obviously, AI remains the key growth driver. So there is no worry on the AI side. Beyond AI, we continue to see opportunity from customer market share again as well, and increase the foundry share. That is for UMC, because we primarily serving the premier smartphone and automotive and industrial space, and we see healthy automotive industry recovery at this point. So I think we have less concern for automotive industry. However, on the smartphone, since we have less exposure in the mid and low-end smartphone, we think we have a less exposure in that space. While we do project the market outlook in that space is weaker than the others. Meanwhile, in near term, we also see some decline in computers because we see a soft demand in the computer AI, LCD controller, DDI, and Wi-Fi space. While the communication and automotive industrial and AI remains, not AI, the communication, the automotive industrial remains robust, we do see some weakness show up in the PC space. Yeah.

Got it. So those kind of warnings are all being considered. Yeah. Thank you. I am sorry, but I hope to clarify a little bit about blended ASP of 10% because I show you the 8-in 12-in, blended ASP and 12-in mix increase will bring up a lot of blended ASP, right? But I want to make sure, are you 8-in that you have a different product line that typical—

I am sorry.

—versus—

Can you break down a little bit?

Yeah.

Can you repeat that?

Oh, okay. Blended ASP increased by 10%. Is that the 12-in plus 8-in blended ASP or a different product mix? You are referring to the 8-in to be up 10%.

It is a company-wide, blended ASP as we reported. Although we did not disclose the number, but we do have an annual number for 2026. The 10% we are referring is for the whole company on the whole year, 2027. It is going to be about 10% year-over-year increase, including like for like price increase as well as the richer product mix.

Okay. Gotcha.

Together, yeah.

Okay. But you do not want to further break it down to, for example, if you remove the 12-in mix change, what would be kind of like [look to] like ASP increase? Would you, able to comment on that front?

It is a very complicated process, and some of the technology are totally different for different products. So it is very difficult to give you a generalized idea. But the 10% is probably the best indication for now.

Got it. Thanks. Last one is probably a kind of a very open question. Because we kind of think AI demand strength is no-brainer, right? I think Jason talked about we are not worried about AI demand. But the recent capital paid about AI processor demand, no matter year cloud financing, there is AR issue, right? I am not sure how visible is kind of UMC and this real kind of AI demand, especially for your product, if you have exposed to. How can you be sure is that still under supply? Just a kind of open question because capital market is a reaction, since do not believe that the AI demand will be so sustainable into next year.

I mean, the AI demand is not in general AI demand. So, when we talk about what we are serving, really in a selected technology and focus, for instance, on the optical space, I think the ramp-up just starting. So it could be a replacement mode. So to us, I think the silicon photonics is really a new space that we are entering, and as well as the industry is entering into. The PIC platform that we are offering, the TFLN, is all associated with that. While the industry is moving from the 1.6T, moving into 1.6T, and for the future 3.2T, we actually think this demands will be continuing robust. The advanced packaging is really a solution that we talk about the DTC, deep trench capacitor solution. A capacity solution, right? And they are more of a device that can be implemented in multiple different areas. The power integration, all those power will continue driving the industry going forward. So I think with all those solutions, I think the market will be continuous. The demand will continue to be robust. Yeah.

Got you. So even, like AI processor slow down, there is a kind of a package migration, like 1.6T to 3.2T, et cetera. Those will continue. So you are not that worried about sort of overall AI demand expansion pace.

Yeah. Because if you just look at it, truly what drives that demand is the compute and also power and—

Right.

—the speed, right? Really, I mean, the optical speed requirement and the power integration requirement is more of a forward-looking type of applications. We're not talking about the AI demand in general. We're talking about technology migration to support the future growth—

Yeah.

—and meeting the system requirement.

Understood. Very clear. Thank you. Yeah, I also agree that, DTC, the deep trench capacitor is a very important notion trend. Thank you, Jason. Thank you, Chitung. That's super helpful. Thanks.

Thank you.

Next one, Gokul Hariharan, JPMorgan. Go ahead, please.

Hi. Thanks, Jason and Chitung, for taking my questions. First of all, just want to clarify, Jason, you mentioned long-term AI revenues can be 25%. Was it 25% of current revenue, roughly $9 billion, or you think it will grow to 25% of revenues, in 2030, which probably a much bigger number? Just wanted to clarify that.

I'm referring to the revenue in next three years, the three years from today. Yeah.

Okay. So in three years it can be 25% of total revenue.

Yes, right.

Okay. Understood. Thank you very much. Could you talk a little bit about any indication on revenue contribution from silicon photonics and the embedded deep trench capacitor for next year, given those seems to be ramping earlier? Also on silicon photonics, any aspirations you have, in terms of where you want your market positioning to be, given you are starting off later than competition, but you also have some unique advantages here?

