Sk Hynix Inc S/Gdr 144A

Good morning, everyone. This is CFO Kim Woo Hyun. Allow me to first introduce the company's performance for the fourth quarter of 2024. Despite a delay in the recovery of demand for consumer products like PCs and smartphones, the demand for AI memory products and data centers where investments are rapidly growing drove overall demand. As a result, we significantly increased the sales of AI memory products, including HBM and enterprise SSDs, and achieved record high quarterly revenue of 19.77 trillion won for the quarter, up 12% from the previous quarter and 75% compared to the same period last year. In DRAM, Sales increase of high-value added products, such as HBM3E and Server DDR5, drove a mid-single-digit percent sequential increase in BIT shipments. Despite price decline in legacy products like DDR4 and LPDDR4, blended ASP rose by approximately 10% due to product mix effect. HBM saw another quarter of strong growth. accounting for over 40% of DRAM sales in the Q4, and shipments of HBM3E 12 high products also began as planned during the quarter. In NAND, although sales of enterprise SSDs increased, inventory adjustments by PC and mobile customers led to a reduction in purchasing demand, resulting in a mid-single-digit percent decrease in both shipments and ASB quarter-on-quarter. Sales contribution from high value-added products, such as HBM and enterprise SSDs, increased significantly, enabling us to maintain relatively stable revenue and profits, even during market fluctuations, and reflect our improved business structure. As a result, operating income in Q4 increased by $1.05 trillion to $8.08 trillion, and operating margin rose by 1% point to 41%, again marking record high operating profit following the previous quarter. Depreciation and amortization for Q4 increased slightly from last quarter to 3.17 trillion won, with EBITDA reaching 11.25 trillion won and EBITDA margin of 57%. Non-operating gain net of expenses reached 1.5 trillion won. Notable items include foreign currency-related net gain of 0.62 trillion won due to strong dollar against won, and other non-operating net gain of 0.87 trillion won, which contains gain from the partial sale of stake in system IC Ushi. As a result, net income before tax was 9.58 trillion won and net profit for the period was 8.01 trillion won with net profit margin of 41%. Now moving on to the annual performance for 2024. 2024 was marked by a significant surge in AI memory demand driven by data centers. thanks to our leadership in AI memory technology and our profit-oriented business decisions, consolidated revenue for 2024 reached 66.2 trillion won, surpassing the record high revenue achieved in 2022. Operating income also exceeded the performance of past super cycle period in 2018, reaching 23.5 trillion won. As AI memory demand grows, the memory industry is transitioning from a commodity market driven by volume and price to a customized market focused on high performance and high quality products. 2024 results were especially meaningful in that we saw the possibility of memory companies generating stable profits by securing the ability to meet customers' high expectations, and support their needs on time. Following our achievement of supplying industry's first HBM3E high products in March last year, we were also the first in the industry to support HBM3E 12 high products in fourth quarter, reinforcing our position as the leading HBM player. Backed by a timely supply and sales expansion of HBM3E8 and 12 high products, annual HBM sales grew more than 4.5 times compared to last year, contributing significantly to our record-breaking DRAM performance. Additionally, we rapidly expanded sales of DDR5 and LPDDR5 products. which offer competitive performance and power efficiency, strengthening the profitability of our conventional DRAM products. Despite limited annual shipment growth in NAND, sales of higher priced enterprise SSDs increased by more than 300% year on year, leading to not only a return to profitability for NAND in 2024, but also record annual revenue and operating income. We also secured new demand areas by proactively developing high density and high performance storage solutions required for the AI infrastructure. At the end of last year, we launched 61 terabyte and 122 terabyte QLC-based enterprise SSD products further enhancing our high-density enterprise SSD lineup. As of year end, our consolidated cash and cash equivalents, including short-term investments, amounted to $14.2 trillion, up $5.2 trillion from end of last year's levels. Interest-bearing debt decreased by $6.8 trillion from the previous year to $22.7 trillion, and as a result, Our debt to equity ratio and net debt to equity ratio improved meaningfully, reaching 31% and 12% respectively. Due to the substantial profit increase in 2024, our annual free cash flow, which is cash flow from operating activities, less acquisition of property, plant and equipment, amounted to 13.9 trillion won. Following our former dividend policy, the Board of Directors confirmed total fourth quarter dividend of 1,305 won per share, which is 301 per share as fixed dividend, and an additional dividend of 1,005 won per share, which is equivalent to 5% of 2024's annual free cash flow per share. In accordance with the amended articles of incorporation the dividend record date for the last fiscal year will be set of February 28 2025 and the actual dividend amount may be adjusted based on changes in the number of Treasury shares until the dividend record date. Next is the company's market outlook and plan. 2025 memory demand outlook. is clouded by inventory adjustments from PC and smartphone OEMs, as well as strengthened protective trade policies and geopolitical risks. Nevertheless, supply-demand conditions are expected to gradually improve towards the second half of the year as IT device demand improves and inventories are cleared. The demand-growth momentum for AI memory products that began in 2023 is expected to continue to be the driving force behind this year's market growth. With the full fledged beginning of the AI era, memory market demand is moving towards higher performance products from low priced products. As a result, polarization of memory market demand is expected to become more severe. PC shipments this year are expected to grow by low to mid single digit percent, driven by corporate PC replacement demand from end of support for Windows 10, as well as effects from AI PC shipments kicking in. AI PC penetration is expected to reach around 30% to 40%, leading to an increase in adoption of not only 16 GB DRAM, which is known to be the minimum required DRAM density for AI PC, but also 24 GB and 32 GB densities. The smartphone market is expected to grow by a low to mid single-digit percent, similar to last year's level. The penetration of AI-enabled smartphones is expected to grow to approximately 30%. driving demand for higher contents as well as higher performance mobile memory like LPDDR5X and LPDDR5T. Furthermore, the inclusion of smartphones