Good morning. Thank you for participating today. We will begin our SK hynix 2022 fourth quarter earnings release conference call. After the presentation by SK hynix, we will be having a Q&A session. To ask a question, please press star one on your phone. Our earnings release today will be interpreted simultaneously, while our Q&A session as consecutive. We will now begin our SK hynix earnings release presentation.
Good morning. Good afternoon and evening to those calling from abroad. This is Park Sung -hwan, Head of IR at SK hynix. Welcome to the SK hynix 2022 Fourth Quarter Earnings Release Conference Call. Before starting the conference call, allow me to introduce the executives present here today. First, CFO Kim Woo-hyun, who will be presenting today, Kwak Myeong-soo, Head of DRAM Marketing, and Park Chan-dong, Head of NAND Marketing.
A reminder that all earnings results and outlooks presented by the company today are subject to change depending on the macroeconomic and market circumstances. We will now begin SK hynix earnings release conference call for the fourth quarter as well as full year of 2022. Mr. Kim will first present on the earnings, followed by the company's plan and outlook.
Good morning, everyone. This is CFO Kim Woo-hyun. Let me begin today's presentation with the company's performance for the fourth quarter of 2022. The fourth quarter saw ongoing uncertainties in the macroeconomic environment due to high inflation and elevated interest rates. The rapid decline in consumer sentiment has driven down IT demand, customers' effort to adjust inventory and to cut costs have continued to weigh negatively on memory demand.
Amidst a difficult market environment, SK hynix has tried to maximize sales across all end applications. High inventory levels and escalating price competitions have led to steep decline in DRAM and NAND prices, and accordingly, our fourth quarter revenue was recorded at KRW 7.70 trillion, down 30% quarter-over-quarter and 38% year-over-year. For DRAM, bit shipment growth was flat sequentially in line with our guidance on back of greater sales centered on new products. NAND shipment growth was higher than guidance, growing by high single digits sequentially, supported by demand for new mobile products and greater sales of data center SSDs. Prices are declining at a pace that has not been seen since fourth quarter of 2008, leading to a sequential decline in revenue.
Fourth quarter operating loss stood at KRW 1.7 trillion, an operating margin of negative 22% owing to sales reduction and further inventory valuation loss as a result of rapid price drop. Depreciation and amortization in Q4 was KRW 3.69 trillion, increasing marginally Q-on-Q. EBITDA was KRW 1.98 trillion, while EBITDA margin was 26%. There was non-operating loss of KRW 2.52 trillion. This includes net interest expense of KRW 0.16 trillion, net foreign currency related loss of KRW 0.21 trillion, valuation loss of KRW 0.62 trillion on financial assets, including Kioxia investment and impairment loss on NAND-related intangible assets of KRW 1.55 trillion.
Net pre-tax loss was KRW 4.22 trillion, while net loss was KRW 3.52 trillion with KRW 0.7 trillion of reverse corporate tax expense as the company turned into loss in Q4. Net profit margin recorded negative 46% for the period. I will now report on the company's financial performance for 2022. Despite an unprecedentedly uncertain market condition in 2022, SK hynix recorded consolidated revenue of KRW 44.6 trillion, KRW 1.7 trillion higher amount than that of 2021. For DRAM, we have proactively ramped up sales of high-density products for PCs and servers. We have strengthened our sales and validation activities of products for high-growth areas such as DDR5 and HBM in response to rising demand for premium products following advancements in computing environments like AI and cloud.
In particular, our next generation strategic product, HBM, has maintained unmatched market share, drawing on its industry-leading product quality. For NAND, cost competitiveness has improved on back of quick ramp-up of our leading 176-layer products, while sales of eSSD has increased by 4-fold year-over-year, driven by stronger product competitiveness and expanded customer base. Operating profits for the year was KRW 7.0 trillion, a decrease by 44% year-over-year due to the rapid decline in demand for memory products during the second half of the year. The company's cash and short-term investments stood at KRW 6.4 trillion at the end of 2022, a decrease by KRW 2.3 trillion from a year ago, while interest-bearing debt was KRW 23.0 trillion, up KRW 5.4 trillion compared to that of a year ago.
Debt to equity ratio and net debt to equity ratio at the end of 2022 was 36% and 26%, respectively. While operating profit declined year-over-year in 2022, annual spending in equipment and infrastructure has risen, the company's free cash flow, which is cash flow from operating activities, less acquisition of PPE, was negative KRW 4.2 trillion for the year. Dividend for the fourth quarter will be KRW 300 per share, while accumulated dividend per share for 2022 will be KRW 1,200. I will now move on to the company's market outlook and future plans.
In 2022, global supply chain disruptions sparked by pandemic and geopolitical tensions, along with deteriorating macroeconomic environment due to interest rate hikes to address inflation, have led to stark decline in memory demand since the second half of the year. In response, suppliers have started CapEx reduction as well as utilization rate cuts. As there is a time lag for the effects to materialize, the imbalance between supply and demand of memory semiconductors has aggravated and led to hikes in inventory levels. The decisive factor behind demand recovery will be macroeconomic conditions, demand growth momentum this year is expected to be stronger that of last year, with rising memory usage on back of price elasticity as memory prices have already declined more than 50% since peak levels.
