Morning and afternoon and evening. Thank you for participating today. We will now begin 2023 Q2 SK hynix conference call. For today's conference call, after SK hynix presentation, there will be a Q&A session. If you have a question, please press star one. Today's presentation will be interpreted simultaneously, while Q&A as consecutive. With that, we will now begin SK hynix presentation.
Morning, and good afternoon, and evening to those calling in from abroad. This is Park Seong Hwan, the Head of IR at SK hynix. Welcome to the SK hynix 2023 Q2 earnings release conference call. Before starting the conference call, allow me to introduce the executives present here with me today. First, Kim Woo-Hyun, CFO, Park Myoung-Soo, Head of DRAM Marketing, and Park Chan-dong, Head of NAND Marketing.
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 SK hynix earnings release conference call. Mr. Kim will first present the earnings for the Q2 2023, followed by the company's plan and market outlook. There will then be a Q&A session with the executives.
Good morning. Allow me to first report to you the company's performance for Q2 2023. Despite prolonged economic uncertainty and a milder-than-expected recovery, Q2 IT demand saw a surge in demand for AI-related memory products due to intensified competition in the development and commercialization of generative AI. SK hynix expanded sales of premium products such as high-density DDR5, high-performance LPDDR5 and HBM, enabling the company to achieve shipment volumes above guidance and blended average selling prices higher than previous quarter.
As a result, revenue in the Q2 was KRW 7.331 trillion, up 44% from the previous quarter. DRAM bit shipment grew by mid-30% from the previous quarter due to strong demand for high-end servers, as well as some inventory build demand from mobile and PC customers. Blended ASP rose by high single-digit % sequentially, despite the continued price decline of DDR4, driven by the greater sales of premium products. In particular, our graphics DRAM sales, including HBM products, has previously constituted only a single-digit % of our DRAM sales, but since reaching 10% of DRAM sales in Q4 last year, this portion has grown rapidly to exceed 20% in the Q2.
The sales of our HBM and high-density DDR5 modules used for AI servers expanded significantly, contributing to DRAM bit shipment and ASP growth in the Q2. NAND bit shipment grew by approximately 50% from the previous quarter due to the expansion of sales across all applications, as well as the base effect of low shipment in the previous quarter. ASP fell about 10%. Although the cost of goods sold increased in the Q2 with rising sales volume, sales and general administrative expense decreased due to cost reduction efforts. The company recorded operating loss of KRW 2.88 trillion and operating margin of negative 39% in the Q2, reflecting lower inventory valuation losses due to the improvement in DRAM pricing environment.
Having passed a trough in the Q1, the memory semiconductor market has started recovering as sales increased and operating losses and margin improved significantly compared to the previous quarter. In the Q2, depreciation and amortization was KRW 3.49 trillion, slightly lower than previous quarter due to reduced investments, and EBITDA was KRW 0.61 trillion, with an EBITDA margin of 8%. There was net interest expense of KRW 0.33 trillion, and net foreign currency-related loss of KRW 0.52 trillion, including translation loss from investment on Kioxia due to weak Yen. Combined net operating loss was KRW 0.91 trillion. Net loss before tax was KRW 3.79 trillion, and net loss for the quarter was KRW 2.99 trillion, with net profit margin of -41%.
Consolidated cash balance at the end of Q2 was KRW 7.49 trillion, up KRW 1.35 trillion from the previous quarter. interest-bearing debt was KRW 30.81 trillion, increasing by KRW 2.05 trillion from that of previous quarter. Debt to equity ratio and net debt to equity ratio at end of Q2 was 54% and 41% respectively, up from 47% and 37% of last quarter. Next, I will talk about market outlook and company's plan. Amidst the ongoing uncertainties in the global economic environment, the recovery of memory demand is taking place at a slower pace. Shipments of PCs, smartphone, and traditional servers have already been corrected downward from the beginning of the year expectation.
However, from the Q2, the market saw strong demand for high density, high performance memory in AI servers, in addition to higher content growth in IT devices due to price elasticity. AI servers use at least 2- 8 times more memory compared to traditional servers for faster computational processing, and adopts high performance memory products such as HBM, which not only drives demand, but also positively impacts profitability. Customers' memory inventories continued to decline in the Q2. With impact of production cuts gradually materializing since the Q2, there was a slowdown in the decline of memory prices compared to the Q1. In the latter half of the year, demand is expected to improve compared to the H1 due to seasonality and release of new mobile products. With greater effects of production cuts to materialize, supply-demand conditions are expected to further improve.
Looking at the demand by application in the PC market, channel inventory decreased on back of promotional events in Q2, sales volume have improved compared to the Q1. This was mainly driven by replacement demand of Chromebooks and low-end laptops, impact on memory content was limited. In the H2 of the year, although macro uncertainties persist, demand is expected to increase compared to the H1, mainly driven by enterprise and gaming PCs. Improvements in mobile demand following Chinese reopening has been rather subdued, falling short of expectations at the beginning of the year, despite large-scale promotions in the Q2. Smartphone demand is expected to improve in the H2 of the year, following new mobile product launches by memory customers and strong content growth from lower memory prices.
Additionally, as demand growth is centered on flagship products, the demand for high density and high performance LPDDR5 is expected to expand meaningfully. The server market is expected to be rather slow this year due to reduced corporate IT spending, stemming from concerns on economic downturn and inventory adjustments by CSP companies. However, the generative AI-triggered demand for high-end servers has increased rapidly since the Q2, and accordingly, the demand for high-density DDR5 and HBM products is also growing rapidly. This year, with reduced level of corporate investments, customers are focusing on investing in costly AI servers, and thus the investment in traditional servers may decline temporarily. However, general purpose servers will still be widely used for the development, deployment, and operation of AI service, and thus, in the long run, will stimulate demand growth for general purpose servers as well.
