Good morning, afternoon, and evening. Thank you for participating today. We will now begin the SK hynix 2023 Q3 earnings release conference call. For today's conference call, after SK hynix's presentation, there will be a Q&A session with the participants. If you have a question, please press star one. Today's presentation will be interpreted simultaneously, and the Q&A session will be interpreted consecutively. With that, we will now begin today's presentation. Good morning, afternoon, and evening. This is Park Seong-Hwan, the Head of IR at SK hynix. Welcome to the SK hynix 2023 Q3 earnings release conference call. Before starting the conference call, allow me to introduce the executives present here with me today. We are joined by CFO, Kim Woo-hyun, Head of DRAM Marketing, Park Myung-soo, and Head of NAND Marketing, Park Chan-dong.
Let me issue a disclaimer that all outlooks presented by the company are subject to change, depending on the macroeconomic and market circumstances. With that, we will now begin the SK hynix 2023 Q3 earnings release conference call. Mr. Kim will first present the earnings for the third quarter, followed by the company's future plans and market outlook. There will then be a Q&A session with the management team. Good morning. Allow me to first introduce the company's performance for the third quarter of 2023. In the third quarter, the overall IT demand continued to show a gradual recovery. Following the previous quarter, there was a strong demand for high-density DDR5 and HBM for AI servers.
Along with the release of new products, there was a noticeable increase in demand for high performance, high-density mobile memory, driven by new product launches and the trend of higher performance flagship models in smartphones. Based on our technological and product leadership, we expanded sales of premium products such as DDR5, high performance, high-density mobile products, and HBM, resulting in a 24% sequential increase in third quarter revenue to KRW 9.07 trillion. For DRAM, sales to server systems increased significantly due to the continued strong demand for DDR5 and HBM, leading to a shipment increase of approximately 20% compared to the previous quarter. The average selling price also rose by around 10% due to price increases in key products and higher contribution from high value-added products.
Despite the base effect of a strong bit growth in the previous quarter, NAND recorded mid-single digit bit growth compared to the previous quarter, driven by the expansion of high-density mobile products and SSD sales. However, ASP declined slightly. As revenue increased compared to the previous quarter, and some of the recognized inventory valuation losses were reversed due to the rise in DRAM prices, our third quarter operating loss improved by KRW 1 trillion compared to the previous quarter, to KRW 1.79 trillion, with an operating margin of -20%. In particular, our DRAM business quickly expanded the mix of high-performance premium products, driven by industry-leading technological capabilities. Therefore, starting from the second quarter, ASP turned positive, surpassing the industry average. Based on this, our DRAM margins turned to profitability after two consecutive quarters of losses since 2012.
Depreciation and amortization expenses for the third quarter amounted to KRW 3.33 trillion, slightly decreasing compared to the previous quarter due to reduced capital spending. EBITDA recorded KRW 1.54 trillion, with an EBITDA margin of 17%. Non-operating loss net of gain in the third quarter was KRW 0.68 trillion. This includes net interest expense of KRW 0.34 trillion, and net foreign currency-related loss of KRW 0.16 trillion as a result of translation loss on foreign currency-denominated debts due to the rise in exchange rate. As a result, the pre-tax net loss was KRW 2.47 trillion. The net loss for the quarter was KRW 2.18 trillion, and the net profit margin was -24%.
Consolidated cash balance at the end of Q3 was KRW 8.53 trillion, up by KRW 0.104 trillion from the previous quarter. Interest-bearing debt was KRW 31.56 trillion, with an increase of 0.75 trillion KRW Q-on-Q. Debt-to-equity ratio and net debt-to-equity ratio at the end of Q3 was 57% and 42%, respectively, rising slightly from the previous quarter's 54% and 41%. Next, I will discuss market outlook and the company plan. Amid ongoing uncertainties in the global economy, the recovery of memory demand is being delayed compared to earlier expectations. However, entering the second half of the year, we are seeing more purchasing demand from customers who are in the final stages of inventory correction.
