ISC Co., Ltd. (KOSDAQ:095340)
127,100
-19,800 (-13.48%)
Nov 20, 2026, 3:30 PM KST
โ View all transcripts
Earnings Call: Q2 2026
Sep 7, 2026
Summary
Q2 revenue and operating profit grew strongly YoY, led by AI, while one-off legacy memory mix and costs weighed on sequential margins. Q3 growth and margin recovery are expected as data-center and flagship mobile products ramp.
Good morning. This is the IR team of ISC. We would like to sincerely thank all analysts and shareholders for joining our 2026 second quarter earnings conference call. Today's presentation is based on our official disclosures and internal management data. We will begin with approximately 15 minutes of prepared remarks, followed by a Q&A session. Before we begin, please note that today's discussion may include forward-looking statements, and these are based on management's current expectations and assumptions. Actual results may differ due to a variety of factors, including market conditions, tech developments, and customer demand. During today's call, we will review our second quarter 2026 performance and provide our outlook for the third quarter. Presentation materials will be available on our IR page after the call in both Korean and English, along with a replay of audio.
In addition, starting this quarter, simultaneous interpretations provided for our international investors. Let me first walk you through our second quarter results. In the second quarter of 2026, ISC recorded revenue of KRW 72.9 billion and operating profit of KRW 21.3 billion. Revenue grew 41%, operating profit 55% YoY, with an OP margin of 29.2%. Compared with the previous quarter, revenue rose 7% while operating profit declined slightly. The QoQ decline in operating profit was not due to any slowdown in demand for AI data center products. It resulted from a temporary shift in product mix as one-off sales of relatively low-margin legacy memory products increased. Despite this temporary change in product mix, our underlying profitability centered on high-value AI data center products remained solid. In particular, second quarter AI revenue reached KRW 59.1 billion, up 71% YoY, and AI's share of total revenue rose to 81%.
We would like to reiterate that in the second quarter, on the back of expanding AI demand, ISC's revenue scale and business structure moved up to a new level. Next, let's go to AI revenue. Second quarter AI revenue was KRW 59.1 billion, up 71% from KRW 34.5 billion a year earlier. AI's share of total revenue also rose 14 percentage points from 76% a year ago to 81%. These figures show that ISC's business structure is shifting rapidly from a traditional memory and mobile center model towards AI compute data center, AI memory, and on-device AI. Let me break this down by segment. AI compute data center revenue was KRW 47.7 billion, up 59% YoY.
Our product portfolio expanded from a GPU-centric lineup to a server, CPU, and ASIC, and our growth was driven by the volume production of major customers' next-generation products and by our rising supply share. AI memory revenue was KRW 5 billion. Although this was down from the previous quarter, we continued to steadily expand the development of high-end AI memory test solutions such as HBM SOCAMM2, as well as our supply base with customers. On-device AI revenue was KRW 5.4 billion, up 20% YoY and 69% QoQ. This reflects growing demand for final test and system-level test sockets as premium smartphone APs integrate CPU, GPU, and NPU more tightly, and on-device AI features expand. Physical AI also generated new revenue of KRW 1 billion.
Although still at an early stage, we see it as a mid-to-long-term growth opportunity as AI applications broaden into areas such as robotics and edge AI. By contrast, non-AI revenue was KRW 13.8 billion, down 20% year-over-year. Next, the breakdown between memory and non-memory revenue. Second quarter non-memory revenue was KRW 56.9 billion, up 29% YoY and 7% QoQ. It accounted for 78% of total revenue, with growth led by AI data center GPUs, server CPUs, and ASICs, as well as premium mobile APs. Memory revenue was KRW 16 billion, up 176% YoY and 5% QoQ. Memory share of total revenue rose from 11% a year ago to 22%. However, this increase in memory revenue includes one-off sales of relatively low-margin legacy products. While it contributed to revenue growth, it weighed somewhat on the overall product mix and operating margin.