With AI server market growing at an average annual rate of 30% in the medium to long term, overall demand for servers, including general purpose servers, is expected to grow at an average annual rate of high single-digit % level. Given such market outlooks by application, this year's demand is expected to grow by mid-to-high single-digit % for DRAM and mid-teen % for NAND. Similar to the forecast of last quarter, the company will actively support demand for high-performing premium products, leveraging our technological competitiveness. In the 3rd quarter, the company is planning for a low-to-mid-teen % bit shipment growth in DRAM and flat bit shipment growth for NAND sequentially due to high base effect. For DRAM, we will continue to support demand for high-density products and HBMs and expand sales of DDR5 and high-performance LPDDR5....
for NAND, we will expand sales of 176-layer-based SSD products, thereby continuing the revenue growth trend. Memory demand has been recovering since Q2 from its record low growth levels seen last year, but the recovery is still insufficient to normalize the industry-wide elevated inventory levels seen since H2 of last year. In particular, due to high inventory levels and lower profitability of NAND compared to DRAM, the company has decided to further reduce NAND production and anticipate this to accelerate the normalization of inventory levels. The company continues to invest as planned, where this year's consolidated investment is to be reduced by at least 50% compared to 2022.
Within significantly reduced level of CapEx this year, the company is striving to secure capacity, as well as investment for high-density DDR5 and HBM3 through efforts such as productivity improvement, accelerating equipment delivery time, and cutting investments in other areas. The impact of the industry-wide supply cuts, which started end of last year, is gradually being materialized and additional reductions are being made according to market demand. The effects of the reduction will become more pronounced in the H2 of the year compared to the H1, and due to the reduced production capabilities of this year, balance of demand and supply is expected to become tight next year.
Yields on our 1a nanometer DRAM and 176-layer NAND have reached mature yields for most of our products, and our next generation, 1b nanometer and 238-layer, are ready for mass production. These products will begin production at small scale this year and have their shares expanded when demand improves next year. In the Q2, by transitioning our 1a nanometer competitiveness to 1b nanometer, the company provided Intel with market's fastest 1b nanometer-based server, DDR5, for validation. The 1b nanometer DDR5, applying High-K Metal Gate process, consumes more than 20% lower power than 1a nanometer. In the H1 of next year, the company will continue to strengthen technological leadership by 1b nanometer to LPDDR5T and HBM3E.
This year, DDR5 and HBM market, where the company has competitive advantage, is seeing significant growth. The company offers a wide range portfolio of 1a nanometer-based DDR5 products, being the sole company to have its 16 gigabit, 24 gigabit, and high-density modules of 128 gigabytes and above validated. Furthermore, our HBM products are now offered up to 24 gigabytes HBM3, the highest density available, with 12 layers stacked, a world-first achievement. Drawing on this competitiveness, this year's sales of high-density 128 gigabyte and above DDR5 server modules and HBM is anticipated to grow by over 2-fold from the previous year and expect to grow further next year. Our strong full lineup of 1a nanometer-based DDR5 products, along with our high market share and competitiveness in HBM, will help us continue to strengthen our leading position in the AI memory market.
Next, I will explain the company's ESG activities and performance. The company recently released its Sustainability Report 2023, which highlights our ESG activities, achievements of the past year of 2022, as well as future commitments. In 2022, SK hynix significantly increased its company-wide renewable electricity usage rate from 4% in 2021 to 29.6% in 2022. Additionally, the company achieved Platinum Zero Waste to Landfill designation, the highest achievable level, by diverting 100% of waste away from landfill for all domestic sites. Detailed information on SK hynix ESG performance in 2022 can be found not only in the Sustainability Report, but also through the company's sustainability reporting system.
SK hynix has completed on-site evaluations of key high-risk suppliers across the supply chain, which the company initiated in 2022. This is a follow-up from the ESG online self-assessment results of the supply chain conduct, conducted last year. The task identified through this on-site assessment will be managed through regular monitoring for further improvement. SK hynix plans to improve the biennial supply chain evaluation process, which involves supplier self-assessment and SK hynix on-site evaluation in order to strengthen the ESG capabilities of the company's partners across the supply chain. Thank you. With that, we are now ready to take your questions.
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 the number star one. For cancellation, please press star two, that is star and two on your phone. The first question will be provided by Subin Lee from CGS-CIMB. Please go ahead with your question.
Thank you very much for taking my question. I have, this is a two-part question. Now, the first part of my question is about the bit growth. We see that the bit growth in the Q2 has outperformed the guidance. Given the fact that the IT demand recovery is yet to catch up, what do you believe, what does the company believe, were the factors that drove the customers' purchases? Does the company also believe that this trend is going to continue into the Q3 of the year? The second part of my question is about the revenue mix for the HBM and DDR out of DRAM in the Q2, and what does the company believe that they will be by the end of the year?
Let me first answer the first part of your question. In the Q2, there was a big growth in the AI demand for the company's DRAM, and as a result, there was more than double growth in the shipment of the DDR5 and the HBM. I would say that it comes down to the company's fundamental competitiveness in our products. Also at the same time, there was some inventory build demand among the PC and mobile customers whose inventory level had dwindled in the Q1. We see that for the PCs, the channel inventory is improving. Also for the mobile consumers, there was a higher preference to purchase flagship products. As a result, there was also some recovery in the demand for high content memory.
Now for NAND, it appears that the demand uncertainty is going to continue throughout the year. Also given the fact that we still need to improve the inventory soundness, we have tried to maximize sales across all applications.
...And also for the, in the Q3, we believe that the server demand for AI is going to keep increasing, especially for high density DDR module and the HBM, and also the demand for high among the mobile customers, demand for high density memory is becoming more and more visible. We believe that in the Q2, and so, moving on from the Q2 into the half, into the H2 of the year, then we believe that the demand is going to remain as healthy as the Q2. For NAND, there was a very high base effect, and as a result, the shipment remained, almost flat from the Q2.