and tablets to China's consumer subsidies is also expected to drive positive developments on set demand this year. Strong growth of server market demand is expected to continue, driven by big tech companies' ongoing competitive investments to secure advanced AI training and inference capabilities. Recently, some big tech firms have even announced annual AI data center investment plans that far exceed market expectations underscoring the importance of large-scale infrastructure as a key to gaining AI competitiveness. As a result, investments in AI servers are expected to continue expanding for the foreseeable future. Demand growth for HBM and high-density server DRAM is also expected to continue, driven by more launches of generative AI-based services and agents. combined with rapid rise in demand for computing resources to support the growing importance of inference technologies in AI. Additionally, within the next two to three years, demand for custom HBM designed to meet the diverse needs of AI customers is also expected to grow considerably. Meanwhile, considering big tech companies' demand to replace legacy platforms to improve power and space efficiency, overall server market, including AI and general purpose server, is expected to grow in the high single digits range this year. All in all, this year's DRAM demand is projected to grow by mid to high 18%. while NAND demand is expected to increase by low teen percentage. Now I will discuss the company's plans for this year. Considering the seasonally weak period for IT devices and relatively high memory inventory levels in the supply chain for some end markets, we anticipate a low teen percent sequential decrease in first quarter DRAM bit shipments, and high 10% decrease in NAND shipments. Since the second half of last year, decoupling of demand between AI memory and conventional memory became more pronounced, and this led to differentiating product mix and sales strategies among suppliers. Based on our differentiated AI product competitiveness, and profit-oriented business decisions, we plan to maintain stable performance even during the period of market corrections. This year, our HBM revenue is expected to grow by over 100% compared to that of last year, driven by strong customer demand. In particular, demand from ASIC customers is expected to increase meaningfully, expanding our customer base. Volume shipment of HBM3E 12 high products is proceeding smoothly, and we anticipate 12 high products to account for more than half of the total HBM3E shipments by first half of this year, as previously planned. In November last year, we developed the industry's first HBM3E 16 high product based on advanced MRMUF process. further demonstrating our leadership in HBM technology. This year, we plan to complete development of HBM 4.12 high product and prepare for mass production as the product will be mainstreamed from 2026. By supporting the product on time of customers' needs, we plan to maintain our leadership in the HBM market. Additionally, by migrating to advanced nodes, We are preparing to increase production of DDR5 and LPDDR5 products, but we have solid technological competitiveness and are showing better demand. This will allow us to produce more competitive products compared to latecomers. To meet the upcoming demand for on-device AI, we are also preparing various product lines such as LPCAM2 and GUFS which offer improved data processing speed and power efficiency, and plan to supply them in time for increasing customer demand this year. For NAND, considering delayed demand recovery in a wide range of applications other than enterprise SSDs, we will continue our profit-focused approach and sales strategy that is aligned with demand, as we have done last year. Meanwhile, we are proactively preparing manufacturing infrastructure to support next generation technologies and products in a timely manner and continue leadership in those markets. M15x fab is currently under construction in Cheongju and is scheduled to open in the fourth quarter of this year. The first fab in the Yongin is also scheduled to begin construction this year, with the target FAB opening in the second quarter of 2027. The construction of these new FABs will lead to a marked increase in infrastructure-related investment this year, though overall investment increment will be limited. However, we will continue to follow the principle of investing only in products that ensure profitability. and make flexible investment decisions based on changing market conditions. In addition, we have completed an advanced packaging investment agreement with the state of Indiana in the US and signed a final funding agreement with the US government under the CHIPS Act for approximately $660 billion won at the end of last year. We will secure advanced packaging facilities to ensure our ability to produce a variety of customized memory products on time and to meet the increasingly complex demands of our customers. Finally, let me introduce our new shareholder return policy. In November 2024, we announced a new shareholder return policy to be implemented over the next three years. We have determined that in order to further enhance our corporate value, it is essential to strengthen our position as a leading AI memory company through continuous investment, while also swiftly improving financial stability that was weakened during the recent downturn. To this end, we have set securing positive net cash and appropriate cash levels as our financial goals, which will enable stable business operations and investments to maintain competitiveness, regardless of future market fluctuations. These objectives have been incorporated into our current shareholder return policy. First, the annual fixed dividend per share will be increased by 25% from 1,201 to 1,501, which increases the total cash dividend paid to shareholders to 1 trillion won per year. This ensures that the total cash dividends paid during this policy period will be at least the same level as that paid under the previous policy. However, going forward, only the fixed annual dividend amount will be paid, and the 5% of annual FCF that was distributed under the previous policy will be used to strengthen our financial soundness. At the same time, we will maintain to use 50% of accumulated three-year FCF as a resource to provide total shareholder returns. Any additional returns beyond the annual dividend will be made within the scope of maintaining financial goals after the policy term ends. Through this new shareholder return policy, we expect to enhance our financial soundness, allowing us to steadily improve our corporate value even amid market fluctuations. And as our performance continues to improve, we expect to be able to offer additional returns to shareholders. Moving forward, we will strive to consistently improve our performance based on our leadership in our AI space and ultimately increase the scale of returns to shareholders over the long term. Thank you. With that, we are now ready to take your questions.