What matters is that non-memory revenue continues to grow in absolute terms, and there has been no change in the underlying trends of AI data center demand and our rising share within major customers. We therefore view the second quarter change in profitability as a result of a temporary product mix, not a slowdown in AI demand or weakening of our competitiveness. Going forward, our memory business will improve the pricing terms of legacy products while expanding the proportion of high-end AI memory products such as HBM and SOCAMM2 to enhance profitability. Now let me walk through the results by application. First, the data center business. In the second quarter, expansion of AI compute applications drove company-wide earnings growth. Next-generation AI accelerators are rapidly advancing towards higher speed and higher power, and chip complexity is rising as pin counts increase and package sizes grow larger.
As a result, demand is increasing for high-specification test sockets that can precisely verify high-speed signals, power stability, and thermal characteristics. ISC has expanded its supply scope from a GPU focus to server CPUs and ASICs, and our share within major customers continues to grow steadily. This means more than simply benefiting from AI market growth. It reflects that ISC's technology, quality, and volume production capabilities are being recognized within our customers. Next, the mobile business. In the second quarter, supply of volume production test sockets expanded, centered on premium APs with high integration of CPU, GPU, and NPU. As on-device AI features strengthen, premium AP architectures are becoming more complex and test conditions are more demanding. Accordingly, demand is rising for high-specification final test sockets and for SLT solutions, which verify chip performance under conditions close to real-world use.
ISC expanded its supply of volume production sockets for premium APs and also secured orders for WMCM-based SLT sockets for a flagship smartphone AP. This demonstrates that our mobile product portfolio is expanding from final test to SLT. In the memory business, the increase in one-off legacy product sales and the expansion of our high-end AI memory business base progressed simultaneously. In the second quarter, product mix was temporarily adjusted by one-off sales of legacy memory products, while the development of supply of high-end AI memory test solutions such as HBM and SOCAMM2 are also gaining full momentum. Beyond supplying sockets as standalone products, ISC is broadening its business into integrated test solutions that combine testers and sockets. We're also continuously expanding our supply of products and applied processes within the three global memory makers. Now let's go to the 3Q outlook.
In the third quarter, we expect growth to continue, led by the data center and mobile businesses, while the memory business improves its product mix centered on high-end products. In the data center business, we expect growth to be driven by expanding volume production orders for AI GPUs, server CPUs, and ASICs. In addition, initial supply of a new ASIC product for a North American hyperscaler customer is scheduled. Our share within major customers is rising, and as the proportion of high-speed, high-power products grows, we expect the revenue share of high-value test sockets to increase as well. Next-generation AI semiconductors require higher socket specifications than before, driven by higher pin counts, larger package sizes, and greater power consumption. We expect this to translate into higher proportion of high-value products and rising average selling prices.
Here, ASP improvement is not a simple price increase, but the result of an improved product mix driven by higher product specifications and technical difficulty. Our mid-to-long-term customer base is also set to strengthen. ISC has newly secured a global data center infrastructure customer and is preparing to supply ISC Link H, a high-speed connectivity socket mounted on AI server racks. This product line is a next-generation solution that supports the verification of high-speed connection performance and reliability of the core modules within AI servers, server CPUs, GPUs, ASICs, and memory, network, and storage. Through this, ISC plans to extend its data center business beyond conventional semiconductor test sockets to products that support high-speed interfaces inside AI servers. In addition, by establishing dedicated production lines for major customers, we will build a foundation for stable supply to our global customers.
In the mobile business, we expect the launch of a new flagship smartphone by a North American mobile customer to be fully reflected. As volume production orders for premium APs increase, we expect final test and SLT socket revenue to expand together. With stronger on-device AI features, premium APs are integrating CPU, GPU, and NPU more tightly. As chip complexity and test difficulty rise, we expect greater adoption of high-specification sockets, leading to improvements in ASP and profitability. Supply of R&D test sockets for an on-device AI is also increasing. Supply at the R&D stage is important because it will allow us to participate in customers' next-generation product development and secure the foundation that can lead to future volume production socket orders. In the memory business, we will improve our product mix by expanding the proportion of high-end products.