Looking at the bit growth, then we believe that, in the Q2, there was also some adjustment, from then, but we believe that there is going to be some. We will continue to try to increase sales based on the real demand and better application mix. Per your question, the second part of your question about the revenue share by the HBM and DDR5. As was also explained during the presentation, for the graphic DRAM, including HBM, the revenue share has grown precipitously since the Q4 of last year. So much so that it has now gone over 20% out of the total DRAM revenue. For the DDR5, we see that, the demand for DDR4 continues to soften.
We are very quickly switching to DDR4, 5 production, although we cannot give you the specific number at this time. I can say that, compared to the market, I would say that our mix strategy is about 2 quarters ahead. About the HBM and the high density DDR, in other words, the products for AI. Now, as I believe I have also mentioned this in the last earnings release conference call, again, we believe that the revenue for these 2 products are going to more than double, and we will continue to see growth in the revenue of these 2 products, more so in the H2 than the H1 of the year. For the year, it is likely that the 2 product share is going to top 20%.
The next question will be presented by Minbok Wi from Daishin Securities. Please go ahead with your question.
Thank you very much for taking my question. I also have two questions. First is about the demand. For the DRAM demand, the company once again projects mid to high single digit growth. According to the other, let's say, analysis organizations, they are downward adjusting the demand forecast. I wonder why the company is a bit different from this trend? Does the company believe that there is going to be further adjustment in the demand projection? Also for 2024, does the company believe that demand will begin to normalize? The second part of the question is about the HBM. Is there any specific technical roadmap for HBM following the HBM3? What does the company believe are the differentiators for its HBM product compared to the other players?
Thank you very much for the question. First, about the demand. Now, the macro uncertainties are persisting, and we see that the soft demand continues across all applications which is likely to continue. sorry, let me repeat. Macro uncertainties are persisting, so we see that the demand remains softer across all applications than we had expected at the early part of the year. As a result, the company has also downward adjusted its demand forecast, which we have already announced in the last earnings conference call. We believe that this demand forecast, which has been downgraded, is going to stay. We see that from the early part of the year, demand has remained weak for smartphones, PCs and servers.
Starting in the Q2, demand for AI servers has risen sharply, offsetting some of the loss in the demand in other products. Moving into the H2 of the year, there is going to be a new launch of mobile products. We believe that compared to the H1, the H2 demand is going to recover. Moving into next year, although Macro uncertainties will remain, the demand improvement is also going to depend on the circumstances. Given the fact that demand had remained soft for the past two years, there's likely that there's going to be stronger replacement demand for smartphones and PCs.
For general purpose servers, because the purchases of these servers had also been sluggish in the past few years, it is highly likely that there is going to be recovery next year. Let me respond to the second part of the question, which is about the HBM roadmap after HBM3, as well as the company's competitiveness. Allow me to focus on what we are doing. For the HBM roadmap, what is important is for us to be in alignment with the graphic processor launch or the schedule by the companies that are leading the accelerator market. Looking at the past three years of trends, for example, the move from HBM2E to HBM3, also looking at the adoption planning for the HBM3E, then we see that the product life cycle is quite fast, at an interval about two years.
Given this means that there is going to be a shift toward HBM4 sometime in 2026. Based on this assumption, the company is also making the appropriate preparation.
... 고객들 피드백을 종합해 보면, 그 time to market 관점에서 제품 완성도나 양산 품질, 나아가서 필드 품질까지 종합해서 SK hynix가 가장 앞서고 있다는 점이 확인되고 있습니다. 저희 입장에서는 HBM 시장 형성 초기부터 지금까지 오랜 기간 동안 경험과 기술 경쟁력을 축적해 왔다고 보고 있고요. 이런 걸 바탕으로 해서 시장 계속해서 선도해 나갈 계획입니다.
About the customer feedback, we see that in terms of the time to market as well as the product readiness and the quality of mass production, as well as the field quality. In all these categories, the customers believe that SK hynix has leadership and has advantages, and that we believe is owed to the experience and the technological competitiveness that we have been accumulating ever since we launched the first HBM quite early in the market. Based on these advantages, the company intends to keep leading the market.
네, 다음 질문해 주십시오.
다음으로 질문해 주실 분은 하이투자증권의 송명섭님입니다.
The following question will be presented by Myung Sup Song from HI Investment & Securities. Please go ahead with your question.
네, 감사합니다. HI Investment & Securities의 Myung Sup Song입니다. 저도 그 HBM 관련해서 두 가지 질문을 드리겠습니다. 첫 번째는, 뭐, 장기적으로 HBM의 수요가 계속해서 크게 증가할 거라는 것에는 이제 이견이 없을 것 같은데. 그런데 이제 최근에 보면은 그 빅테크 업체들이 자기네 자체 LLM 서비스 출시를 위해서 올해부터, 굉장히 빠른, 그런 강력한 그런 투자에 지금 나서고 있는 상황인데, 이게 과연 내년 뭐 1분기, 2분기까지 이런 강력한 투자가 이루어지고 난 다음에 이런 투자가 일단락이 되면 그다음에 단기적으로는 좀 그 투자의 휴지기가 올 수 있고, 그러면, 빨리 성장하고 있는 HBM 수요가 역시 내년 하반기 정도에는 조금 그 slow down 될 수 있지 않냐? 이제 그런 우려도 좀 있을 수 있을 것 같습니다. 그래서 그 관련한 전망 좀 말씀 부탁드리고요.