Additionally, as strong demand for AI servers continues, along with the added effect of new mobile product launches, the demand for high density and high-performance memory products is increasing. Meanwhile, as the reduction in memory semiconductor production takes effect to a meaningful level in the latter half of the year, prices of key products are stabilizing. Therefore, we believe memory industry is finally passing through the severe downturn and is now entering the stage of full recovery. While next year's demand conditions may still be under the influence of the macroeconomic environment, we expect set shipment growth in all applications after two consecutive years of correction. With channel inventory in the PC market normalizing, next year's PC shipment is expected to grow by mid-single digit, driven by replacement demand.
Furthermore, as decline in memory prices ease cost burdens, and as the AI PC market, which supports AI chips or AI software, emerges, memory content per PC is expected to continue growing by double digits. The smartphone market has entered a mature stage, with the used phone market expanding and replacement cycles elongating. Nevertheless, considering the replacement cycle, next year, it is expected to grow in the mid-single digit range compared to this year. Additionally, in the latter half of this year, new product launches by major mobile customers and the increasing proportion of flagship products have accelerated the demand for high capacity and high-performance LPDDR5 memory. In the long term, as the utilization of AI functions increases in smartphones, the adoption of DRAM above 16 GB and NAND above 512 GB is expected to accelerate.
Next year's server market is expected to show a gradual recovery trend as server build demand improves, particularly among CSP customers, and investments increase compared to this year. Although this year saw a temporary decline in demand for general servers due to the rapid increase in AI server-oriented spending within limited budgets, ultimately, investment in general servers is deemed essential for the utilization of AI services. Therefore, starting from next year, there will be replacement demand for general servers as well as new investments. And with the intensifying technological competition in the AI market, investments in AI-oriented servers are also expected to continue. As IT device shipments shift to positive growth and lower burden of memory costs leads to increase in content similar to levels of previous years, next year's demand for DRAM and NAND is expected to grow by high-teens%.
Meanwhile, demand growth for this year has been revised downward from the previous quarter expectations to mid-single-digit% for DRAM and high single-digit% for NAND. The company's shipment growth is expected to exceed market demand growth as we proactively respond to demand for high-spec products based on our technological competitiveness. For the first quarter, our sales bit growth is expected to be around 10% for DRAM due to higher DDR5 sales. NAND shipment is expected to decrease by around 10% sequentially, with lower sales of low-margin products to improve profitability amidst muted demand recovery. To respond to next year's demand, our company plans to expand stable production and support of products in which we have a competitive advantage.
We are prioritizing investments for tech migration to 1a nm and 1b nm, and to secure HBM and TSV capacity in order to maintain a leading position in the DDR5, LPDDR5, and HBM markets next year. Accordingly, our company's CapEx for next year will increase compared to this year, but we aim to minimize the increase by considering investment efficiency and financial stability. Yields of our 1a nm DRAM and 176-layer NAND have reached maturity for most of our products last quarter. Our next generation, 1b nm and 238-layer products yields are also improving smoothly according to schedule. Furthermore, following this year's mainstream product, HBM3, we started providing samples of 1b nm HBM3E products to our customers in August.
Our HBM3E boasts world-class performance in all aspects, ranging from speed, an essential specification for AI memory, as well as heat treatment and user convenience. Furthermore, our HBM3E product has improved heat dissipation performance by 10% compared to existing products by applying advanced MR-MUF technology. It also supports backward compatibility, allowing customers to use it in systems designed with HBM3 without any design or structural changes. Based on our industry-leading HBM development and production capabilities, we are committed to full support of our HBM3E products to customers next year, solidifying our position as a key player in future AI infrastructure. Regarding the export restrictions on semiconductor equipment to China, we have recently been notified by the U.S. government as a validated end user.
With this revision of the VEU, we believe that the uncertainty regarding our operations in China has been significantly resolved, as we will now be able to import designated items without separate export licenses. We will continue to work closely with various stakeholders to prepare for external uncertainties and ensure smooth business continuity. Next, let me share our ESG management activities and performance. Amidst the increasing global disclosure requirements related to sustainability, we have recently published a TCFD, or Task Force on Climate-related Financial Disclosures report, that outlines key climate change related risks and opportunities, as well as our response strategies. The distinguishing feature of this year's TCFD report, the second one issued following last year, is the enhanced analysis of physical risks related to climate change.