While the inventory levels across the industry are expected to peak during the first quarter, the effect of suppliers' response towards market condition will start to materialize from the first quarter, which will gradually bring inventory level lower and eventually improve supply and demand conditions towards the second half. Looking at demand by application, PC shipment is expected to decline again this year, but DRAM contents for PC is expected to grow by over 10%, driven by lower sales of Chromebook and higher sales of high-spec laptops and gaming PCs. In addition, demand for client SSD is expected to grow at low 20% level on back of lower cost pressures and rising contents from greater supply of PCIe Gen 4. The mobile market saw shipment decline by more than 10% on the heels of weak smartphone demand, especially in the Greater China region.
The ongoing bearish consumer sentiment curbs expectations on this year's shipment growth, but we expect demand recovery from the second half as digestion of inventory in certain channels is underway and as impacts of Chinese reopening and economic stimulus measures unfold. The smartphone market will remain polarized this year in terms of memory content. Performance-oriented flagship models will continue to see increasing density driven by greater adoption of LPDDR5X and UFS 4.0 and competition over market share. In contrast, low- to mid-end models will see slower content growth. Continuing trend of adopting more discrete NAND over MCPs. Server market is projected to see slowing demand growth due to tightened corporate IT investment, spurred by concerns of economic contraction and inventory adjustments by CSPs. The launch of long-awaited CPU will bump up demand of high spec servers adopting DDR5 in the second half of the year.
Increased content from falling memory prices will lead to this year's demand for server DRAM to grow by high 10s% and eSSD by high 30%. Demand growth in terms of system build is expected to be low 10s% for DRAM and low 20% for NAND. In terms of DRAM and NAND shipments, we will respond with flexibility to meet the demand growth level as we have sufficient inventory. Due to low seasonality in the first quarter, and with elevated level of inventory across the industry, demand is expected to contract more than the usual seasonality. We are planning a double-digit percent Q-on-Q decline in DRAM bit growth and high single-digit percent Q-on-Q decline for NAND in the first quarter.
With memory prices falling steeply and profitability quickly deteriorating, suppliers are taking actions to bring balance in the demand and supply situation. Taking such market conditions in consideration, the company will reduce this year's capital spending by more than 50% from the KRW 19 trillion of CapEx last year. Meanwhile, in order to be ready for the coming upturn, we will be continuing critical investments to ramp up our new products, such as DDR5 and LPDDR5 and HBM3. That will be the demand drivers, as well as investments in R&D and infrastructure to prepare for the future growth. In addition, by maximizing equipment efficiency, we will strive to maintain, even in upturns, the CapEx efficiency that we have obtained in this downturn.
In order to normalize the inventory level and bring forward the balance of demand and supply, we have lowered our wafer inputs in some of the legacy and lower profitable products in the fourth quarter. Considering the natural wafer loss from the technology migration, our wafer production in DRAM and NAND will be reduced compared to that of last year. Together with slower tech migration, we expect negative bit production growth for DRAM and minimal growth for NAND this year. Once the industry's efforts to lower production growth takes effect from the first quarter and production capabilities are reduced following the CapEx costs, we expect that not only will inventory levels normalize this year, but also a better-than-expected upturn may come next year.
Therefore, despite the current challenges in the market, we will remain dedicated to developing our technology and products, which will form the foundation of our long-term growth and further solidify our position as an industry leader. As of 2022 year-end, shares of our main products of 1a-nanometer DRAM and 176-layer NAND reached 20% and 60% of the total production respectively. Several of our 1a-nanometer and 176-layer products have already reached mature yield levels, and even new products have reached stable yield rates, enabling us to increase mass production once demand improves.
Although production proportion expansion of 1a-nanometer and 176-layer products are expected to be limited this year due to reduced CapEx, we will maintain industry-leading product competitiveness by securing readiness for next generation 1b-nanometer and 238-layer mass production by mid-2023. Growth of DDR5 market, where we have competitive advantage, is expected to accelerate in 2023. The company will advance its place in the DDR5 market with a full lineup of industry's first Intel-validated 1a-nanometer-based DDR5 products, including high density, 16 Gb and 24 Gb products. Furthermore, the company has developed and provided samples of world's fastest LPDDR5 Turbo, which are to be adopted in smartphone flagship models, and is planning to start mass production in the second half based on 1a-nanometer node.
This product operates at a data rate of 9.6 Gbps, faster than the previous maximum speed of 8.5 GBps. In combination with its ultra-low power consumption, it will enable a differentiated performance for customer products. Last but not least, will be the company's ESG management activities. The company has established the sustainability reporting system, which provides comprehensive ESG related data. More than 500 types of digitized, visualized ESG data that have been accumulated over 4 years are made available on our company's website. We have also added to the accessibility of our sustainability report, as it can now be downloaded as per selected key sections. SK hynix will continue to respond in advance to the increasing ESG needs of our stakeholders and to fulfilling our reporting responsibilities.
In efforts to achieve net zero by 2050, the company is strengthening joint efforts across the industry. In this regard, we have joined as a founding member of the Semiconductor Climate Consortium, newly established by Semiconductor Equipment and Materials International last November. SCC is the first global consultative body focusing on reducing GHG emissions across the semiconductor value chain, and is joined by major global ICT companies representing various sectors of the ecosystem. SK hynix will continue to remain committed to joint effort in reducing GHG emissions and will annually report our performance with transparency as part of our effort to strengthen our ESG management. Since the second half of last year, the memory market has been going through unprecedented challenging conditions.
The company will strive to further strengthen our position in mobile and cloud memory markets, as well as finding new growth drivers by securing automotive and AI customers. Taking from our experiences of advancing further by overcoming repeated challenges, we will aim to firmly establish our position as a global leading semiconductor company. Thank you.