두 번째는 그 HBM3 시장을 이제 독점해 오고 계신데, 지금 이제 뭐 시장 보도나 뭐 업계 소식 따르면 경쟁사가 조만간 하반기 중에 진입이 가능하다, 이렇게 지금 되어 있는 상태고, 특히 지금 그 보도나 이런 거에 따르면 경쟁사가 2.5D package와 그다음에 HBM3을 동시에 이제 공급하는, 이제 그런 쪽으로 고객에게 제시하고 있다, 이렇게 지금 나오고 있습니다. 과연 이렇게 됐을 경우에, 뭐 당분간은 이제 경쟁사가 TSV 케파가 부족한 걸로 알려져 있기 때문에, SK hynix가 이제 계속 그 우월적인 위치를 이제 차지하실 수 있겠지만은 내년 상반기부터는 조금 경쟁 상황이 좀 달라질 수 있지 않을까, 그렇게 볼 수도 있겠습니다. 관련한 설명을 좀 부탁드리겠습니다.
I also have two questions about HBM. Of course, for the long term, I believe that there is no disagreement that there is going to be a sharp increase in the demand for HBM. When we look at some of the big tech's movements, then in order to launch more LLM services, they are investing very quickly and very intensely into this field. But then now, after the investment is pretty much completed by the Q1 or the Q2 of next year, then there are some concerns that there might be some idle period in investment, meaning that there could be some dampening of the very fast growing demand for HBM by that time. There are some concerns that the HBM demand is going to somewhat slow down in the H2 of next year.
What does the company see? What does the company think about this projection? The second question is about, again, the HBM. Of course, the company currently has a near monopoly on the HBM3, there are some news and press news about how another player, a competitor, is going to jump into this market in the H2 of the year. The rumor is that the competitor is offering a 2.5D package together with the HBM. Given the fact that the competitor does have a shortage in TSV capacity at this time, it is likely that Hynix is going to maintain the comparative advantage for the time being. This probably means that the competition landscape is going to somewhat change starting in the H1 of next year.
Again, what does the company see about this news?
네, 굉장히 복잡한 질문을, 어려운 질문을 주셨는데, 뭐, 제가 이해하고 전망하는 선에서 충실히 좀 답변을 드리겠습니다.
Thank you very much for what is, quite a complex and difficult set of questions, but I would like to respond faithfully, to the best of my ability and my knowledge.
클라우드 업체들, 플랫폼 업체들이 이제 단기적으로 CapEx 투자를 확 늘리고, 말씀하신 것처럼 일종의 휴지기로 갈 수 있느냐? 저희 메모리 입장에서는 일종의 캐즘일 텐데요. 그러한 부분들이 없을 거라고 완전히 부정하기는 어려울 것 같습니다.
Now, I believe that what was inside your question was about how the cloud companies or the platform companies are very fastly increasing their CapEx investment to be followed by an idle period in investment, which could be something like a chasm for memory companies. I must say that, we cannot take the possibility off the table entirely.
다만, 그 내년도까지 이제 질문을 주셨기 때문에 내년도까지 좀 놓고 보면, 어, 어느 정도, 어, 저희 고객들과 롱레인지 플래닝에 대한, 아, 비저빌리티는 저희가 확보해 놓고 있다라고 이렇게 보여지고요. 이유는, 어, 플랫폼 같은 경우에는 지금 현재, 어, 트레이닝 위주로, 어, AI 서버들을, 그 준비를 하고 시장 서비스를 하고 있는데, 이 부분들이 결국은 인프런스 서버 관련된, 그 보완, 어, 보충들이 필요합니다. 그래서 그러한 것들이 플랫폼 업체마다, 어, 번갈아가면서 일어날 거라고 보고 있어서, 어, 특정 업체 기준으로 만약 고객 기준으로 본다면, 어, 투자, 그다음에 Harvest 기간, 그다음에 재투자, 이런 것들이 나타날 수 있겠지만, 이런 것들이 시장 내에서 다양한 플랫폼 업체들 간의 교차가 된다면, 어느 정도 이러한 캐즘들은 매우 짧거나 아니면은 그, 어, 정도가 아주 약할 걸로 그렇게 보고 있습니다.
Since you also asked about the projection into next year, now, for our customers, we believe that we do have some visibility into their long range planning. That is because now, of course, for the platforms, what they are currently focusing on is about a training for the AI. In order to prepare the AI server and to launch the services in the market, this means that they also need to keep making improvements and strengthen their inference servers as well. Having said that, now, if this were to take place across different platform companies at different time intervals, then we believe that this is going to mean that even if there is a chasm, it is going to be short or very weak.
If it's just a single company, then yes, there would be period of investment followed by harvest and then reinvestment. Given that there are several platform companies, making such investment and going through these iterations, then that also means that again, the chasm is, if there is one, then it is likely to be quite short and weak.
수요의 백그라운드 측면에서 보면, 아직 이 발전 방향이나, 스코프를 예단하기 어렵긴 한데요. 지금은 이제 LLM 기반의 ChatGPT, 이런 부분들이 전면에 나서 있지만, 이 부분들이 결국은 개인화 과정을 거치면서 이제 맞춤 서비스 형태까지도 진화해야 되는 이러한 방향성들이 하나 있고요. MS Copilot 같은 이제 기업용 서비스를 본다면, 결국은 이 부분들이 일종의 pre-trained 서비스 형태라서, 종국에는 실시간 혹은 지속되는 training, 이런 쪽으로 연결이 되어야 될 겁니다. 그렇게 된다면, 앞서 말씀하셨던 그러한 이제 휴지기라는 부분들이 조금은 또 완화될 수 있는 그러한 배경이 될 수 있다, 이렇게 보고 있습니다.
Another factor that we have to consider is about the technological development trend. Of course, in terms of the background to the demand, then, it is difficult to predict what the actual demand, what the actual development pathway or the scope is going to be. On one hand, it is true that today, it's mostly LLM based, led by, for example, ChatGPT. We believe that it is ultimately going to go the path of personalization, meaning that there are going to be customized services, arising from the AI services. Also when we look at specifically the enterprise services like MS Copilot, then they would come pre-trained. This would also mean that, in order to keep providing the services, then they would have to have real time and almost ongoing training.
That means that even if there is going to be some breakage in the investment, again, that is not going to last long.