Now, Q&A session will begin. Please press star 1 that is star N1 if you have any questions. Questions will be taken according to the order you have pressed the number star 1. For cancellation, please press star 2 that is star N2 on your phone. The first question will be provided by Kim Seon-woo from Merit Securities. Please go ahead with your question.

Hello, I'm Kim Seon-woo from Merit Securities. I have a question about commodity memory uphangs. Since the second half of last year, the price of commodity memory uphangs and supply increases in Chinese companies have been decreasing. It seems that this trend will probably continue until the first half of the year. I would appreciate it if you could tell us how you see the overall situation this year. Do you think this will end with a short-term adjustment or do you think it will be a similar adjustment period from the second half of 2022 to the first half of 2023? Thank you.

So the question is, commodity memory prices have started to decline from Q4 due to weak demand and increased supply from Chinese manufacturers, and this trend seems likely to continue into the first half of the year. So how do you view the memory market for this year, and do you see it as a short-term adjustment, or do you expect a correction similar to 2023?

Thank you for the question. I'll answer the question first. As you said, the market has been growing Thank you very much for the question.

I will first answer the part related with the DRAM. As you mentioned, commodity memory prices began to decline from Q4 last year, due to a continuous luggage demand from major applications. The price decline has been particularly evident for DDR4 and LP-DDR4 products, which are impacted by high inventory levels and increased supply from Chinese manufacturers.