We have expanded the scope of physical risk assessment to include not only our domestic and overseas facilities, but also over 60 key partner companies. In addition, we have applied the latest climate change scenarios utilized by the Intergovernmental Panel on Climate Change in our analysis. We will strengthen our analysis and management capabilities regarding climate change risks and transparently disclose our climate response strategies, contributing to the transition to a carbon neutral society. Thank you. With that, we are now ready to take your questions.
We will now begin the Q&A session. [Foreign language] Now Q&A session will begin. Please press star one, that is star and one, if you have any questions. Questions will be taken according to the order you have pressed star and number one. For cancellation, please press star two, that is star and two on your phone. [Foreign language] The first question will be presented by Yong Gun Kim from Mirae Asset Securities. Please go ahead with your question.
[Foreign language]
Thank you very much for the chance to ask my questions. I have two questions. Now, first is about the AI inference, memory market, and, what does the company believe that this is going to have an impact, in terms of the impact on the AI inference on the memory market overall? And the second question is about the outlook for the AI server demand in 2024, and in linkage to that, also the outlook for the HBM server demand. And also, in the future, then in general, what does the company, project for the HBM memory demand to be in terms of its share out of the overall memory demand?
[Foreign language]
Now, allow me to respond to the first part of your question. Now, the question was about the shift of the AI demand into the inference market and what kind of impact it is going to have on the overall market. And now, to state it from the beginning, I believe that it is going to be directly related to increased demand in memory.
[Foreign language]
Now, as the number of parameters increase for the purpose of training, this means that the number of users will also increase, which would further drive the demand for inference as well. This also means that there is going to be a general shift from text-based generative AI to multimodal. That also means that the language model itself is going to diversify further, meaning that not only the current large language model, but perhaps also mid-size or even small-size language model, meaning that there is going also to be diversification in the language models as well.
[Foreign language]
So today, the demand for AI is mostly on chat, chat, meaning like the chatbots. But then, for the businesses, if the use of the AI can continue to increase, so for example, for product solutions, but also for the social media and social contents and also business marketing and smart assistants, this means that the data generated from all this usage is also going to explode. And that also means that we will be requiring even higher process power to process the train data as well as the inference data, meaning that, the even the devices would then be requiring even higher and stronger power. And this is going to boost, bode well in terms of increase in the memory content, not only for the accelerators, for the AI server, but also for the general devices as well.
[Foreign language]
Now, your second question was in relation to the outlook for the AI server demand, as well as the HBM demand and also the share out of the overall memory. Now, in terms of the AI server, of course, it would differ by definition, but then in general, I would say that for the AI server, it is going to be in the high single digit or up to 10%.
[Foreign language]
Now, in terms of CAGR for the next five years, then the projection now is that it is going to be by over 40% annually for the AI servers. But then again, that is the projection now. As we see today, the players are increasingly jumping into the AI ecosystem, meaning that if there are also calls for AI-only data centers or even AI-only computing power, then the AI server market growth could potentially accelerate.
[Foreign language]
For the HBM market, the current projection is an average of 60%-80% growth for the next 5 years. The CAGR of 60%-80% for the next 5 years. Having said that, again, we are seeing signs that there could be potentially even faster growth in this market. So, what we are seeing now is not a definitive demand, because we are definitely seeing even higher possibility for accelerated growth in the HBM demand. So in the end, the real growth might actually be higher than what we are expecting today.
Foreign language.
And in terms of the HBM's share out of the total market, now of course, that would also largely depend on how large the overall DRAM market is going to be next year. But, our assessment now is that out of the total market, HBM would account for about mid- to high-teens %. But again, there's also the possibility that this share might even be higher.
[Foreign language]
[Foreign language] The next question will be presented by Ricky Suh from HSBC. Please go ahead with your question.