With that, we will now begin our 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 the number star one. For cancellation, please press star two, that is star and two on your phone. [Foreign language]The first question will be provided by Hyunwoo Do from NH Investment & Securities. Please go ahead with your question.
Good morning. I have two questions. The first one is, if we look back in the past, we would say that since 2012, it would have been the company's consistent logic that industry consolidation has allowed through cycle profitability over the years, and we also have agreed with that logic in the past. However, in this present downturn, we wonder if that logic still held true, whether this downturn would have proven to be so severe as we are seeing at present. In the company's viewpoint, what would be some of the major factors that may have contributed to aggravating this downturn as we see it? The second question is about EUV. The company plans to deploy EUV in full scale since the one beta or 1b-nanometer.
Compared to your competitor, which has been deploying EUV since the 16 nanometer node, do you see any significant tech gap?
Between your EUV process technologies and the competitors, if so, how much time do you expect to take to catch up on the EUV learning curve to your competitor?
Let me first take your question. First question on industry consolidation and the relation to the down cycle. To cite the most recent down cycle after industry consolidation has been established, we can refer to the 2019 down cycle which happened after the significant cloud boom as you may recall in the years past. Back then we would ascribe that downturn as a function of memory supply side volatility.
If you look at the present downturn, it is not a function of only volatility in the memory side, but rather a more broad and persistent volatility that is present in the macro environment, which is compounded by the geopolitical issues that we see on the landscape today. We would say that the present downturn is rather a broader down cycle that is affecting the overall IT industry and ecosystems, not just the memory industry. While the duration of this down cycle is indeed an important matter for the company as well, the direction going forward is where we have our greater focus on. As you would know, the demand until today has been driven mostly by hardware devices, including PCs and smartphones. Into the future, we expect more to come from data-driven growth.
That is growth that comes from the increasing proliferation of tech and the platforms. We believe that there will be sustained momentum from such data, tech, and platform-driven growth. Meanwhile, on the supply side, there is a clear trend of deceleration in the overall supply capacity and scaling up. That would mean beyond the medium term, we continue to believe that the industry consolidation will continue to contribute to returning supply and demand to a greater balance over the years. First I would like to say thank you for the question, and now I will take your second one on EUV deployment. First, I would like to note that it's different for each DRAM player in terms of their technology considerations as well as their product development roadmaps and production strategy.
Those factors will play a part in determining the different timing of EUV deployment for each and every DRAM player. I would say that it is difficult naturally to determine which is the exact and correct timing for EUV deployment, and that is indeed the reality that is playing out. The company, in consideration of the very high investment required of EUV machinery as well as the maturing of related technologies, has focused to this time on maximizing the efficiency we can gain from EUV processes. As a result, we have now achieved industry-leading productivity in terms of EUV use.
이를 통해서 1a-nanometer 밑에서는 원가 절감 효과가 가장 큰 공정 한계에 대해서 EUV를 적용하고 있고요. 기존 DUV 대비해서 공정 스텝 수를 줄이고 양산임과 어떤 수율 안정화를 통해서 EUV를 적용하지 않는 이전 tech 수준의 원가 절감률을 달성하였습니다.
As part of this approach, the company has taken one process on the 1a-nanometer node that would have the most significant cost down effects from using EUV, and went on to deploy EUV machines on that single process. As a result, we have seen a reduction in the required process steps compared to when we used traditional DUV. As a result of further deployment in mass production as well as yield stabilization, we have been able to achieve with EUV cost downs that are equivalent to what we saw on the prior node.
EUV 적용을 어디까지 얼마를 적용한다고 단언을 할 수 있는 상황은 아니지만 일단 1B 나노미터나 1C 나노미터에서 EUV 적용 개수를 점진적으로 늘려나갈 계획을 가지고 있습니다.
While it is difficult to say for certain the extent to which we will be deploying EUV on a full scale, we can say at this moment that on the following 1b-nanometer and 1c-nanometer, we will begin to gradually expand the processes that do apply and use EUV.
공정 전환에 따른 EUV 적용에 대한 장비에 대해서도 최적화하고 높은 효율성을 유지할 계획을 가지고 있습니다.
We also remain focused on the optimization of the EUV equipment deployment for necessary for process transitions and also in continued efficiency gains.
네, 다음 질문해 주시죠. 다음으로 질문해 주실 분은 다이와 캐피털 마켓의 SK 킴 님입니다. The following question will be presented by SK Kim from Daiwa Capital Markets. Please go ahead with your question.
네, 안녕하세요. 질문 기회 주셔서 감사합니다. 저는 수요 관련해서 두 가지 질문이 있습니다. 그 지금 응용처 전반적으로 지금 재고 조정이 지속되면서 이 시장에서 이런 우려가 지속되고 있는 것 같은데요. 그 첫 번째로는 그 중국이 지금 리오프닝을 하면서 그리고 그 모바일 쪽의 업체들에도 재고 조정도 어느 정도 완료된 것으로 보이는데요. 중국 모바일 수요가 개선되는 어떤 시그널이 있는지 그런 현황과 그리고 두 번째는 그 데이터센터 수요가 전반적으로 지금 어떻게 보고 계시는지, 지금 전반적인 수요 상황과 향후 전망을 포함해서 설명해 주시면 감사하겠습니다.