We expect that the commodity memory price will be low for the time being due to the seasonal demand and customer demand.

We anticipate that the commodity memory market will continue to show a weak pricing trend in the short term because of the influence from the slow seasonal demand in the first half and inventory adjustment by customers.

However, as the demand for high-end and high-capacity memory for AIPC and smartphone products increases in the latter half of the year, even if memory suppliers expand their supply of DDR5,

However, starting in the second half, we expect increased demand for high-end, high-density memory in AI PCs and smartphones, and this demand coupled with supply expansion of DDR5 from memory manufacturers is unlikely to cause significant imbalance between supply and demand.

Therefore, we predict that there will be a continuous growth in HBM AI memory market, whereas conventional GM market will go through a gradual smooth correction. And most suppliers are focusing investment on ATBM, where demand visibility is high and profitability is secured.

This cautious investment focusing on ATBM and prioritizing AI memory production over commodity products with limited capacity suggests that the cycle, the stand, will be different from past downturns.

Let me also touch upon the NAND. 수익성 중심의 사업 운영을 지속할 계획입니다.

NAND has been more significantly affected by demand slowdowns in traditional applications and some suppliers have already announced production cuts. We also plan to continue our flexible investment and production strategies that we have maintained since 2023 to focus on profitability.

이에 따라 NAND도 지난 다운턴 때보다는 So in this slide, we anticipate that NAND will also experience a milder correction compared to previous downturns.

However, for the market situations to improve, both supply adjustments and a recovery in end demand are critical, and the pace of net market recovery will ultimately depend on how quickly demand for general-purpose applications rebound.

Next question.

The next question is from Nicola Godoa of UBS. The following question will be presented by Nicolas Godoy from UBS. Please go ahead with your question.

Good morning. Thanks for taking my question. Regarding HBM, could you give us a bit more details on your preparation stages for HBM-4? In particular, when would production actually start for HBM4-12i? And when do you expect to ship in volumes to customers initially? We've had the ramp in 2026. Do you have also a broad timeline for when 16i HBM4 would come? And lastly, for the logic-based die, can you confirm you would use TSMC from the start for HBM4-12i? Thank you.

Yes, hello. Thank you for the opportunity to ask questions. I would like to hear more details about the preparation of HBM. In particular, when will the production of HBM-4-12 start, and please tell us about the plan for the ramp-up in 2026. And when will the 16th stage come out, and in the case of logic-based AI, will you use TSMC? Please answer these questions.

Yes, thank you for the question. As I have already mentioned, our HM4 product is already in development by applying a division of raw materials that is already proven to be technical stability and mass production, and our goal is to finish the development and mass production preparation in the second half of this year and start supplying.

Thank you for the question. As we mentioned earlier, our HBM-4 product is being developed based on the proven 1B nanometer technology, which ensures both technical stability and mass producibility.

We aim to complete development and be ready for ramp-up by the second half of this year. HBM-4 supply will initially begin with 12 high products, and 16 high products will follow based on customer demand, and shipments are expected in the second half of next year.

And regarding 16-high packaging technology, we anticipate that we can leverage our experience with the advanced MRMUF process, which was first applied in HBM-3 to mass-produce HBM-4 16-high products. For the first time in HBM4, we plan to use logic boundaries for the base die to enhance performance and power efficiency. To achieve this, we are collaborating closely with TSMC as one team.

Next question, please. The next question is from JP Morgan. The following question will be presented by Jay Kwon from JP Morgan. Please go ahead with your question.

Thank you for the question. You said that capex will increase compared to last year. Is there a possibility of a reduction in the current demand environment in Goryeo? Thank you for giving me the chance to ask questions.

You mentioned that the CAPEX investment will increase this year compared to last year. Considering the current demand environment, is there any possibility of reducing the planned CAPEX? And what will be the approximate breakdown for infrastructure investment versus the FAB investment this year? And can you also compare the investment this year to last year?

Yes, first of all, thank you for the question. I will answer it. This year's investment budget is to invest in HBMs that have already cooperated with customers for supply. So thank you for the question.