[Foreign language]
Thank you very much for the explanation. My question is about the inventory level per application, and also what is the company's inventory level? Also, in terms of the normalization, what is the company's expectation in terms of the timeline for the normalization of the inventory? The second question is, it appears as if there is a very rapid shift to DDR5 by the company, and does the company believe that this trend will continue? Also, by the end of the year, what does the company expect the share of the DDR5 to be? And going, looking ahead to next year, when does the company believe that the crossover will occur?
[Foreign language]
Now, I will respond to your question about the inventory. Customer's inventory is improving in general, but inventory sell down is somewhat slow as demand recovers primarily for high-spec products, while customers' inventory is mostly legacy products.
Foreign language.
PC and mobile inventory seems to be gradually coming down as customers try to sell them off, but server inventory is believed to have slightly increased due to soft demand for general servers compared to AI servers.
[Foreign language]
AI memory is one of the most challenging applications, as the customers' needs are quite demanding, with high performance and reliability requirements. Supply remains tight versus demand, not a good environment to build up inventory.
[Foreign language]
Now moving on to the company's inventory. The company's inventory level at quarter end showed meaningful decline compared to Q2. While demand remains soft overall, sales of premium products, such as DDR and LPDDR5 and HBM, have increased. But demand for products in the inventory, like DDR4, LPDDR4, is slow to recover, and inventory normalization has been moderate compared to other recovery times.
[Foreign language]
Nevertheless, it is projected to decrease significantly by the end of the year, with gradual demand improvement and clear impact from production cut in the second half. Inventory level is expected to normalize in the first half of next year, primarily in DRAM.
[Foreign language]
Now, allow me to respond to the second part of your question about the shift to DDR5.
[Foreign language]
Now, for the overall market, it is expected to achieve the crossover to DDR5 in the first half of next year. So in the market, meaning the computing segment, like PCs. So that is for the general market, in the, first half of next year. But then for the company, thanks to our product competitiveness, we were already able to achieve the crossover in the computing segment, like PCs, in the third quarter of this year.
[Foreign language]
I expect the trend to continue, and this is also aligned with the shift in the product performance from 4800 to 5600, and also our high density product readiness. It also depends on the successful supply of the 32 giga mono for next year's CPU, and also the continuous expansion into 2025. Based on all these trends, we will continue to migrate to DDR5 very quickly.
[Foreign language]
In the earlier question, we talked about the inventory, and then now, for the... So in this process, the DDR4's share is going to start to decline in the second half of this year, including on the production side as well. So, according to this trend, in the first quarter of this year, the supply and demand for DDR4 is going to be somewhat balanced. And then now, for the DDR5, of course, the demand crossover is going to be important. But again, looking at the current situation now, we are at a stage where supply is going to be very tight.
[Foreign language] The next question will be presented by Nicolas Gaudois from UBS. Please go ahead with your question.
Good morning, and thanks for taking my questions. The first one relates to supply. You are likely accelerating technology migration to 1a and 1b nanometer in 2024. Micron commented this would lead for them to a loss of net wafer capacity. Would that be similar for SK hynix? And could you quantify, is your assessment similar for the overall DRAM industry? And if so, how long do you think it could take for the industry to recover 2022 wafer installed capacity? At the same time, you're also underutilized as your peers are. What is the current base for how capacity utilization rates will evolve in 2024 for DRAM and NAND, and when could you be back, you think, to full utilization? And then the second question relates to your China operations.
We've had an update from the U.S. Department of Commerce on the permanent reprieve for U.S. equipment vendors to be able to ship and service tools to China for SK hynix. With that in mind, and considering the other constraints, such as EUV restrictions, having separate process technology in Dalian versus Korea for NAND flash, what is your latest thinking about those two China fabs longer term? Thank you.
[Foreign laguage]
[Foreign language].
Thank you very much for your questions. So I take it there were, in total three. Now, first was about the wafer production reduction and also when we can go back to the 2022 wafer capacity level. Now, for the company, this year, we are executing CapEx within the given scope, while adjusted to product-specific priorities. But the current investment and production volume are not likely to be sufficient to allow us to fully respond to the growing demand while maintaining technological competitiveness. As such, in 2024, the company will continue to operate with CapEx efficiency by focusing more on tech migration rather than on capacity ramp up.