We expect revenue from high-end AI memory products such as HBM and SOCAMM2 to grow, and our supply scope within global memory customers to expand. For legacy memory products, we will improve pricing terms by customer and by product to raise average selling prices and profitability. In the equipment and materials business, we expect related revenue opportunities to expand as a major customer's investment in the Yongin Semiconductor Cluster gains full momentum. Initially, revenue will come mainly from the supply of test equipment and process materials, and business opportunities are expected to grow in stages, such as production lines are built and operations ramp up. Next, our capacity expansion plan for mid-to-long-term growth. Our current annual production capacity is approximately 2 million units. We plan to expand capacity in stages, in line with customer demand and the approval and volume production schedules of our key products.
Our target is to expand annual capacity to 2.7 million units in 2027 and 3.7 million units in 2028, reaching up to 5 million units in 2029. The corresponding revenue contribution begins in 2027. However, it will be reflected in stages depending on customer product approvals, volume production schedules, and the utilization rates of the new production lines. This is precisely why we are expanding capacity in stages rather than all at once. By concentrating investment on products and customer projects with confirmed demand, we will manage the risk of excess capacity and improve investment efficiency. In the second half of 2026, we plan to complete the expansion of Vietnam Fab 1 and begin full-scale operation.
The Songdo Fab and Vietnam Fab 2 will enter full operation in 2027, after which we will sequentially expand the capacity of the Songdo Fab and Vietnam Fab 2 in line with customer demand and volume production schedules. At the Songdo Fab, we will apply manufacturing AI and automated process to raise productivity and quality management, building a volume production system capable of serving major customers and high-value products. The essence of this capacity expansion lies not simply in increasing the number of units we can produce. First, it is to meet growing demand for high-value products such as AI GPUs and server CPUs, ASICs, and premium APs. Second, it is to translate our rising volume production share with the major customers into actual revenue growth. Third, it is to improve productivity, quality, and cost competitiveness together with capacity expansion through automation and process optimization.
By concentrating investment on areas with confirmed demand, we will manage the risk of excess capacity while proactively securing a production base that can convert future customer demand into actual revenue and profit. Alongside capacity expansion, ISC is also strengthening its technology and cost competitiveness. First, we plan to apply our recently developed new material for test sockets to volume production processes, reducing raw material and manufacturing costs while improving the high-speed signal characteristics and durability of our products. We expect the broader adoption of this new material to contribute not only to cost reduction, but also to product performance, differentiation, and improved profitability. In addition, starting with the supply of next-generation high-speed connectivity sockets for servers, we are expanding our data center product portfolio. AI servers are evolving into architectures that combine not only CPUs, GPUs, and ASICs, but also various high-speed modules such as memory, network, and storage.
In response to these changes in server architecture, ISC will broaden its business scope from semiconductor test sockets to high-speed interface products for servers. ISC will continue to invest in production capacity and technological competitiveness to capture growth opportunities in the AI semiconductor market. At the same time, we will maintain financial soundness so that we can safely secure the funding for our mid-to-long-term investment. We believe growth investment and shareholder returns are not conflicting concepts, but tasks that must be pursued together to sustainably enhance corporate value. The company is reviewing a mid-to-long-term policy to expand shareholder returns in stages while maintaining financial soundness, and we will share our specific growth investment plans and shareholder return direction with the market during the second half of this year. Let me summarize.
In the second quarter, ISC recorded revenue of KRW 72.9 billion and operating profit of KRW 21.3 billion, growing 41% and 55% YoY, respectively. AI revenue reached KRW 59.1 billion, up 71% YoY, and AI share of total revenue rose to 81%. The QoQ change in profitability stemmed from a temporary shift in product mix as one-off sales of low-value legacy memory products increased. Excluding this effect, our underlying profitability centered on high-value AI data center products remained solid. in 3Q, we expect to sustain our growth momentum on the back of expanding volume production of AI GPUs, server CPUs, and ASICs, the launch of a new flagship smartphone, and growing HBM SOCAMM2 revenue. Over the mid-to-long term, we plan to gradually expand our annual production capacity from the current level of 2 million units up to 5 million units by 2029.