Let me answer your question. This year, we plan to have a slight increase in investment compared to last year, and this will be mainly driven by HBM investment that we agreed upon with customers, and the construction of M15X and Yongin Fab that lay foundation for future growth.

As we mentioned before, when making investments, we put priority on products with proven profitability

and we are guided by principles of responding to market changes with agility and flexibility.

This year, we want to make an effort to secure future growth based on FAP and other FAP-based technologies so that we can do less business with leading technology competitiveness.

and this year our primary focus will be on investment essential to meet market demands and at the same time we'll work hard to lay groundwork for future growth including fab construction to ensure that our leading technologies will turn into business at the right time um

The majority of investments this year will be concentrated on HBM and infrastructure, and infrastructure investment is expected to increase significantly compared to last year.

We will continue to optimize resource allocation for each product to expand FCF and maximize investment efficiency by prioritizing technology transitions and also seize growth opportunities through agile market responses.

Next question. The next question is from Kim Dongwon of KB Securities. The following question will be presented by Kim Dongwon from KB Securities. Please go ahead with your question.

Thank you for your question. Please tell us how much the ratio of DDR4 and LPDDR4 will be reduced this year. Thank you.

So how much DDR4 and LPDDR4 proportion be reduced compared to last year?

Yes, thank you for the question. In the case of DRAM this year, high-performance and high-end products such as HBM and DDR5 are tight due to increased demand, while legacy products such as DDR4 and LPDDR4 have increased demand

This year in the DRAM market, high-performance, high-end products like HBM and DDR5 are expected to see tight supply because of increasing demand, whereas legacy products such as DDR4 and LPDDR4 are likely to experience an accelerated decline in demand, which will deepen the decoupling trend between product types.

The company is continuing to lead the DDR market based on HBM and DDR5 competitiveness, so we will continue to focus on the demand for these products, and we will continue to reduce the production of legacy products such as DDR4 and LPDDR4.

So our competitiveness in HBM and GDR5 underpins our leading position in the Ethereum market. And in order to focus on meeting the demand for these products, we'll continue to scale down the production of legacy products like GDR4 and LPDDR4 and streamline and clear inventory.

Therefore, the sales of legacy products such as GDR4 and LPDDR4

As a result, the share of legacy products such as DDR4 and LP-DDR4 in our revenue is projected to drop significantly from around 20% last year to single digits this year.

Next question.

The next question is from SK Kim from Daiwa Securities. The following question will be presented by SK Kim from Daiwa Securities. Please go ahead with your question.

Yes, hello. Thank you for the question. First of all, congratulations on your good performance. I have a question related to HPM. The trend of AI technology in existing training First of all, congratulations on your great and excellent performance.

I do have a question regarding HBM. As AI technology trends shift from the training phase to the inferencing phase, There are concerns that demand for high-performance HBM may also slow down. What's your view on this?

Yes, I'm Kim Il-tae. I don't think so. In the early AI market, for training AI, there were expectations that high-performance HBMs with GPUs were essential. On the other hand, when moving to inference AI, relatively low-end HBMs Thank you for the question.

I don't really think so. In the early AI market, GPUs equipped with high-performance HBM were essential for training AI, and it was expected that the inferencing phase may require relatively low-end HBM, but that was only the initial expectation. With the recent development of inference AI, the market is now shifting toward a more advanced inference, and this only highlights the growing importance of higher memory capacity and bandwidth.

Ultimately, in order to develop into an AGI that is close to human intelligence, a large amount of computing power is required in the influence process. As a result, we expect that the high-end HBM demand will be reduced because the AI market will expand into an influence. Rather, we expect it to be the most important factor in HBM growth.

So to ultimately develop artificial general intelligence or AGI that resembles human intelligence, substantial computational power is required even during the inferencing phase. And therefore, rather than slowing down demand, we expect that expansion of the AI market toward inference will become a critical driver for HBM growth.

Moreover, we anticipate that major tech companies' ongoing competitive investment will continue to secure AI leadership

and also to achieve precise training and inference results. AI investment plans are now being announced, not only by corporations, but also at the national government level, indicating that AI demand will continue to surpass our growth expectations.

