Good 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 2021 fourth quarter earnings release conference call.
[Non-English content]
Before starting the conference call, allow me to introduce the executives present here with me today. First, Kevin Noh, Chief Business Officer, Park Myeong Soo in charge of DRAM Marketing, and Park Chan Dong in charge of NAND Marketing, and the new Head of Finance, Kim Woo Hyun.
[Non-English content]
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. The CBO, Kevin Noh will first present the earnings for the fourth quarter in 2021, followed by the company's plan and market outlook.
[Non-English content]
Good morning.
I wish everyone and your family health and happiness in the New Year. 2021 saw an uncertain business environment continue with global supply chain disruptions, geopolitical risks, and a pandemic that remains unabated. DRAM sales actively increased in PC, graphics, and Chinese mobile phones where non-face-to-face demand was solid. Based on its quality competitiveness, we were able to maintain our position as market leader among major server customers. Furthermore, we already secured industry-leading quality competitiveness in next-generation strategic products such as DDR5 and HBM3, expected to be met with rapid demand growth, and began successful development and mass production of 1α nm technology-based products using EUV equipment. NAND recorded bit shipment growth of over 60% for the year, achieving the highest ever revenue and turning around to profit for the year, thanks to smooth production ramp-up of the 128-layer and outstanding product competitiveness.
In the mobile SSD market, by focusing on demand for high-density solution products, we drastically improved our business portfolio with more than six-fold increase in data center SSD sales and double in PC SSD in only two years.
[Non-English content]
Last December, we completed the first closing of Intel's NAND business acquisition, which will help improve the DRAM dependent business structure while adding growth momentum to our NAND business. We thus acquired Intel's SSD business and Dalian Fab. The acquired NAND business, which will be incorporated as a subsidiary, was launched on December 30th as Solidigm.
[Non-English content]
With this acquisition as a stepping stone, SK hynix will firmly position itself as a leader in the NAND industry by leading technological development and meeting customer needs with a broad product portfolio. Furthermore, the company will not only expand the market share, but also continue to evolve into a global memory solution provider with understanding and extensive know how in computing architecture.
[Non-English content]
I will now report on the company's financial performance in the fourth quarter 2021.
[Non-English content]
Consolidated sales in the fourth quarter was KRW 12.38 trillion, up 5% QoQ and 55% YoY, recording the highest quarterly sales ever for the company.
[Non-English content]
DRAM bit shipment grew by mid- to high-single-digit in line with the guidance, thanks to strong demand for servers and graphics, as well as for new mobile products and PCs. ASP fell by mid-single-digit. In particular, server DRAM sales were the highest since the record boom period of third quarter 2018, driving the overall sales growth.
[Non-English content]
NAND bit shipment growth was in the low teens percent on par with the guidance backed by data center SSD demand and increasing densities in mobile products. ASP fell by around 10% due to the component supply disruption and soft mobile demand in the Greater China Region. Despite the price decline, NAND remains on a fast-paced growth, with sales achieving record highs for three consecutive quarters, especially data center SSD reaching the highest ever quarterly and annual revenue.
[Non-English content]
Meanwhile, MCP's portion of the revenue fell from 24% to 19% as both bit shipment and price fell on the heels of slowing mobile demand from the Greater China Region. MCP revenue for the year was also a record high, growing 59% YoY.
[Non-English content]
The company's fourth quarter operating profit was KRW 4.22 trillion, up 1% QoQ and up 337% YoY. Operating profit margin was 34%.
[Non- English content]
As for profitability by product, for DRAM, despite the higher sales share of 1z-nm process and sales of premium products like GDDR6, cost reduction was not able to offset the price decrease due to the initial startup cost burden as mass production began in M16.
[Non-English content]
On the other hand, NAND achieved a double-digit cost reduction with sales portion of the 128-layer reaching 80%. This improved profitability following the previous quarter and helped achieve a profit turnaround for the year.