[Foreign language] DRAM, DDR5, HBM3E.
For DRAM, we are committed to increasing the portion of 1a nanometer and 1b nanometer to over half of total production by the end of 2024, by migrating to leading technology to increase production of high value add products like DDR5 and HBM3E, that will lead demand growth in 2024.
[Foreign language]
Having said that, increasing investment and utilization rate will take place gradually in line with market conditions. The outlook at this time is that it will take quite some time for the industry to go back to the level of capacity in Q4 2022, which was the peak before production cuts.
[Foreign language]
The second part of the question, I understand it to be about when we can go back to full utilization, and then also based on this, what would be the outlook for production next year?
[Foreign language]
Recovery from production cut will be gradual, in line with inventory level and market conditions. Given that NAND's inventory is higher than DRAM's, conservative production will be maintained for NAND for the time being.
[Foreign language]
The industry-wide reduction in wafer input in the fourth quarter of last year was mostly in legacy tech. When the time comes to go back to higher utilization, we need to migrate to next gen leading technology, for which the unused equipment will have to be used.
[Foreign language]
Along this line, CapEx next year will also be prioritized for the tech migration rather than for capacity up.
[Foreign language]
This will help turn around overall production growth in the DRAM and NAND industry to a positive growth next year, but only by a single digit.
[Foreign language]
Now moving on to the third part of your question, which was about the company's plan on the China fab operations.
[Foreign language]
As you have mentioned, it is indeed positive that uncertainties for our China operations have been considerably lifted by being selected as a VEU.
[Foreign language]
So basically, our fab operation strategy warrants multi-dimensional review as it involves various factors like future geopolitical situations, market demand, and the long-term need for fab space.
[Foreign language]
[Foreign language]
We are working on a plan to utilize China fabs in consideration of future technologies, product mix options, and customer demand.
[Foreign language]
We will remain on the path of production and investment cuts due to soft demand and high inventory this year, but future investment and operations will be determined based on the market situation and production facilities available at the headquarters.
[Foreign language]
[Foreign language] The next question will be presented by Soobin Lee from CGS-CIMB. Please go ahead with your question.
[Foreign language]
Thank you for taking my questions. I also have two questions. First is about the DRAM bit growth, which I see that has once again over outperformed the guidance in the third quarter, following the second quarter as well. So given that the downstream demand recovery continues to be sluggish, what were the reasons or the factors that made it possible for the company to overachieve on the DRAM bit growth for two consecutive quarters? And the next question is, now to improve profitability and the free cash flow, what is the company's plan on operating the NAND business?
[Foreign language]
I will respond to the first part of your question, which is about the DRAM bit growth in the third quarter.
[Foreign language]
Now, it is true that overall, the recovery in demand has been slow. But then in the third quarter, in the smartphone market, especially after the inventory correction, there has been growth in demand among the customers to rebuild their inventory for the finished products. And I would say that the first factor was the fact that the company was able to very effectively respond to this demand.
[Foreign language]
The second factor that I would like to point out is the fact that in the products where we have stronger competitive nuts, like, for example, DDR5 and high density servers, as well as the HBM, the growth in these products was higher than average, which I believe was also quite conducive to driving bit growth for the company.
[Foreign language]
I will respond to the second part of your question.
[Foreign language]
Now, about NAND profitability, yes, it has been sluggish. Now, NAND has been characteristically showing high price sensitivity, where demand surges with falling prices. The industry tried to reduce cost through high bit growth, but the pace of cost cutting decelerated with increasing investment burden from higher stacking.
[Foreign language]
We have particularly seen demand growth fall short of expectations over the past two years, despite a sharp decline in prices. But supply growth had continued, leading to a sharp rise in inventory level and worsening profitability, prompting all companies to begin cutting production.
[Foreign language]
We are now working on strategies for development, production, and investment for NAND that will allow us to respond to the changing market environment.