Also, the development of physical AI and various AI agents introduced by CBS this year

In addition, the technologies presented at CES this year, such as physical AI and various AI agents, they are expected to serve as long-term growth drivers for HBM demand and will continue to push forward the development of both processors and software.

Thank you for your answer. Next question, please.

Next question is from Ryu Young-ho from NH Investment & Securities. The following question will be presented by Ryu Young-ho from NH Investment & Securities. Please go ahead with your question.

Thank you for the question. I'm Ryu Young-ho from NH Investment & Securities. I'd like to ask a question related to the server. Last year, the demand for the server DDL5 was very good, but I'm curious if there's any recent decline in customer demand. I would like to ask a question in relation with servers.

Server DDI5 demand last year was very strong, and are there any signs of slowdown in customer demand recently? And what's your assessment of the hyperscale customer's DM inventory level?

Thank you for the question. So,

Thank you for the question. Hyperscale customers have continuously expanded investment to secure high-performance computing infrastructure. And accordingly, AI servers has also continuously exhibited strong demand growth. And also investment in conventional servers have also been made alongside in tandem, which ensure robust server DRAM demand this year following last year.

Also, when establishing AI data center, there is a growing emphasis on efficient space utilization and high energy efficiency.

Taking into consideration the replacement cycle of servers where massive investment was made in 2017 and 2018, we expect that this year we will see a strong surge in replacement demand.

Also, if the new CPU that supports DDR5 is ramped up this year, and if we consider the relatively low DDR5 supply of customers,

In addition, with new CPUs that support DDR5 ramping up this year and customers holding low DDR5 inventory levels, server DRAM demand, particularly for DDR5, is expected to remain very strong. Even if GDI5 prices undergo temporary adjustments, server customers will still have opportunities to build inventory at favorable price, and therefore we don't really have a serious concern about potential slowdown in GDI5 demand. By proactively addressing the demand for high-performance, high-density server DRAM modules this year, we will continue to secure our leadership in the server market.

Next question.

The next question is from Lee Se-chul of Citigroup. The following question will be presented by Peter Lee from Citigroup. Please go ahead with your question.

Hello, I'm Lee Se-chul from Citigroup. I'd like to congratulate you on the decoupling of memory and CU. I'd like to ask a question about DRAM. Once again, congratulations for your excellent performance despite the deepening decoupling trend in the memory market demand.

There is a perception that your 1C nanometer process is superior to competitors. What is the SK Hynix status of 1C nanometer development, and what are your plans for expanding investment?

Thank you for the question. I will answer it. Our company is based on the excellent development of 1B nanometers, So based on our excellent 1B nanometer technology, we completed the development of 1C nanometer products in the second half of last year, and we also built potential for mass production. The maximum support movement speed is 9.2 Gbps per second, which is about 28% compared to previous generation products, and the power efficiency has been improved by more than 9%. As the AI era comes, the speed of data processing and the trend of low power characteristics are required,

Our 1C nanometer-based DDI5 product supports a maximum operating speed of 9.2 Gbps, approximately 28% faster than the previous generation, with over a 9% improvement in power efficiency. As we transition to the AI era, which demands rapid data processing and low power consumption, These products are designed to be optimized to meet future server demand.

In addition, once the nanometer products exceeded the target yield for the initial ramp of fate,

already in the development stage, indicating that the expanded production and ramp-up could lead to significant cost savings.

From the second half of the year, we will start to mass-produce with general DLNs, but we will focus on HBMs and infrastructure investments that have already secured customer demand. Considering the future demand and supply situation,

So while we will begin applying 1C nanometer to ramp up conventional DRAM products starting from the second half, a significant portion of investment this year will focus on HBM and infrastructure where we already secured customer demand. And therefore, we will keep a close eye on future demand and supply dynamics in order to make the right investment decisions for ramp up.

And also, as for the 1C nanometer process that shows the excellent performance and stable

initial yield, so we are going to apply that to the HBM4E in the future in order to ensure the timely development and supply of the next generation HBM products to secure and maintain market leadership.