We will execute investment in stages based on customer product approvals, volume production schedules, and actual demand, converting our expanded capacity into revenue and profit. Finally, ISC's growth does not rely on expansion of AI market alone. Our applications are broadening from GPUs to server CPUs and ASICs. Our share within major customers is rising. Our ASPs are improving as a proportion of high-spec product growth. Adding phased capacity expansion to this will connect the business opportunities we have secured today to mid-to-long-term revenue growth. ISC will balance growth and capital efficiency, building sustainable growth in the AI semiconductors test market. This concludes our earnings call for the second quarter of 2026. We will now move on to the Q&A session. Please feel free to ask any questions. Thank you. We will now begin a Q&A session. If you would like to ask a question, please press asterisk and number 1.
on your phone. If you would like to cancel raising your hand, please push asterisk and number 2. The first questioner is Oh Seong-hyeok from Citi Global Securities. Please go ahead. Hello. Your AI share is up and your performance is great. Congratulations. I have three questions. Number 1, Q3 OP margin guidance is what I would like to understand. Number 2 is, the second half, you will have a better understanding about which sector is going to grow more, which application is going to grow. Number 3, for second half data center, I know that your revenue is great. What about the growth rate? Do you think this is going to be positive too? Let me answer your questions in order. Number 1, related to revenue and OP margin of Q3. Right now, we have to look at the orders that we got and their production schedule.
I believe our revenue and OP margin will grow significantly compared to Q2, especially mobile flagship AP, North American smartphone is about to be released, and data center, next-generation GPU, CPU, ASIC mass production is scheduled. Also, Q2's low margin product mix is going to go away. As a result, our profitability is going to grow. I cannot share specific numbers, but I believe Q3 OP margin is going to be at the higher 30s. I believe both revenue and profitability are going to be captured by ISC in Q3. I believe the biggest product that is going to drive growth of Q3 is mobile AP and non-memory chips. Especially for mobile AP, I believe that WMCM-based new flagship smartphone AP mass production is what we have secured recently. High value-added product is going to impact us in a positive way.
For data center, CPU or GPU are something that we are going to supply additionally, and these products are going to also positively influence us. For CPUs, our market share in major customers is almost 100%. With this stable market share, we're going to continue to grow. For GPU, new products are coming up, so high value-added products are going to drive our growth too. For second half data center growth, revenue growth, I think this is going to be higher than the first half of this year. First half was more about R&D and validation process, whereas the second half is going to be the time where these products are going to be manufactured in large scale. So data center-related revenue is going to grow significantly.
As we have shared earlier, ISC Link H, which is connectivity socket, is going to be recognized as revenue for the first time in history. So, I believe relevant products are going to be sold in bigger volume. However, because when revenue is going to be recognized can differ depending on the schedule for clients, so please bear with us. Let's go for the next question. Next is Kim Min-kyung from Hana Securities. Please go ahead. Hello, my name is Kim Min-kyung from Hana Securities. I would like to ask a few questions too. Q2 profitability went down because of the lower margin legacy products, but can you share more details about it? Second, you said profitability is going to be recovered in Q3. What's the background for that? Next-generation interface test sockets are something that you mentioned, but what's the application for this next-generation test sockets?
LTA and designated line construction, how are these going to? Sorry, just to ask more questions. Once the capacities are added, what's going to be the surplus or the newly added capacity? I know you mentioned it, but can you share with us in terms of gradual stage-based added capacity volume? Well, first and foremost, talk about the OP margin of Q2. Well, if you look at revenue, it went up. Revenue went up by 7%, but our profitability went down by 9.7%. So, you can see that the OP margin went down from 35% to 29%, which is about 6%, and the major reason for that is because product mix, low-margin product mix and the proportion went up. Secondly, one-off expense
was incurred, and that was around KRW 3 billion because of the mass production ramp-up cost was something that we had to spend, which was KRW 2 billion. We also had to provide some of the performance-related incentives, and we built a reserve for incentives, which was KRW 1 billion. So if you include that one-off expenses, I would say our profitability was around 36%, but this was recognized in Q2. Because we recognized it in Q2, Q3 will be better. Just to talk about the recovery to 30% level from Q3, the reasons are, number one, North American smartphone AP mass production, GPU and CPU mass production, and ASIC mass production. So, these three elements are going to affect both revenue and profitability because those are high value-added, high-margin product. I believe our OP margin is going to stay around 35%-40%.