Thank you for taking my question. I have 2 on demand. First, we are now seeing inventory correction happening across customer applications, which is causing concern in the market. That brings me to my 1st question, which is, as China's economy is reopening, we are also hearing news that Chinese mobile brands are close to completing their inventory adjustments. In light of these developments, does the company see any particular signals that are signing towards a true recovery in Chinese mobile demand? The 2nd question is, we want to know your view on overall data center demand. Both questions would be directed at the demand picture that you're seeing at the moment, as well as your outlook on how it will evolve going into the future.
네, 질문 주신 내용 말씀드리겠습니다. 우선 그 모바일 시장 관련돼서 뭐 아까 이제 첫 질문에 답드리면서 말씀드렸지만 올해의 수요 예측은 사실은 이제 매크로 volatility의 관점에서 상당히 뭐 이 변폭이 크다고 보여집니다. 특히 이제 스마트폰 같은 경우에는 전년 대비해서 globally 그리고 중국만 놓고 보더라도 저희가 flat 정도의 shipment 성장을 현재로서는 보고 있는데 이 부분은 뭐 경우에 따라서는 좀 변화의 range가 좀 있지 않나 이렇게 보고 있습니다.
Thank you for the question. To answer your question on the mobile side, I have touched upon this briefly in my last response, but, we want to note that the demand predictions for the current year would have significant volatility. They may be subject to significant change because of the macroeconomic uncertainties that we are seeing in the landscape, especially for smartphones. If you look at the global situation, and if you just look at China, we are projecting currently flat smartphone shipment growth for the year. Depending on how macro factors evolve, we believe there is room for change to these current projections.
서버의 경우에도 마찬가지입니다. 현재 높은 불확실성과 경기 둔화 우려 때문에 빅테크 기업들도 이제 투자 절감 및 비용 절감을 하고 있어서 당분간은 아마도 이런 관점에서 수요는 재고 소진에 우선 집중될 걸로 보이고 있습니다.
The situation is similar for the server side. Due to high uncertainty as well as concerns for an economic slowdown, even the big tech firms are reducing or cutting their spending and prioritizing inventory adjustment for first and foremost. Still, we do see some opportunity factors present. Some potential upside could be, for example, on the Chinese mobile side after China's economy truly reopens. If there are any policy changes, for example, any new subsidies given out to new smartphones, we believe that will have an upside effect on the new smartphone models that are to launch in the second half this year, specifically for our side in terms of a boost for the high density POP components.
Server. You would be well aware of the development on the server side, that is the launch of the new CPUs, which are clearly set to be an opportunity for the market, and that will serve to boost demand for high density DDR5 products. If you look at the buildup trends and the refresh cycle on the data center side, we believe that given that the time frame between this year and the next, between 2023 and 2024 would be the time when data centers begin to launch their new refresh cycle for the existing servers they have on site. We believe all in all this presents a new opportunity for existing product lineups, and that will be all the more reinforced by the launch of the new CPUs, which are again a clear market opportunity.
Even our customer feedback leads us to understand that our customers are also cautious and conservative on the first half outlook. On the second half of the year, our customers are also saying they are seeing stronger demand momentum compared to the year prior. We believe that a bulk of our customers demand will be focused and concentrated on the second half of the year and believe this is an opportunity truly that the company has to capture on.
The following question will be presented by J.J. Park from J.P. Morgan. Please go ahead with your question.
[Foreign language]
Thank you. I have two questions. First, the company has announced a close to 50% reduction in CapEx for this year. Do you see any further downside revision risk to that number? That is, do you have any intent to make further CapEx cuts? The second question is, I have recognized that the company has reported an over KRW 1.5 trillion in impairment losses related to NAND related intangible assets. I found those numbers in the PPT, they were recorded as not operating losses, but instead as part of the non-operating expenses. On these impairment charges on the NAND intangible assets, could you offer some more detail perhaps?
Thanks for the question. I'll take the first one on whether the company plans any further CapEx cuts beyond what has been announced.
As was said in the Q3 2022 earnings call, I wish to reiterate the message from that call today that the announced CapEx cuts already result in a more than 50% reduction to our investment year-on-year for 2023.
Given that the CapEx spending after our announced cuts arrive at a significantly smaller level if we consider the fab scale and the critical infrastructure that we need to maintain and invest in. These CapEx spending for this year will already be significantly down, a record low. That means given that at the present moment, the company is not considering further CapEx cuts beyond what has been already announced.
The basic factors or basis for our decisions on any revisions to CapEx plans would be, of course, market developments, but at the moment, we are not seeing any significant market change on that, and that would merit a change to our direction on CapEx. At the moment, given what we know right now, we are not considering any further CapEx cuts. While we are going to execute the significantly large CapEx cut announced for this year, we will continue to make investments into our products such as the DDR5 and HBM3 DRAM, which are set to rise in demand from this year onwards. Also our products built on the 1a-nanometer and 176-layer processes, which are our competitive, cost competitive offerings. On the investment side, we will ensure that we can secure full readiness to truly deploy these products to scale.
About your second question on the NAND related impairment charges.
[Foreign language]
In the second half of 2022, memory demand slowed down significantly and sharply, bringing down prices with them, and this truly deteriorated NAND profit margins.
[Foreign language]
In reflection of these market developments as well as higher interest rates, the company proceeded to reevaluate its assets on its balance sheet at the year-end. As a result of that asset revaluation, we incurred non-recurring costs in the fourth quarter.