한편으로 그, 이 플랫폼 대비해서 이제 그쪽에 올라가야 될 앱이나 API, 뭐 이런 부분들은 이제 막 시작 단계라고 볼 수 있습니다. 이 부분들이, 커머스 분야하고 연결되는, 그 폭발력도 갖고 있기 때문에, 이제 그러한 것들이 일종의 기저 수요가 된다면, 다양한 플랫폼 서비스 경쟁들이 일어난다면, 이러한 캐즘 들은 생각보다 영향이 적을 수 있을 것 같다, 이렇게 전망하고 있습니다.
Another factor that we have to consider is not only the platform, but also the app applications or the APIs that would be placed on top of them, which are only in the beginning phase. They also have the potential to be connected with commerce, meaning that they could also, potentially lead to explosive growth. They could serve as the base demand. Also given the growing competition on the platform side, again, I would say that all these factors lead to the conclusion that even if there is going to be a chasm, it is going to be short-lived.
내년도 하반기로 가면서 이제 그 단기적으로 앞서 말씀하신 휴지기가 올 수 있느냐? 수요가 줄 수 있지 않느냐? 이제 우려, 그다음에 경쟁사 관련된 질문을 하셨는데, 뭐 경쟁사 이야기는 저희가 하기는 어려울 것 같고요. 일단 내년도 하반기까지 본다면, 앞서 언급드렸던 그 이 로드맵상 내년도 하반기는 HBM3E가 이제 본격적으로 양산 그다음에 확대되는 시기입니다. 그거에 맞춰서 저희 주 고객들과 지금 이야기를 해나가고 있고요. 캐파 준비 하고 있는 걸 보면 일단 내년도까지 HBM 쪽에 성장세에 큰 우려는 지금 없다고 이렇게 믿고 있고요. 다만, 고용량 모듈 같은 경우에는 이제 워낙에 이 CapEx성 투자에 대한 부담이 좀 있기 때문에, 다양한 고민들을 아마 그 플랫폼 업체 쪽에서 하게 될 것 같습니다. 그래서 그런 부분들은 보면서 대응을 해나갈 계획입니다.
Now, you asked about whether there is going to be some break in investment and that there are concerns about slowdown in demand, perhaps in the H2 of next year. I would like to respond to this. Of course, you also asked about, you know, competition's movement, I must say that it's not appropriate for us to comment on the competitor. Again, looking ahead to the H2 of next year, according to our roadmap, that is going to be the period of mass production of HBM3E and also expansion of supply of HBM3E. In our discussion with the main customers, that is the assumption.
When we look at the capacity planning as well, then we believe that the growth in the HBM demand is going to stay quite healthy well into next year. One caveat is about the high density module, because this is an area where the CapEx investment might be a bit burdensome for the platform companies. I believe that for the high density module CapEx, the platform companies would have to perhaps think through this further.
네, 다음 질문해 주십시오.
다음으로 질문해 주실 분은 JP Morgan의 JJ Park님입니다. The following question will be presented by J.J. Park from JPMorgan. Please go ahead with your question.
질문 기회 감사드립니다. 너무 HBM 질문만 있어서요. 딴 거를 좀 질문드리겠습니다. 2분기 보면 지금 굉장히 DRAM, 특히 NAND 쪽에 많이 나왔는데, 그 기말, 2분기 말 기준으로 재고 수준이 어느 정도 되는지 궁금하고요. 늘 말씀하셨듯이 2분기 때 재고 관련된 평가 손실 규모랑, DRAM 가격도 이제 리커버가 되는 것 같은데, 향후에 추가적으로 얼마 발생 가능성 좀 말씀해 주시면 감사드리겠고요. 두 번째는 신문 미디어 쪽에서 나오는 Kioxia랑 Western Digital 합병 관련돼서 논의가 되고 있는데, SK hynix 어차피 주주로서 그 입장이 어떤지 말씀해 주시면 감사드리겠고요. 이거는 약간 추가 질문보다는 그 계속 말씀하신 게 2분기에 HBM 플러스 그래픽이 전체 DRAM 매출의 20%인지, 아니면 HBM 플러스 하이덴서티 DDR5까지 포함해서 전체 DRAM 매출의 20%인지 좀 clarify 해주시면 감사드리겠습니다.
Since most of the questions had been about HBM, let me ask something else. In the Q2, we see that the bit shipment growth was quite rapid for both DRAM and NAND, but especially for NAND. What was the level of inventory at the end of the Q2? For the inventory valuation loss in the Q2, what was the extent of the loss? For the DRAM price, it seems as if there is now some recovery in the price as well. What does the company believe is going to be the extent of the price recovery? The second question is about the news about consolidation between Western Digital and Kioxia. What would be the company's view regarding this potential consolidation, given that the company is also a shareholder?
Point of clarification, now, in terms of the share of the HBM, was it at 20% by HBM plus graphic, or was it over 20% for HBM plus high density DDR5?
우선 맨 마지막에 어 보완 질문 주셨던 부분, uh, clarify를 먼저 하겠습니다. 그 uh, 이십 프로라고, 이 분기에 이십 프로라고 말씀드렸던 부분은 uh, HBM과 graphic segment로 어 발생한 매출입니다. 그래서 저희가 통상 그 PC, server, mobile, graphic, 뭐 이런 식으로 응용 분야 uh, segment를 나누기 때문에, 이제 그 관점에서 HBM이 포함된 graphic uh, 비중으로 말씀드렸다고 이해해 주시면 될 것 같습니다.
First, about the clarification. For the 20% in the Q2, so that was HBM plus graphic segment, because the company segmentation is by PC, server, mobile, graphic. In this case, we meant the segment that includes HBM, which is the graphics segment.
... 문의 주신 재고 관련해서 말씀을 드리겠습니다. 저, 정확한 숫자로 말씀드리지 못하는 부분은 미리 양해를 부탁드리겠습니다.