[Foreign language]
First, faced with constraints of cost reduction and profitability improvement, we plan to reduce business volatility by strengthening the high ASP premium product lineup and maximizing investment efficiency through appropriate resource allocation.
[Foreign language]
So driven by the activities that I have just explained so far, I hope that we will be able to turn NAND into a sustainable business structure.
[Foreign language]
[Foreign language] The next question will be presented by Sung Kyu Kim from Daiwa Securities. Please go ahead with your question.
[Foreign language]
I also have a two-part question. First is about CapEx, which the CFO had already alluded to earlier, and the guidance was that the CapEx is going to increase next year. But then I would just like to ask for a more specific breakdown. For example, I believe that investment is going to be much more important in increasing the HBM capacity and also into the tech migration to 1b nanometer in DRAM. So if the company can elaborate a bit more about your plan on investment in these areas? And the second question is about NAND, which is also related to the earlier response. Now, it was mentioned that in DRAM, turnaround was achieved in the third quarter, but then in order to improve the enterprise-wide performance, I would say that there is also a need for NAND turnaround.
Then when it comes to the NAND business, then there's not only the SK hynix NAND business, but then also Solidigm, which the company has acquired from Intel. So what is the company's plan for operation of Solidigm?
[Foreign language]
Thank you very much for your questions. I will respond to the first part of your question, which is about CapEx.
[Foreign language]
Yes, as it was mentioned, for the company, we also believe that investment into HBM and also into migration to 1b nanometer is the most important part of our investment.
[Foreign language]
Now, our CapEx this year increased where we needed to respond to demand for AI memory, but otherwise decreased, with the company-wide CapEx down by over 50% year-on-year.
[Foreign language]
Looking ahead to next year, investment expansion appears to be unavoidable as there will be recovery in memory demand, and we need to adequately respond to demand, especially for products where we have a competitive advantage.
[Foreign language]
The size of CapEx itself is set to grow over this year for purposes of ramping up leading technologies, expanding HBM and TSV capacity, and securing mid- to long-term competitiveness. But we still plan to minimize the increment by exercising discipline and sticking to investment priorities.
[Foreign language]
I cannot give you the specific number at this time, but we will make sure it will allow us to sufficiently respond to the growing market demand and continue to lead the premium market.
[Foreign language]
I will take your second question.
[Foreign language]
The question was on the timeline for turnaround in NAND business and also the operation for Solidigm in the future. Now, regarding NAND, NAND has a higher inventory level than DRAM, and impact from AI demand is limited. It will, it will thus take longer for the market to recover compared to DRAM.
[Foreign language]
Given this context, we plan to maintain conservative NAND production next year, as has been the guidance, while focusing on improving profitability by optimizing product lineup.
[Foreign language]
For Solidigm, challenging business environment continued with an unprecedentedly low memory demand environment. Efforts were made to reduce cost through production cuts, personnel restructuring, and to eliminate redundant efficiencies with the head office.
[Foreign language]
We plan to combine Solidigm's strength, which is the diverse customer base and firmware technology, with SK hynix capabilities, to optimize the eSSD roadmap and strengthen our product portfolio in time for market improvement.
[Foreign language]
We will expand our competitiveness in the high specification eSSD market, including the PCI Express Gen5 market this year, diversifying our NAND portfolio and expanding our customer base to secure a more steady business base.
[Foreign language]
[Foreign language] . The next question will be presented by Dong Hee Han from SK Securities. Please go ahead with your question.
[Foreign language]
I also have two questions on HBM3E. The HBM3E market is set to take off next year, and I wonder whether the company is ready to lead in the HBM3E market as it has done so in HBM3 market? And if yes, then could you also explain how and why? And the second question is, with the takeoff in the HBM3E market, is it expected that the prices in the other, HBM or HBM3 products are going to fall?
[Foreign language]
Thank you. Let me first respond to the question about whether we can continue our leadership in HBM3E market.
[Foreign language]
Now, this is one area where I can be quite confident, including both HBM3 and HBM3E. Our capacity for next year has been sold out, and we are still getting additional requests for additional demand. So, at least on the demand side, we have a very strong visibility into next year.