[Foreign language]
One-off costs related to NAND were incurred in our Kioxia investments as well as in relation to Solidigm. For our Kioxia investments, the company conducted a fair value measurement of our Kioxia investments at the year-end, and as a result, incurred and recognized close to KRW 600 billion in losses from valuation.
[Foreign language]
The rest of the non-recurring costs as a result of the asset value revaluation would include the write-off of the goodwill that we assigned to Solidigm at the time of acquisition, as well as other NAND intangible assets that belong to Solidigm. Because of the write-off of these intangible assets in value, in line with the deterioration in the NAND market, we arrived at the non-recurring costs that we saw in the first quarter.
The following question will be presented by Simon Woo from Bank of America. Please go ahead with your question.
Yes. Thank you for the opportunity to ask my question. I am Dong-je Woo or Simon Woo of Bank of America, and thank you indeed for providing us insight into the company's strategy despite these difficult market circumstances. I wanna know, as you go ahead with your more than 50% in CapEx cuts, I wonder because CapEx is an important leading indicator of how the company will build out its competitiveness for the ensuing years, including the next one. My concern is by going ahead with this CapEx cut, wouldn't this be affecting your future competitiveness and potential? What would be your strategy for the medium to long term to potentially prepare yourself even amidst such CapEx reduction?
I would like to first and foremost note my gratitude towards your interest and concern, of course, for the company, and would like to take your question on the effect of CapEx cuts on our long-term technological competitiveness. Already, the company's main products built on the 1a-nanometer and 176-layer processes. The main products in our portfolio have already reached mature yield, while the new products that we are introducing on these same nodes are also exhibiting stable yield levels. The CapEx cuts this year will not have a material effect on altering the advanced node portion of our tech mix, but rather have some effect in containing the production growth. At the same time, we will continue to execute the CapEx spending necessary for our next generation process technologies, including the 1b-nanometer and 238-layer nodes.
We will continue to spend on the development of these process technologies, as well as ensure that mass production can proceed early. To say that again, we will ensure that we do spend to ensure that pilot runs to ensure mass production readiness can proceed on track for these next generation technologies so as to prepare ourselves for the market to come in 2024. The 1b-nanometer process exhibits more than 40% net die efficiency gains compared to our prior 1a-nanometer node. It goes to show 1b-nanometer's cost competitiveness. On top of that, our 238-layer process already has had the core products developed, and within this year, we plan to introduce a 1 terabit device on 238-layer stack that will yield close to 50% in net die efficiency gains.
Going forward, the company will continue to prepare ourselves by continuing to enhance our investment efficiency as well as preparations for future core competitiveness building so as to prepare ourselves truly well for the next upturn.
네, 다음 질문해 주시죠.
다음으로 질문해 주실 분은 HSBC의 Ricky Seo님입니다. The following question will be presented by Ricky Seo from HSBC. Please go ahead with your question.
네, 질문 받아주셔서 감사드립니다. 그 재고 관련해서 한 가지 질문드리겠는데요. 지금 그 단기적인 그 메모리 가격의 재고 트렌드가 가장 중요할 것 같은데, 저희 회사에서 파악하고 계시는 고객사별로, 키 애플리케이션 고객사별로 재고 수준이 사분기 말에 어느 정도였는지, 삼분기 대비해서 어느 정도 트렌드를, 어떤 트렌드를 보이고 있는지 설명해 주시면 감사드리겠고요. 향후에 그 이제 그 프로덕션 컷 하는 게 이제 임팩트가 나타나기 시작될 텐데, 어떤 어플리케이션에서 가장 빠르게 재고가 하락할 걸로 예상하시는지 설명 부탁드리겠습니다. 감사합니다.
Thanks for taking my 2 questions, both on the inventory side. Inventory, we would say, is the most critical factor in determining short term and near term memory pricing. Given that fact, I want to know the company's view. What's your visibility? What are you seeing on the customer side? Across key applications and across customers, what was the exiting inventory level coming out of Q4 2022? How did Q4 inventory levels change compared to the prior Q3? The second question is, as we expect the effect of your production cuts to materialize going forward, which applications does the company see as going to be the first to see inventory levels meaningfully come down?
네, 그 질문 주신 그 업계 및 고객 재고 관련된 두 가지 질문인데, 하나로 다 묶어서 답을 드리도록 하겠습니다.
You've asked question on the inventory side for both the industry inventories and the customer side, and I will try to answer that in a single response.
앞서도 얘기가 나왔었지만 전반적인 고객 재고는 지난 1990년대의 다운턴과 유사한 상황으로 좀 보여지고요, 여기에 이제 공급사 재고까지 같이 보면, 업계 전반의 재고는 아마도 이제 사상 최대 수준인 것 같다, 이렇게 판단하고 있습니다.
As noted earlier, our customer inventories are close to levels that we observed during the 2019 down cycle and supplier inventories combined that would bring the overall industry inventory level at a record high.
응용별로 보면 아무래도 그 스마트폰 관련된 모바일 쪽, 이 상대적으로 재고 수준은 좀 그나마 낮은 것 같고, 상대적으로 낮은 것 같고요. 그다음이 PC. 반면에 서버 같은 경우에는 재고 수준이 업의 특성을 감안했을 때 다른 consumer oriented 시장보다는 좀 상당히 높은 걸로 이렇게 이해를 하고 있습니다.