Allow me to respond to your question about the inventory level, first off, please understand that I cannot give you the specific numbers.
Uh, 이 분기 말 당사 재고는 판매 확대와 감산 효과가 나타나기 시작해서, uh, 전분기 대비 소폭 감소했습니다.
At the end of the Q2, the company's inventory level fell slightly as a result of growing sales and the effects of production cuts.
그중에 제품별로, 경향성은 좀 다르게 보이는데요. DDR5와 HBM 새롭게 수요가 성장하는 제품 중심으로 판매가 늘어났기 때문에 재고가 감소한 반면에, DDR4, LPDDR4 같은 제품에 대한 수요는 전방 수요 약세 지속으로 인해서 회복세가 좀 더딥니다. 여전히, 낮지 않은 수준의 재고를 보유 중에 있습니다.
Again, we see different trends by product category. Where the demand is strong, for example, DDR5 and HBM, of course, sales grew and the inventory level has gotten lower. Where the demand remains soft, like DDR4 or LPDDR4, we see that the downstream demand continues to remain weak, meaning that the recovery also has been quite slow. For these products, the inventory level remains not low.
하반기에는, 상반기 대비해서, 개선된 수요와, 감산 효과가, 영향을 미치면서 연말 재고는 현재보다, 상당 부분 축소될 것으로 기대하고 있습니다.
In the H2, compared to the H1, we believe that as demand continues to strengthen and the effects of production cuts continue to spread, then by the end of the year, the inventory level is going to be lower than what it is now.
네, 이어서 질문 주신, 재고 평가 손실에 대한 규모와 추가 발생 가능성에 대해서 답변을 드리도록 하겠습니다.
Allow me to respond to another part of your question, which is about the inventory valuation loss. What was the extent of it, and are there any possibility of additional losses?
Uh, 아시다시피 이 분기에는, 그, 낸드 가격이 지속적으로 좀 하락하는 상황이었기 때문에, 그럼에도 불구하고 당사는 약 한 오천억 정도 수준의 재고 평가 손실을 인식했습니다. Uh, 해당 수준은 전분기 대비해서는 크게 줄어든 금액입니다.
as you would know, in the Q2, NAND prices continued to fall. Despite this, the company recognized about KRW 500 billion in inventory valuation loss, which is considerably lower than what it was in the Q1.
Uh, 추가 발생, 발생 가능성 부분에 대해서는, uh, 하반기 쪽은 아마 재고 감소 가속화와, uh, 가격의 어떤 안정화가 예상되는 상황이기 때문에, uh, 추가적인 어떤 재고 평가 손실이 발생할 가능성은 일단 낮다고 예측을 하고 있습니다.
Whether there is going to be additional recognition of inventory valuation loss? Well, currently, the expectation is that inventory reduction is going to accelerate in the H2, even as the prices continue to stabilize. We believe that at this time, there is the likelihood of additional valuation loss is going to be low.
만약에 발생을 하더라도 그 규모는 미미하다고 예상을 하고 있습니다.
Even if it does happen, meaning that even if there is additional recognition of the loss, the extent is going to be minimal.
일부 제품에 대해서는, 경우에 따라서는 환위도 발생 가능하다고 생각을 하고 있고, 해당 경우가 발생을 한다면, 손익 회복에 긍정적인 효과로 작용할 것으로 생각하고 있습니다.
For certain products, there's also the possibility of reversal in the inventory valuation loss, in which case it is going to, of course, be positive in the company's recovery of the profitability.
두 번째 질문 주신, Kioxia와 WD 합병에 관한 내용에 대해서 답변을 드리도록 하겠습니다.
Now for your question on the potential consolidation between Kioxia and Western Digital.
다들 아시다시피 어떤 NAND 사업 자체가 지금 매우 어려운 사업 환경에서 지금 전개되고 있고, 그러다 보니 최근 일부 언론에서 Western Digital하고 Kioxia의 합병 추진 논의에 상당한 속도가 붙은 것처럼 언급되고 있기도 합니다.
Now, as we all know, the NAND business now is struggling in quite an adverse environment. Perhaps as a result of this, there have been some news reports about the potential consolidation between Western Digital and Kioxia and about how the negotiations are now picking up pace.
저희 회사는 해당 그 언급된 내용에 대해서 지속적으로 모니터링을 하고 있습니다마는, 아직까지 합병 관련해서 구체적인 어떤 조건 등이 확인된 바는 없습니다.
For the company, we are keeping an eye on the situation, we are very closely monitoring the situation, as far as we understand, there have been no specific terms of consolidation that have been agreed on.
현재 양사의 합병이 Kioxia에 어떠한 영향을 미칠지 등에 대해서 면밀히 검토를 하고 있고, 종합적으로 판단해서 저희의 입장을 결정할 예정입니다.
About what kind of impact the consolidation would have on Kioxia if it does happen, the company is also very closely reviewing the different possibilities and scenarios, and we will define our position after reviewing all the relevant information.
네, 다음 질문해 주십시오.
다음으로 질문해 주실 분은 UBS의 니콜라스 가우도이스 님입니다. The following 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 is: considering the upside in DRAM, higher density, DDR5 and HBM demand, how will you address requirements for more 1a nanometer and below capacity to support that? In particular, would you actually accelerate potentially technology migration in your capacity in Korea, such as M14 from here and into 2024? And, and possibly even considering some new capacity in M16 at some point next year. Could that therefore trigger an overall increase in CapEx next year as well for DRAM? Secondly, could you update us on your utilization rates reductions? I think you mentioned in the presentation that you took further steps for NAND flash in particular.
Could you try to for us to quantify this for both DRAM and NAND, where we are now, and where would you expect to be by the end of the year as well, for both DRAM and NAND flash, for capacity, for production reduction? Thank you.