We are currently running at full capacity, so, we said this year will be somewhere at the $300 million range. For the new add, the mass majority of the capacity will start coming in line in late 2028. So 2027 is still going to be under the constraint. So, while we try to optimize, like I said, we try to reconfiguring some of our facility. Hopefully, we can optimize some to shorten, but we are under the constraint for 2027. So, we would probably be able to share some light in the next quarter. But again, it is not going to be significant, because the capacity is not available yet. So that is the revenue for 2027. What was your second question? I am sorry.

Yeah. Secondly, I think, what is your aspiration from a silicon photonics market share perspective, given you are starting a little bit later than your competitors, but you also are migrating straight to 12-in. You are already starting to see some at least test results which are comparable or better than your competition. Just wanted to highlight and think about where we could end up in silicon photonics market share.

Of course, yes. The aspirations are very high, and we have very high confidence in terms of decision and moving into a 12-in directly, without going through the 8-in. Right now, I think the decision actually did pay off. We have demonstrated with our 12-in PIC technology. We already demonstrated better performance in many of the key parameters. Our customer and their end customer. Right now the product is being validated and qualified by the end customer as well. With the production yields get to the manufacturer level, we think our 12-in PIC platform is ready to support the customer, for their current 1.6T optical application and going into the next generation as well. The TFLN is a little bit different situation. I think we much ahead about anyone else in the TFLN development.

Yes.

I think the customer also see a benefit between the integration of the putting TFLN modulator with our 12-in PIC solution. Because that, our advanced packaging preparation is also bear fruit now because with the hybrid bonding, the chip-to-wafer capability, I think those actually will help us to even become even more competitive because combination of the silicon photonics solution along with the packaging capability, that is putting us in a unique position. We feel very competitive in those space. Yeah.

Okay. Just to follow up on the TFLN engagement, Jason, is this mainly going to be HyperLight, whom you are going to be supporting, or are there a broader set of customers that you already have got beyond HyperLight, which I think you already invested in them and publicly announced as well?

Yeah. We announced the HyperLight partnership because HyperLight was a very early adopter of the TFLN. I think we are very pleased with that program. Because we already demonstrated with HyperLight, we have seen a very strong interest in the TFLN solution now, not just from HyperLight, from many different customers in the space. Yeah.

Okay. Maybe on 2027, given your capacity constraint, any read on how much capacity can you still squeeze out on an overall basis, let us say for 12-in? Can we do mid-single digit capacity growth next year, or is it still low single digit potentially next year?

Well, I guess my reaction is just not enough. I think we want to improve that number any way we can. Right now, the best way to explain that is it is not enough.

I think you are right. Very minimum. Yeah.

Okay. Chitung, just wanted to also lastly examine the gross margin puts and takes. Obviously, price increase limit ASP of 10% or thereabout is quite positive. If you carry the utilization gains from Q4 into all of next year, that also is positive. Could you give us some indication on how depreciation is going to shape up? I think last quarter you mentioned depreciation will be moving higher, but it looks like the pace might be a little bit more. Are we saying that we are largely going to be in the similar gross margin range, or can we still squeeze out some gross margin expansion into next year, even with heavier investment coming through?

Depreciation is going to be certainly higher and maybe a couple of years in a row. We are talking about low single digit depreciation growth, at least for 2027, if not more for 2028. For gross margin, this is going to be a big number to offset. On top of that, we also have rising raw material costs, et cetera. Like I mentioned earlier, we feel more confident to have a healthy EBITDA margin structure. But for gross margins for next two years at least, depreciation, along with the rising raw material cost, is going to put pressures on our overall financial gross margin.

I also like to echo what Chitung just said. We are expecting ASP improvement, and we are expecting higher utilization, and all those will provide support to our profitability for 2027. At the same time, we are anticipating a rising cost structure and additional depreciation, which associate with our investment decision. Our priority now is to maintain a healthy profitability profile while continue invest in different differentiated technology and support our long-term growth. So that will give you some color on how we view about this gross margin question. Yeah.

Got it. Thank you. Thank you very much. Yeah.

Thank you. Ladies and gentlemen, we thank you for all your questions. That concludes today's Q&A session. Now I will turn things over to UMC Head of IR for closing remarks. Thank you.

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.

Thank you. Ladies and gentlemen, that concludes our conference for 3Q 2026. Thank you for your participation in UMC's conference. There will be a webcast replay within two hours. Please visit www.umc.com under the Investors Events section. You may now disconnect. Thank you again. Goodbye.