To break it down by application, I would say the mobile, including smartphone side to have relatively low inventory levels compared to other applications, relatively speaking. The next would be PC in terms of the inventory correction progress that we are observing across applications. It's a slightly different picture for server because if you consider the nature of the server and data center industry, their inventory is currently high, relatively high compared to the more consumer-oriented segments.
업계, 뭐 저희뿐만이 아니고 안팎에서 이제 그러다 보니까 규모 관점에서 우려가 가장 큰 서버 같은 경우에는 조금 더 들여다봐야 될 부분이 있습니다. 뭐냐 하면, 앞서도 이제 하반기 수요 모멘텀의 그 한 부분이 이제 신규 CPU 및 DDR5라고 말씀을 드렸는데요. DDR5는 사실 지금 업계에 재고가 없다라고 봐야 될 것 같습니다. 이제 재고 빌드가 돼야 되고 그다음에 이 분기부터 이제 그 관련 서버 세트들도 빌드를 해야 되고. 결국은 DDR4와 DDR5로 갈라서 본다면, 서버 재고 부담은 DDR4에 집중되어 있고, 한편으로 DDR5는 이제 저희가 늘려가야 되는 상황입니다. 오히려 그런 면에서는 재고를 우려할 수준이 아니고 저희가 공급량을 늘려야 되는 상황으로 되어 있고요. 그러다 보니 이제 저희 그 믹스 전략의 핵심은 DDR4를 좀 줄이고 그다음에 DDR5 믹스를 고객들하고 적극적으로 늘려가는 그러한 방향이 중요할 것 같습니다.
Given all the news that we're hearing, and this is not just an outlook coming from the company, it merits close attention to really closely monitor and look deeper into the situation in terms of inventory on the server side because of the sheer size that we're currently seeing. As was noted earlier, the key driving factors of momentum into age 23 would be the new CPU introductions as well as DDR5. If we take DDR5, currently the industry does not have enough inventory of DDR5. Inventory levels are quite low, and we are at the stage where the industry has to build up DDR5 inventory. In relation to that, we expect that from Q2 this year onwards, the DDR5 equipping server setShipment builds will also begin to grow. Related DDR5 server builds will grow from Q2 onwards.
If you look at the server, DRAM inventory, most of it is focused on the DDR4 technology at the present moment. While for DDR5, we are in a situation where we have to begin building up the inventory. We would say that, the DDR5 for server inventory is not at concerning levels. Rather on the contrary, suppliers have to begin building up the DDR5 inventory for the industry. In light of that, the company's strategy in terms of product mix is to gradually reduce the share of DDR4 in our server DRAM mix, while closely engaging with our customers to gradually expand the DDR5 mix.
To sum up, if we are to see all the drivers that we mentioned in this response really transpire, that is, if we do see the slight demand momentum returning in the second half of the year-over-year basis, and also on the supply side, if we do see the effects of the CapEx adjustments as well as a more efficient mixing of our DRAM product modes through engagement with customers, if these do transpire successfully, we believe that towards the second half of the year, inventory will become much healthier on both sides for both the customer and the supplier side. The following question will be presented by Myeong-seop Song from HI Investment & Securities. Please go ahead with your question. Thanks for taking my two questions.
First is, I wonder that in light of your announcement that you will be reducing CapEx by a significant margin this year, while we are also seeing that you exited the year with negative free cash flow as well as significantly higher borrowings. Given these financial developments, well, I wonder if the company is considering any possibility of a change to your dividend policy, your broader shareholder return policy. Of course, I understand that this would depend on how your finances evolve going into the future, how your cash balance evolves. It would might be difficult to say at this point, but there is a real concern in the market that if this year's results also turn out to be much worse than expectations, the company might have to consider conducting equity capital raisings.
Given that that is a real concern being raised in the market, I want to know what the company's views and plans are on this side. The second question is about your Q1 bit growth projection. You have projected that both for DRAM and NAND, there would be significant declines in bit growth for the first quarter. Your competitor has announced that for the Q1, both DRAM and NAND bit growth would still be at the low single digit. Whereas the company has had higher bit growth than the competitor in the fourth quarter of last year, although of course you were lower in Q3, I wonder what could explain this significant bit growth projection gap between you and the competitor for the first quarter?
Let me answer your first question about any changes to our shareholder return policy and any potential for equity capital raisings. As you would also be aware, what we saw last year was that compared to our expectations early in the year, the market unexpectedly deteriorated very sharply, resulting in us exiting the year with negative free cash flow. This speaks to the persistent volatility in the memory industry that we are, of course, continuing to see today. While we are seeing the external uncertainty further heightened this year and the difficult market conditions also persisting, the company is responding to these circumstances by making ambitious CapEx cuts as well as reduction to our expenses overall. With these efforts, we will continue to strive to create a positive free cash flow going into the future. Currently, we are not planning any changes to our shareholder return policy.
Given that for the near term, there are persisting macro and memory industry-related uncertainties, the company's policy on our overall cash stock is to carry a more stable level of safety cash balance for the near term to navigate these elevated uncertainties. For the long term, we will continue to be disciplined in our CapEx spending so as to bring free cash flow back to positive territory and also at the same time, gradually reduce the size of our borrowings. As for equity capital raisings, we are currently not considering equity capital raisings as a financing instrument or tool at the moment. I'll take your second question on the difference in the bit growth estimate for the Q1. I don't think I would have to really split this into a discussion for a DRAM and NAND separately.