Next question.

The following question will be presented by Handong Hee from SK Securities. Please go ahead with your question.

Yes, hello. I'm Handong Hee from SK Securities. I have a question about land. If the price of land continues to fall, I have one question regarding NAND.

If the current downward trend in NAND price continues, there is a possibility that the industry might return to losses in the second half of the year. Do you have any additional plan for NAND production cut? If so, to what extent and for how long?

Yes, I will answer your question. During the downturn period of the last 22-23 years, the company has been operating mainly on legacy tech. In the last 24 years, we have increased some wafer production to meet the increasing demand for enterprise SSDs.

So during the previous downturn, we cut production mainly in legacy technologies while increasing wafer production in 2024 to meet rising ESST demand.

However, products other than enterprise ESSTs have been limited due to the delay in demand for general applications.

However, for products other than enterprise SSD, we have maintained limited production considering the delayed recovery of demand in general applications.

NAND will continue to maintain the same operating standards as it does now until demand improvement is confirmed. We plan to adjust production to fit the market situation and focus on stock normalization.

Moving forward, until we see clear signs of demand improvement for NAND, we'll continue to maintain current approach and flexibly adjust production in line with market conditions, and we'll focus on normalizing inventory levels.

Thank you. Next question, please.

The next question is from Kim Young-gun of Future Asset Securities. The following question will be presented by Kim Young-gun from Future Asset Securities. Please go ahead with your question.

Hello, I'm Kim Young-gun from Future Asset Securities. Thank you for your question. I have a question for you. When do you expect the HBM volume to be confirmed in 2026? Thank you for the opportunity.

When do you expect to finalize HBM supply volumes for 2026? And do you think the strong growth trend of HBM will be sustained in the future?

SK Hynix has been working with some customers So we SK Hynix have already initiated discussions with some customers regarding ACBM supply volumes for 2026.

So we expect that we'll be able to gain some visibility for most of the next year's volumes by the first half of this year.

Considering the high investment cost and long TAT associated with the HBM business, we aim to enhance business stability and visibility

by securing a long-term contract through a proactive negotiation with customers.

And as we mentioned earlier, the AI market holds significant growth potential

far exceeding our expectation, and they are mainly driven by advancements in learning and inference technologies, as well as the integration of AI services across different industries.

In the meantime, the fourth industrial revolution, which was represented by big data, artificial intelligence, robot technology, and autonomous vehicles, is clear to see the changes in the industry through the innovation and visualization of AI technology. The acceleration of these industrial changes

The fourth industrial revolution previously represented by big data, artificial intelligence, robotics, and autonomous vehicle is now transforming industries with strong and visible AI innovation. This accelerating industrial transformation is expected to continuously generate high computing demand And therefore, we have no doubt about the sustained long-term growth of HBM demand.

Next question.

The following question will be presented by Choi Bo-young from Kyobo Securities.

Please go ahead with your question. Hello, I'm Choi Bo-young from Kyobo Securities. I have one simple question for you. Let me ask you a simple question.

In the sales performance, could you please elaborate on the significant non-operating profit recorded and the factors that influenced it?

Our company, SK Hynix System IC, has a 100% stake in the system IC, and the process of converting it to JV was completed in November last year. As a result, the profit was reflected in the sales revenue of 1.3 trillion won in the process of discharging part of the stake that the party had.

So our subsidiary, SK Hynix System, held a 100% stake in System ICOC. And they completed the process of converting the System ICOC into a joint venture last November. And during this process, some of our stakes were divested, resulting in approximately $1.3 trillion in gains. And it was recognized as non-operating income in Q4.

In addition, in December, Kioxia went public, and our investment assets in Kioxia were re-evaluated to reflect the year-end closing price.

And this resulted in the evaluation loss of approximately 100 billion won. Sorry, 200 billion Korean won.

According to the increase in exchange rate, foreign exchange-related net profit of about 6 trillion won was generated. Therefore, the net profit of the fourth quarter was 1.5 trillion won.

In addition, due to rising exchange rate, we recorded a net foreign exchange gain of approximately 600 billion Korean won. And overall, all in all, the total non-operating profit for Q4 amounted to about 1.5 trillion won.