[Non-English content]
Depreciation and amortization in the fourth quarter was KRW 2.79 trillion, slightly increased from the previous quarter. EBITDA was KRW 7.01 trillion, and EBITDA margin 57%.
[Non-English content]
There was a net non-operating profit of KRW 0.45 trillion. The fair value valuation at year-end generated KRW 0.51 trillion gain from our investment in Kioxia. Net profit before tax was KRW 4.67 trillion, up 1% QoQ. Net profit was KRW 3.32 trillion with a net profit margin of 27%.
[Non-English content]
Consolidated cash balance at the end of the year was KRW 8.67 trillion, up by KRW 3.72 trillion from the previous year. Interest-bearing debt was KRW 17.62 trillion, increased by KRW 6.37 trillion YoY. The company's debt-to-equity ratio and net debt-to-equity ratio at year-end stood at 28% and 14% respectively.
[Non-Englsih content]
Despite the increase in CapEx, free cash flow, which is cash flow from operating activities, less acquisition of plants, property, and equipment, was KRW 7.31 trillion on the back of improved profitability. As a result, total dividend payout per the company's dividend policy will be KRW 1.06 trillion, higher by 32% from the previous year. Dividend per share was set at KRW 1,540. Next is the company's market outlook and plan. Memory market this year is expected to see supply chain issues continue through the first half, which is one of the most important variables for demand. We expect this to gradually ease as we move into the second half, improving the overall demand supply situation.
[Non-English content]
Looking at demand by application, the PC market, where shipment has grown more than 10% for two years in a row since the pandemic, is expected to see solid demand this year led by commercial PC. The set shipment is expected to be flat with growth rates slowing down due to the high base effect.
[Non-English content]
DRAM content per PC is expected to grow at a high single digit percent, with lower sales of the relatively low content Chromebook and higher sales of laptops and gaming PCs. Demand for client SSD is expected to be driven by the pent-up demand from supply disruptions and increased content in PCIe Gen 4. For the client SSD, it is expected to grow in the low 30%.
[Non-English content]
The smartphone market this year is expected to grow in the mid-single digit level, similar to last year's.
However, there will be growing polarization in memory content owing to the rising costs due to supply disruptions. While high-end models are increasingly adopting 8 GB or higher DRAM and 256 GB or higher NAND, and switching to LPDDR5, lower price models are expected to see slower growth in memory densities.
[Non-English content]
In the server market, investment is expected to continue to increase this year with the growth in 5G network and MEC. Replacement demand continues for servers from heavy investment in 2017 and 2018. Especially with the expansion of DDR5 supporting CPU adoption from the second half of the year, we anticipate demand for high-spec servers to increase. Demand recovery is expected to continue not only for cloud service, but also for corporate server customers, which has been improving since the second half of last year.
As a result, this year's demand for server DRAM is expected to grow by high 20% and enterprise SSD by low 30%.
[Non-English content]
The long-term outlook for high-performance memory products, including server products and HBM, is highly positive as demand for high-performing IT infrastructure is rising due to factors such as the acceleration of non-face-to-face infrastructure buildup, dissemination of AI machine learning and emergence of NFT and Metaverse as new applications.
[Non-English content]
Overall, DRAM demand growth is expected to be in the high teens percent this year, and we are aiming for the company's bit shipment growth to be in line with market growth. However, demand may soften in the first quarter due to seasonality, and we plan to respond more flexibly in light of our lower inventory level. Accordingly, the company's DRAM bit shipment in the first quarter is planned to decrease by mid- to high-single-digit QoQ.
[Non-English content]
Following last year, the company will maintain profitability-oriented DRAM business this year. We plan to adjust our inventory level with flexibility in response to changes in demand and strive to reduce industry volatility and maximize profitability.
[Non-English content]
NAND demand growth this year is expected to be about 30%, and the company is planning for bit shipment growth that outpaces demand growth again for the year. This is based on our existing NAND business and when Solidigm's volume is included, bit shipment growth is expected to almost double compared to last year.