However, as I have mentioned related to quarterly OP margin, it will differ depending on mass production schedule, or when the revenue is going to be recognized. To talk about ISC Link Edge, as you have asked, I think our clients would call this a connectivity socket. Unlike testing the individual semiconductor chips, like conventional test sockets, this connectivity socket validates the connection performance between different modules that are on AI server racks, and the safety of it. It checks the interoperability between different chips, and safety and stability of the connections between different parts. As you know, at servers, there is CPU, GPU, ASIC memories, network, and different chips are connected, which are all high-speed modules. High-speed signal transfer, contact, stability, interface, performance are checked by the connectivity sockets.
This product is now provided to one particular client we have, but I believe that this product can also be sold to other clients that are in charge of AI servers and data centers. Once qual tests are over with other clients, I believe next year, we will be able to see this as a big growth driver. Number three, you talked about LTA and the designated line for different customers. Because of the signing of NDA with major clients, we cannot share everything with you. As I have said, we are going to supply connectivity sockets to one major client, and the initial revenue is going to take place from the second half, and this will be a big part of our revenue going forward.
Our major clients, both memory and non-memory, CPU, GPU, ASIC clients will now have designated production lines, and with them, we get initial payment from them to cover the cost of designation. Revenue visibility is something that we are expecting. From 2027, I believe LTA clients are going to take up 50% or above from our total revenue mix, but it could go even higher. Our growth, speed, and profitability are going to benefit from the designation of major client lines. As I shared from previous slides, when it comes to capacity, from sockets perspective, every year we produce around 2 million sockets. Because we are adding new production capacity, this number is going to grow, and we are going to add more equipment to the site gradually. 2029, we will have a production capacity of producing more than 5 million units of sockets.
Once we are done with CapEx investment, by 2027, we will produce 2.7 million, and by 2028, 3.7 million, and 2029, 5 million. Basically, we are going to have 3x production capacity than the current level. This is not a one-time addition. We will gradually and increasingly add production capacity as we talk to our clients about their actual demand and order volumes. Right now, we are at the 100% of utilization rate. Maintenance, quality assurance, and production transfer time is what we consider. I wouldn't say we have a lot of room to produce more, especially AI and data center products. Per client, spec request is different and quality standards are different. That is why we believe it is best to have a designated process and designated production line for different clients. That is why we have to rush to invest in capacity addition.
I would say that the capacity addition is not to complement our capabilities, but also I would say it's an investment to proactively respond to CPU, GPU, ASIC, and next-generation product needs. Now let's move on to the next question. The next question is Henry Jeongmin from Bio Capital Market. Please go ahead. Hello. Thank you very much for giving us this opportunity to ask a question. First and foremost, because GPU is the biggest portion that you have, could you give us a specific guidance for GPU in the second half? According to what I heard earlier, you said GPU is not going to grow that much compared to the first half. Why are you so conservative in your projections? Next, you talked about Songdo fab, and you're going to add some automation and AI features in the production.
Well, when I went to your site earlier, I was able to see some of the human laborers. Compared to your existing site, I want to ask you what the automation level is going to be like in the new fab. Will we be able to check the change? My third question is quite simple. For the last few years, corporate ASP price has been growing at which percentage point level? Do you believe that ASP is continued to grow? I want to know the trajectory of your ASP growth so far. Well, let me answer your questions. First, if you look at our first half GPU volume, around KRW 66 billion is our revenue level, and because our target is KRW 130 billion-KRW 140 billion, I believe it's going to be around the same. I wouldn't say I'm conservative about GPU growth.
When we give guidance every month to the market, we have to look at mass production schedule of our clients, and new products seem to go through some delay when it comes to mass production. Please consider that we're seeing both our schedule and our production capacity. This year, we had the issue of capacity. That is why we're rushing into CapEx investment to add more capacity, but it takes time, and we need some time to actually ramp up our production. Now, given that, I would say that GPU, CPU, you may think that we're being conservative, but it has to do with our production capacity level. Nextly, related to process, maybe if you recall your tour to Vietnam, we have a lot of workers for both pogo and rubber sockets. It's not just assembly.