네, 이, 첫 번째 질문은요. 그 최근 고용량 DDR5와 HBM의 수요 강세를 감안할 경우에 향후 1b 나노미터, 그리고 그 이후의 tech 전개 계획은 어떻게 되는지 하고요? 또 1b 나노미터 전환 가속화 혹은 M16 capa 확대까지 필요하지 않을는지, 만약에 그렇게 된다면 내년에 투자가 크게 늘지 않을는지가 이제 지금 첫 번째 질문이고요. 두 번째로는 이 NAND와 DRAM의 감산의 현황에 대해서 여쭤보고자 합니다. 이제 이미 이에 대해서 발표를 하신 바가 있는데요. 그것도 이제 앞서서 발표 때 잠깐 언급해 주셨지만, 이제 DRAM과 NAND와 관련하여서 감산은 어느 정도까지 하실 것인지? 연말이면은 이 각각의 생산 capa가 어느 정도 될 것이라고 예상하시는지요? 우선 질문 주셔서 감사합니다.
Well, first of all, thank you very much for your questions.
첫 번째 질문 주신 나노미터와 그 이후의 테크 전략과 그다음에 내년도 투자에 대해서 말씀을 드리도록 하겠습니다.
Let me respond to the first part of your question, which is about the 1a nanometer tech migration or tech development plan, as well as potential investment next year.
저희 회사는, 올해 늘어난 고용량 DDR5와 HBM 수요 대응을 위해서 필요한 테크별, capa 믹스들을 연초 대비 다소 조정하고, 다소 조정해서 운영하고를 하고 있습니다.
The company has already adjusted slightly the capacity mix by technology from the year, early part of the year planning. Again, I mean, we have already made adjustments to the planning that we have made from the early part of the year regarding the capacity mix by technology, so as to respond to the growing demand for high density DDR5 and HBM.
This year, for HBM mass production expansion, the necessary 1z nanometer proportion expansion, and next year, for HBM3E, which is for supply from the H1, we are preparing to secure 1z nanometer and tsv capa.
For this year, what we are focusing on is increasing the portion of 1z nanometer, which we need to increase the mass production of HBM. We are also preparing to increase the capacity for 1b nanometer and TSV capacity to prepare for the HBM3E that is to be supplied starting in the H1 of next year.
이를 위해 올해는 제한된 투자 범위 내에서 타 영역의, 투자분을 일부 절감하고, 효율적 운영을 통해서 HBM 제품 수요에 대응하고 있습니다.
To that end, we are reducing investment in other categories within the investment scope that is given for the year, again, to better respond to the growing demand for HBM. Looking ahead to next year, in order to maintain our technological competitiveness and to sufficiently respond to the continuous growth in demand, the minimum level of investment that we have for this year is probably not going to be sufficient next year. For now, the company's priority is on investing to increase mass production of HBM. Across the company, rather than capacity increase, we would focus more on tech migration, meaning that we would be focusing more on improving the efficiency of our CapEx. Moving on to the second part of your question, which was about the plan for production cuts.
The company has been undertaking production cuts from the Q4 of 2022, mostly for legacy products as well as low margin products. For the year 2023, the production for both DRAM and NAND are going to be lower year-over-year. Especially for NAND, as was explained earlier, its inventory level remains higher than that for DRAM, and its profitability margin is also quite sluggish. As a result, we have decided on an additional production cut of between 5% to 10%.
The following question will be presented by Giuni Lee from Goldman Sachs. Please go ahead with your question.
Thank you. I also have a question about investment. Of course, the company must continue with the investment in order to maintain its leadership in the rapidly growing HBM and DDR5 markets. Also because of the free cash flow and the balance sheet, having yet to recover, there are also some concerns that the company might potentially fall behind competition. In order to prepare for this, then are there any specific plans for funding or investment that the company currently is considering? If there are, then, if you could provide us with more details, then that would be appreciated.
Again, thank you very much for the questions. Allow me to respond to your question about investment and funding. As we all know, adverse market conditions are continuing.
The external uncertainties persist and customers' demand as well as prices remain soft. The company is making preparation and is trying to address the current situation by adjusting the mix around higher profitability and also reducing cost and reducing investment in facilities.
... 그리고 올해 급성장하고 있는 DDR5와 HBM3 제품의 고객 수요에 대응하기 위한 투자는 현재 차질 없이 집행하고 있습니다.
Also, I must emphasize that the investment necessary to respond to the customer's demand for DDR5 and HBM3 is underway as planned.
Uh, 투자 대응을 하고는 있지만, 현재 말씀드리고 있지만, 전사 투자 규모 자체는 기, 기존에 발표드린 대로 작년 대비 오십 퍼센트 이상 축소를, uh, 유지할 생각입니다.
Investment on that front, the DDR5 and HBM3 is underway. It's continuing, as I explained earlier, for the company on the whole, for the group on the whole, the investment scale is about 50% lower than last year's, which has already been announced, and this is going to stay.
연초에 선제적으로 그 자금 조달로 리소스를 확보를 했고, 2분기부터는 매출이 증가하면서 반면 투자금액 및 현금 지출 규모는 제한적인 만큼, 하반기에는 현금 흐름이 지속적으로 개선될 예정입니다.
Now, the company has already secured some resources from, preemptive, funding in the early part of the year. Given the fact that sales have been rising since the Q2 with a limited, investment as well as cash expenditure, we believe that the cash flow is going to continue to improve in the H2.
Uh, 따라서 하반기 자금운용은 일부 차입금 차환 발행 외에 상반기와 같은 대규모의 어떤 자금 조달 계획은 현재는 없습니다. 하지만 매크로 환경 및 업황에 따라서 유연하게 대응할 예정입니다.
Because of this, in terms of financing, in the H2, there is not going to be large scale financing that we have had in the H1, aside from part of the refinancing. Having said that, of course, we would be responding flexibly to any changes in the macro environment as well as the market situation.