For all of DRAM and NAND, we have projected a very conservative estimate for the bit growth outlook for Q1 because that is based on what we're hearing from the customers in our talks. This bit growth estimate is modeled closely to the true or natural demand that we are sensing from our customers out in the field. On top of that, it is also aligned with the Q1 demand outlook that the company currently has. Our market share with strategic customers has not changed for the first quarter of the year, and we believe this would be a more important indicator rather than the slight changes to the bit growth estimate. Overall, market share remains unchanged.
To go into a little more detail, there would be two potential reasons that can explain the difference in the estimates for bit growth. The first would be the fact that for each supplier, the customer portfolio is different. If we were to also factor in the different market share that each supplier has amongst this different base of customers, then that would be material enough to really change or alter the bit growth estimates as we are seeing in the Q1. For example, if you consider a supplier that not just provides memory but also smartphones, then the smartphone business would help to get the supplier stronger demand in the Q1 helped by the smartphone business. That would, of course, contribute to a higher bit growth estimate on their end, whereas in contrast, SK hynix is not that case.
To add a bit more detail on the company's practice, well, SK hynix of course engages with our strategic customers in yearly negotiations on the yearly volume and pricing deals. At the for each year, we plan out the business with our customers and execute according to the terms of our yearly negotiations. Of course, there is some flexibility when it comes to the quarterly actual implementation, depending on the market circumstances then. Of course, the overarching aim would be, of course, to maintain our market share with the strategic customers. Since it is difficult for us to predict the market situation for the first quarter, we are sticking to that broader aim of maintaining our market share with our strategic customers and flexibly adjusting as the market evolves.
The following question will be presented by Marcus Shin from Mizuho Securities. Please go ahead with your question.
Thanks for taking my question. Since there were many good questions asked about the company, I would like to pose one question about Solidigm. I believe that the revenue and results impact for the company were in part a function of Solidigm's performance also worsening. I would like to ask you for more color on Solidigm's Q4 2022 financial results, as well as your outlook for the full year 2023 revenues and profitability. If possible, could you provide perhaps a timing that you expect Solidigm to be able to return to turning normal levels of profit?
Thank you for your question. I will be answering your question on Solidigm's performance and our potential expected timing for Solidigm returning to be able to turn normal profit levels. Solidigm was not exempt from the unprecedented slowdown in memory demand that the company experienced.
From the 2H, the second half of 2022, Solidigm also began to see the top line declining at a similar pace with the company. As you would know, 2022 was the first year of the Solidigm acquisition, which entailed various standup costs as well as costs incurred as a result of acquisition accounting such as the purchase price allocation at the time of acquisition. As a result, these non-recurring costs were incurred and also further impacted Solidigm's performance. For the near term at least, we believe impact and headwinds would be inevitable on Solidigm's revenue and profitability given that the NAND market downturn is continuing. We have to note that for the full year 2022, we saw Solidigm's existing and current solution capabilities allowed the company and Solidigm to boost our overall product competitiveness in the market.
Thanks to Solidigm, the wider customer base allowed us to really see our Solidigm data center SSD sales and revenue grow by a significant margin for the year. Given the many external uncertainties that remain, the integration going forward will not be an easy process, but we will continue on this journey. If the company can under its original vision in acquiring Solidigm, continue to pursue our integration with the end of achieving economy of scale and also in line with our original intent, improve further our data center SSD capabilities, we believe that this could be a factor that helps us achieve a faster recovery than would have been possible in a usual NAND market downturn.
The last question will be presented by Sei Cheol Lee from Citigroup. Please go ahead with your question.
Thank you. I have two questions. First, we saw that inventory write down charges grew significantly in the fourth quarter. Could you provide an exact number on the Q4 charges related to inventory write downs as well as your estimate for Q1? One about demand. On the demand side, there is a lot of market interest going into recent AI technologies such as the ChatGPT chatbot. How would this affect memory demand exactly going into the future? Thank you. I will answer your first question about our inventory valuation loss or write down charges. In Q4, we observed that inventory levels grew while ASPs fell down compared to the prior quarter, which resulted in the company recognizing between KRW 600 billion and KRW 700 billion in inventory write down charges.
While we do believe that further inventory write downs may be possible in Q1 should ASPs fall further, it is difficult to provide you an estimate at this time because the exact amount would depend on the inventory levels then as well as the actual extent of the price falls.
다만 일분기에 어 사분기와 같은 어쨌든 급격한 가격 하락이 예상되지 않고 있기 때문에 어 규모가 사분기 대비 줄어들어지지 않을까 조심스럽게 예측하고 있습니다.
Since we do not expect the price declines in Q1 to be as substantial as in Q4, we cautiously project that the size of inventory write downs, if any, would be smaller in Q1.
예, 두 번째 질문 주신 그 ChatGPT 관련된 답변드리겠습니다.
Okay. Now I'll take your second question about ChatGPT.
어, 이 ChatGPT라고 이번에 이제 많이 회자되고 있는 이 어 내용이 결국은 AI 관련된 내용인데요. 어, 언어 모델의 확장성, 그리고 이게 더 중요한 건 대중을 상대로 한 어 인공지능 활용의 일반화 및 상용화라는 데에서 파급성이 크다라는 의미를 둘 수 있을 것 같습니다.
There is indeed lot of market talk about this ChatGPT system, which basically ties into the AI application. ChatGPT uses a large language model and its scalability as well as its commercialization and targeting of the mass public. Its usability for the mass public is what gives ChatGPT and similar algorithms such disruptive potential.