We'll take two more questions and next question, please.

The next question is from Lee Joon-hee from Goldman Sachs. The following question will be presented by Lee Joon-hee from Goldman Sachs. Please go ahead with your question.

Congratulations on your good performance. Thank you for the question. I also have one question. There seems to be a lot of discussion in the market about Chinese oil companies. So I'd like to ask you about Hynix's opinion. How do you evaluate the technology of Chinese oil companies? Congratulations for your good performance and thank you for giving me this opportunity.

There are increasing debates on the Chinese GM companies nowadays. So how does SK Hynix assess the technological capabilities of these Chinese GM companies? Do you think that they can enter the DDR5 and HBM market?

Yes, thank you for the question. I will answer it. Since the second half of last year, the price of China's DDM supplier's DDR4 or LP4 product supply has expanded and the demand has plummeted, leading to a decrease in the price of legacy products. Recently, some

Since the second half of last year, Chinese direct manufacturers have expanded the supply of DDR4 and LPDDR4 products, and coupled with slow demand, the prices of these legacy products have continuously dropped. And recently, some reports even suggest DDF-5 development and sales by Chinese manufacturers, raising concerns about potential market impact.

However, compared to the current high-performance product response, there is a considerable difference in the technology applied by subcontractors compared to the standard process applied by major suppliers.

However, the technologies applied by these late-comer companies are significantly behind the advanced nodes used by major leading suppliers, and therefore, this will lead to a clear wide gap in quality and performance of DDR5 products.

Moreover, with ongoing restrictions against Chinese companies and Chinese manufacturers are expected to face heightened uncertainties in developing advanced nodes and technologies. We will continue to maintain a gap with the second-hand companies in all aspects, including product portfolios, performance, quality, and customer service, through the development of advanced tech development, high-end products, and various AI and memory products including HBM.

So by focusing on advanced node development and ensuring timely readiness for high-performance products and developing various AI memory products, including HBM, we'll continue to maintain a strong lead over competitors in various aspects, including product portfolio, performance, quality, and customer services. Now we'll take last question.

The last question will be presented by Simon Wu from Bank of America. Please go ahead with your question.

Yes, thank you. This is Udongjae from Bank of America. My question is about the increase in the production rate of DRAMs other than HBMs. Since the increase in the capacity of HBMs is taking place, Thank you for giving me this chance.

My question is about your prediction for the production growth rate for DRAM excluding HBM because there is an ongoing CapEx investment centering around HBM, so I want to know about your plans for a general conventional DRAM.

Thank you for your question. Over the past two years, we have utilized the advanced nodes with priority in order to reduce

conventional DRAM production in response to market downtrend, and also to mass-produced profitable HBM products. And therefore, we have seen a very limited increase in the production of conventional DRAM except HBM.

However, in order to reduce the production of legacy DRAMs, which have worsened the supply situation since the second half of last year, and to increase the supply of DDR5 and LPDDR5, Since we are expanding the 1A nanometer process, it seems that the production of general D-RAM wafers will increase significantly this year.

Starting in the second half of last year, we have scaled down the legacy D-RAM production due to worsening supply and demand conditions. But at the same time, we are expanding the adoption of 1A nanometer processes to increase supply for DDR5 and LPDDR5, which show strong demand. and therefore oil and the wafer production for conventional DRAM is expected to grow slightly this year.

Based on the increase in production and the increase in wafer production for general DRAM, the increase in production for general DRAM this year is expected to increase to the level of market demand growth.

Considering the production increase due to migration to advanced process and wafer production increase for conventional DRAM product, we expect that general DRAM production this year is expected to grow at a rate aligned with market demand.

This year, we plan to expand the production of HBM based on HBM demand, which is strong in the AI market. This year, with solid and strong HBM demand in AI market, we'll maximize HBM production, and for conventional D-RAM products,

we will have a product mix prioritizing profitability within limited capacity. And by pursuing transitioning to advanced nodes, we will ensure timely supply for customers and maintain healthy and sound inventory levels for legacy products. So this will conclude the earnings call conference call. Thank you for your attendance.