[Foreign language]
What we are hearing from the customers and other market stakeholders is that the company has an overwhelming dominance in terms of the market share in the HBM3E capacity.
[Foreign language]
Now for the longer term, including next year, the company is currently engaging with various companies, both the AI players and other potential customers, about further expanding our business areas. Also, what we are getting from the customer VOC is that the company inspires confidence and trust, thanks to our product competitiveness and our track record of large scale mass production. Based on this performance, we are currently discussing with potential customers about becoming the primary vendor.
[Foreign language]
Now, when we look at the evolution of the HBM demand, then going from 2E to 3, then 3E, now to 4, we see that the product migration has been very fast. In this process, the mass production quality, as well as the product performance and also the overall experience, have served the company very well in building up our advantage in this segment. As a result, we are discussing the potential technical cooperation and also capacity with our key partners, not only for 2024, but also all the way into 2025.
[Foreign language]
Now let me respond to your question about the potential decline in prices in the legacy products as we shift into a new gen product.
[Foreign language]
Now, looking at the product development cadence, then in terms of the HBM pricing, we can see two characteristics. One is that it is mostly large project based and also that it is based on long range planning. As a result of this, what we are seeing is that in most cases, it is not a parallel development, meaning that it's not the kind of situation where we have the commodity product and then the next generation product existing together.
[Foreign language]
For the new products, they would come with higher performance and also come from higher cost. As a result, the ASP would be higher. On the other side, this also means that the old products are not going to fall behind the market price.
[Foreign language]
Now, especially given the fact that next year the DRAM market is set to recover and grow very fast, then for the customers who would be adopting the HBM3 newly, then we can set the price according to the market situation. And then for those who are shifting from HBM3 to HBM3E, then I would say that this is going to be a different arrangement of supply. But then when we look at the general cadence, when we look at the general situation, then it is usually in terms of the generation change, then it is usually on a one-year cadence, meaning that from HBM2E to HBM3 and HBM3E. And again, in this process, the company has maintained leadership and superiority. So I believe that we are still in the position to do so.
Now, then, to sum up, in the conventional DRAM, then perhaps, there had been price interference between old tech and new tech, but that is not what we will be seeing in HBM.
[Foreign language]
Thank you. We will take the last question.
a question is Joon Lee from Goldman Sachs.
The last question will be presented by Joon Lee from Goldman Sachs. Please go ahead with your question.
[Foreign language]
Thank you very much for taking my question as the last one. Now, according to news reports recently, the company is opposed to the merger between Kioxia and Western Digital. I wonder whether the reports are true, and if yes, then what are the reasons for the opposition? And that being the case, then how does the company plan to utilize the invested asset in Kioxia? And then the second question is about the partial reversal in the inventory loss valuation. And so there was the reversal in the third quarter. So if you could share with us the specific size?
[Foreign language]
Thank you very much for your questions. I would first respond to the question about the Kioxia and the Western Digital consolidation.
[Foreign language]
The company is not agreeing to the deal at this time, in light of the overall impact on the value of the company's investment in Kioxia.
[Foreign language]
Please understand that we cannot disclose the specific reasons or comment on the deal process, due to the confidentiality agreement with Bain.
[Foreign language]
But I do wish to make one thing clear. We will be making the decision for the sake of all stakeholders, not only the shareholders, but also Kioxia as well.
[Foreign language]
Now, moving on to the second part of your question about the inventory valuation loss.
[Foreign language]
There was partial reversal in the inventory valuation loss, according to LCM in the third quarter, as DRAM prices rose and NAND price decline slowed. It was mostly for NAND products, with a total value of reversal of KRW 60 billion.
[Foreign language]
Going into the fourth quarter, further inventory reduction and price stabilization are expected, which is likely to increase the amount of reversal in inventory valuation loss. Thus, it is also expected to keep contributing to improving business performance.
[Foreign language]
Thank you very much. With that, we conclude the SK hynix 2023 Q3 earnings release conference call. Thank you very much for your participation.