We have pre-process, which is processing materials, and then the post-process, which is the packaging, and we have a lot of workers for validation and inspection. Our raw material, mixing, and packaging, assembly, these are where workers are put in. I would say the emphasis that we're putting in for Songdo fab is for pre-process and post-process, we would like to use AI to make sure that we use minimum number of human workers. Let's say we used to need 100 workers for the production line. I would say assembly requires 50, and the rest will be 50 for both covering pre and post processes. But if we add AI and if we stabilize this, I would say that we would only require one-tenth of the current level, 10% of the current labor force. This is the result of our simulation.
I believe that besides socket assembly, almost 100% of automation will apply. For assembly process, it is difficult to apply automation because individual chip has different pin counts, and pins pitch or structure, layout, those are all very different. Assembly will require some people, but besides assembly, the other process will use 100% automation and AI. If you look at corporate ASP trends so far, between 2016 to 2020, memory share was higher, so ASP growth rate was only a single digit, less than 10%, I would say. Depending on our clients' business, the growth rate would go up, sometimes go down. From 2020 to now, if you divide our product into memory and non-memory, memory has been just staying at a single-digit growth rate, but non-memory, I would say on average, it has been growing by 20% on average every year.
Based on that, in, I would say, 2020 and on, our growth was mainly led by non-memory sector. Memory side profitability, how we can drive that using HBM and SOCAMM2, is something that we have to look at. High-end memory profitability drive is something that we're expecting, and probably memory ASP will also be better with this new trend. Now let's move on to the next question. No one has asked for a question. If you would like to ask a question, please push asterisk and number one on your phone. Let me have your attention once again. There is no person who has asked to raise a question. I'm going to give some more time to wait for a question. Next is Oh Sun Young from Citi Global Securities. Please go ahead. I would like to ask two more questions.
Number one, you talked about North American hyperscalers, first delivery of the product is scheduled. Could you give us more details? Which product are you thinking of? Number two, I think seasonality is something that we have to consider. What do you think about the seasonality affecting our performance in the second half? Let me give you the answer for the first question. Hyperscalers, we had some major clients, but we added new clients. Data center cloud, our related business is what this new client is working on. We have launched a new product recently, but for mass production ramp-up, we will have to follow the investment schedule of this new client. From last Q3, I don't believe that we have seasonality issues anymore.
As you know, Q1, we grew, and Q3, Q4, I believe that both revenue and OP margin are going to grow as well, because we don't have large portion of memory. Also, if you only look at the memory too, there's no such thing as cycles anymore. For non-memory, we're doing well in both R&D and mass production. I would like to declare right now that there's no seasonality issues impacting us anymore. Thank you. We have Kim Dong Kwan from Meritz Securities. Please go ahead. Hello, I'm Kim Dong Kwan from Meritz Securities. I know that many had asked questions already, but I would like to ask two additional questions for you. First, data center ASIC business is what I would like to ask about.
In the previous quarter, there was new ASIC clients mass production project was mentioned, and you also shared some of the major businesses of your ASIC client, and they are doing well in terms of their business. Can you share your projection and prospect for your ASIC supply and delivery? Thank you so much for your question. Let me answer you. Related to ASIC, since last year up until now, I would say, one client took up almost the lion's share of our revenue, and cloud and data center investments were made by them, and we have been supplying according to their schedule. But unlike our expectation, at least, or if you look at our first half, it is similar to last year's level, but from the second half, it is going to go up. The revenue is going to go up.
Last year, we did KRW 33 billion for ASIC, but this year, if you look at the trend, we should be able to have KRW 50 billion of cumulative revenue for ASIC. Some clients are developing and launching some products, but up until the moment where they can start mass production, it will take some time. So, timeline-wise, I think it is taking more time till mass production, even though they launched new products. I would say KRW 50 billion of revenue from ASIC is what we are expecting, and we are diversifying our client portfolio. We are talking to different clients in ASIC too. Probably in the second half, we will share great results from ASIC too, both Q3 and Q4. Now let us move on to the next question. We will receive questions later separately through emails and phone calls. Thank you very much for your attention so far.
This concludes Q2 earnings call of ISC. Thank you.