The following question will be presented by Sung Kyu Kim from Daiwa Securities. Please go ahead with your question.
Yeah. 안녕하세요, 질문 기회 주셔서 감사합니다. 전 두 가지 질문이 있습니다. 첫 번째 질문은, 그 중국 생산 관련된 내용인데요. 그 미국의 반도체법, 그 CHIPS Act에 따라서 향후 중국 내 어떤 Capa 증설이라든지, 뭐 투자에 좀 제약이 예상되고 있는데요. 현재 지금 올해 October에 종료가 예상되는 그 1 year 유예 기간이, 뭐 추가적으로 연장이 가능한지, 지금 중국 쪽 그 생산 상황에 대한 업데이트 부탁드리겠습니다. 두 번째 질문은, 사업 관련해서 지금 DRAM 쪽은 말씀하신 대로 좀 AI 수요가 강세를 보이고 있고, HBM3라든지, DDR5 등, 우리 회사의 그런 기술 경쟁력 기반으로, 당초 생각했던 것보다 상당히 좀 조기에 좀 수익성이 반등할 것 같은 기대감이 있는데요. 반면 NAND 사업은 지금 적자 축소가 좀 쉽지 않고, 상당히 좀 어려움이 지속되고 있는 것 같은데, 이 NAND 사업의 좀 현황과 대응 방안에 대해서 좀 구체적인 설명 부탁드리겠습니다. 감사합니다.
I also have two questions. First is about the company's production in China. Now, with the CHIPS Act being enacted, it appears that there will be some constraints in increasing capacity in China and also investing in the Chinese facilities. In October, the 1-year grace period is going to be expired, and does the company believe that they can, it can be extended? Overall, the question is about an update on the company's production in China. The second question is now, looking at the different businesses, then yes, it is true that the AI demand for DRAM remains quite strong. The company is responding quite favorably to this thanks to its technological competitiveness in such products as HBM3 and DDR5.
It appears as if the company is set to recover its profitability earlier than expected. That's for DRAM, and looking at NAND, then it appears as if it is very difficult to narrow the losses, and it seems like the difficulties are continuing for NAND business. For the NAND business, how does the company plan to address this? If you could just give us an overall update about the NAND business as well?
Yeah. 먼저 질문 주셔서 감사합니다.
Thank you for the questions.
중국 장비 수출 유예 관련에 대한 업데이트를 먼저 좀 드리도록 하겠습니다.
Let me first give you an update about the grace period, about the restriction against the export of equipment to China. We consider the overall aspects such as long-term geographic risk, market demand, and efficiency in fab operations. We are currently looking into the various aspects of our plan in Chinese operations, looking at such factors as the long-term geographic risks, market demand, and efficiency in fab operations. To ensure stable business operations, the company is currently engaging in very close consultation with the different stakeholders, including the governments of different countries, in relation to the export of equipment as well as control thereof.
Allow me to respond to the question about NAND business.
As the questioner has rightly observed, for NAND, the impact or the benefit from the upside in surging demand for AI servers is less so than in DRAM. Across the industry, the investment level still remains relatively high. Again, the inventory level across the industry remains high for the NAND business. As a result, how the company plans to address this is, now, in addition to the production cuts that have been announced already, the company will undertake further production cuts in order to accelerate the normalization of inventory. Another area where we will work on is enhancing efficiency as well as reducing costs. Across the group, we will strengthen the management of CapEx and OpEx, and we will also try to reduce the further redundant costs between our NAND business and Solidigm.
For example, consolidating some of the redundant individual capabilities and simplifying some of the cost structure as well. In the H2, we will focus on improving the profitability by trying to sell more in response to real demand and improving the application mix. Also for the high value-add products like the eSSD 16-channel, we would also try to increase their mix out of the total products so that once there is improvement in the market, then we will be able to quickly make improvement on our profit and loss status. Also we will strengthen the competitiveness of SOC like controllers.
We will take one last question. The last question will be presented by Dong- Hee Han from SK Securities. Please go ahead with your question.
Thank you. I also have two questions. First is about the HBM. Now, HBM itself appears to be somewhat like a contract-based business. Does the company believe that based on this business model, the company will have some bargaining power in terms of the pricing? The second question is, now, about the HBM demand. With the demand for HBM rising very sharply, then it is likely that it could also smooth out the profit variability at different cycles? Does the company also believe that this would have an impact on the DRAM market share as well?
I see that HBM is the focus of intense attention today. I believe that the question was about whether the company will be able to maintain its price bargaining power for the long term.
For a product like HBM, which is a System- in- Package type, from the very beginning of product planning, we would have close collaboration with the customers as well as the related companies for development cooperation. It would also require large scale R&D resource input, and meaning that it is the kind of business that would require dedicated investment. In terms, we would then also look at the appropriate ROI across the life cycle, which I mentioned was around 2 years. This negotiation is conducted almost on a yearly basis. In other device businesses, there is a separate validation process, meaning that the process company would validate the memory ahead of time.
For the HBM that goes to the accelerators, then as I mentioned earlier, the planning is done with the customers, meaning that the planning as well as the validation pretty much occur at the same time. The capacity within the value chain is also predetermined, so this is conducive for such long-range planning. Down the road, as we move on to HBM4, the ecosystem is likely to become even more complex. Yes, the business model that is different from the off-the-shelf devices is likely to be maintained. This also means that given its nature of that is similar to a contract-based business, then the question was whether it is going to smooth out the profit variability.
We believe that the answer would be yes, given its contract-based business nature as well as the pricing structure. About the second question on whether this will also have an impact on the overall DRAM market share. On one hand, growth inside the market is almost a given, it depends on the company's product competitiveness as well as the supply competitiveness. If we maintain our competitiveness on both fronts, then I believe, yes, it would also have a positive impact on the company's market share and even the bit share.
Thank you very much. That concludes the SK hynix 2023 Q2 earnings release conference call. Thank you very much for your participation.