이런 그 소위 말하는 Generative AI 기술하고 그다음에 현재 웹의 검색 엔진이 결합되는 이런 활용, 그리고 향후 웹이 발전하는 뭐 예를 들면 Web 3.0 같은 이러한 그 기술적 진화에 따른 메모리 뿐만이 아니고 업계 전반의 이 활용 확장 가능성은 매우 크다고 믿고 있습니다.
We believe that if we combine potential use cases for ChatGPT and such Generative AI technologies with existing search engine technology, on top of that, we can also expect further broader evolution, such as the advent of Web 3.0. This would all combine to bring about a significant change and also offer scalability to many different use cases and the demand upside as well for not just the memory industry, but for the broader IT industry and ecosystem.
어, 이러한 이제 어 서비스를 위한 학습 과정에서 데이터가 텍스트만을 다루게 될 거는 아니고, 앞으로 이미지나 비디오 뭐 또는 생체신호까지 세상에 존재하는 모든 데이터 형식을 커버하는 멀티모달 AI로 진행할 거기 때문에 결국은 학습과 추론을 위한 서버 인프라 투자의 확대가 수반이 되어야만 합니다.
Since the training of these models would involve not just text data, but also all the forms of data that exist in the world, including image, video, and biometric data, would be utilized in the training of these multimodal models or networks. We believe that the investments into servers that are needed for the training and the inference stages would only continue to grow.
메모리 관점에서는 이제 중요한 게 two 가지 포인트인데요. 하나는 속도, second는 용량입니다. 이 결국은 속도가 핵심인데 그거를 지원하기 위해서는 병렬 처리를 위한 그 고성능 DRAM 및 그다음에 고성능 Computational Storage Device가 필요하게 될 겁니다.
There would be two factors relevant to memory in light of these developments. One would be speed, and the second would be memory capacity, and speed would be the more essential factor in this case. In order to support the higher computational speeds required of these models, we would have to support parallel, excuse me, parallel processing, by developing and marketing more high performance DRAM as well as storage that has computation functions.
결과적으로 HP, HBM 같은 고대역 메모리는 곧바로 활용이 되고 있고요. 그 기존 서버 메모리 중에서 특히 128 GB-class 이상의 모듈 수요가 빠르게 성장하고, 나중에 64 GB에서 128 GB로 넘어가는 크로스오버 시점도 당길 수 있는 그러한 이벤트라고 생각을 합니다.
Already we are seeing HBM that is memory with high bandwidth already in use for such AI and compute intensive applications. For the existing server memory offerings, we are particularly seeing strong demand growth in the 128 gigabyte and higher density server memory modules. We believe potentially the timing for the bit crossover of server memory modules from the 64 gigabyte config to the 128 gigabyte configuration can be pulled in. It could happen earlier with this development.
한편으로 그 NAND 관점에서는 그 기존의 저장 매체보다는 더 빠르고 어 강력한 성능의 그 이 Computing Power 지원을 해야 되기 때문에 결국은 이 부분이 어 QLC 기반의 병렬 처리용 SSD를 빠르게 개화시키는 그러한 매개체가 될 수도 있다고 보고 있습니다.
On demand side, since a faster performance and stronger performance would need to be supported beyond what current storage medium offers. In order to support that compute capability, we believe this development could potentially be one that pulls in and really primes the stage for deployment of QLC-based SSDs that are capable of parallel processing. With these technical developments, we believe that ultimately this will lead to the creation of the memory-centric architecture, which involves aforementioned technologies such as CXL, Compute Express Link, as well as processing and memory that is PIM, and the computational storage devices, and also memory pooling in systems. In the medium to long term, we are certain that this would serve as a growth engine for the memory side.
Even in the near term, we are already seeing some of our customers asking for more 128 GB or larger size server memory modules compared to the prior year. We believe this development is also not irrelevant to the current development. I believe this circles back to your earlier question about our CapEx reduction potentially affecting future technological competitiveness. Well, capacity would of course be one factor in determining our competitiveness, but we believe that in light of future trends, such as ChatGPT and the developments that we have described, what would be most critical, more critical even, would be the portfolio that we have, as well as the tech and product leadership that the company enjoys. In that aspect, SK hynix indeed is leading the industry on both DRAM and NAND fronts in terms of our technology readiness for the future.
For DDR5, as an example, we are already completely prepared to begin shipping the DDR5, begin marketing the DDR5. We're selling it into the new server CPUs, as can be seen from the fact that we have achieved validation from the CPU maker. You can see that we are currently leading in terms of the validation progress from this CPU maker. For our High Bandwidth Memory, HBM, we are currently maintaining an absolute lead over the market, over other competitors. We have the LPDDR5, the low power memory for which we have recently launched a turbo version or specification. LPDDR5T, which is its name, delivers more than 30% gains in the performance for power. With DDR5, HBM, and LPDDR5T, we have now established a full lineup to really respond to the future market.
We have the graphics memory, GDDR. We have long supported GDDR6 to the market, but now we are accelerating the transition to GDDR7 in order to maintain our leadership on this front. We can also mention that in the automotive market, we are really focusing on our LPDDR5 offerings to ensure fast and certain growth on this segment as well. Overall, regardless of what our short-term policy is on, is for CapEx, we can certainly say with confidence that we are doing all the preparations necessary to ensure our leadership into to capture future technology trends.
Thank you. This concludes the Q4 '22 conference call from SK hynix.