[Non-English content]
However, given similar seasonality as DRAM in the first quarter, NAND bit shipment is expected to decrease by a high single-digit percent, with the possibility of falling further depending on the market environment such as supply chain issues. Meanwhile, when including Solidigm, approximately 20% bit shipment growth is planned QoQ.
[Non-English content]
Through Intel NAND business acquisition, we have set the stage to step up from being a late entrant to becoming a number two player. We will focus on growing the consolidated NAND business while maximizing the synergy between our existing NAND business and Solidigm in order to solidify our position as the industry leader that leads in technology and product development, as well as satisfies global customer needs.
[Non-English content]
Let me now turn to the company's technology and product roadmap, which will be the basis of sustained growth even in an unpredictable market environment. For DRAM, we plan to increase the portion of 1z nm and 1a nm.
1anm, where EUV is applied, will take up over a quarter of total production by the end of the year.
[Non-English content]
In the case of DDR5, for which demand will be in full swing this year, we began shipping for computing applications in the fourth quarter last year. We also recently shipped samples of 24 Gb DDR5 based on 1a nm, which is the largest density possible on a single DRAM chip.
This product improves the cost efficiency of data center operations and will be supplied first to cloud data centers, then are expected to be used in high-performance servers for big data processing or Metaverse deployment.
[Non-English content]
The 176-layer NAND, which started mass production in Q4 last year, is based on the same tech platform as the 128-layer. With minimum changes in the process, we plan to continue the success of the 128-layer, which showed industry-leading production ability and performance.
With mass production smoothly underway, the 120 layers share out of production is expected to be around 70% by the end of the year.
[Non-English content]
Let me make a correction. Now, regarding the percentage for the NAND by the end of the year, it is not the 128 layer, but the 176 layer. So it will be the NAND 176 layer that will reach around 70% share out of total production by the end of the year.
[Non-English content]
Our CapEx in 2021 was KRW 13.4 trillion.
This will likely increase this year, mostly due to large investments in construction and infrastructure for future growth, such as purchase of Yongin land site and the R&D center in the U.S.
[Non-English content]
On the other hand, equipment spending is planned to be similar to last year and will remain below the annual depreciation and amortization cost. As equipment lead time keeps getting longer, we are striving to minimize the impact of scheduled equipment delivery.
Given that disruption cannot be completely ruled out, we are looking into a number of options to address the issue. There will be additional CapEx requirement for Solidigm, but the impact will be limited considering its own cash generation capability.
[Non-English content]
Next, I will report on the company's ESG management activities and performance.
[Non-English content]
Through the CEO-led ESG Management Committee, the company last year focused on discussing mid- to long-term ESG strategies and improving on important and urgent issues.
[Non-English content]
To achieve our RE100 by 2050, we set an interim goal of using renewable electricity for 33% of the power in all our workplaces by 2030. To manage the specifics, we have decided to establish the Carbon Management Committee. The committee will oversee the entire process from our GHG reduction plan to implementation, and play a central role in addressing climate change and renewable electricity sourcing.
[Non-English content]
The company's MSCI ESG Rating was upgraded for the second consecutive year to grade A. This was achieved by a positive evaluation of our improvement in water stress management, such as reduction in water intensity and target setting for longer term water reuse through our task force activities. In addition, for 12 years in a row, we are included in the Dow Jones Sustainability Asia/Pacific Index, which selects companies based on both financial factors and non-financial factors, such as environmental and social impact.
[Non-English content]
Thr ough such active ESG management, we will try to differentiate ourselves in sustainability aspects and reach a balance between qualitative and quantitative growth.
[Non-English content]
Finally, let me introduce our new shareholder return policy designed to better share the fruits of our future growth.
[Non-English content]
For the past three years, from 2019 to 2021, we set a dividend policy of KRW 1,000 per share as fixed dividend and 5% of the annual FCF as additional dividend. FCF being the cash flow from operating activities, less acquisition of plant, property, and equipment to ensure predictability in shareholder return while reflecting earnings volatility caused by the industry cycle.
[Non-English content]
Accordingly, we paid KRW 1,000 per share dividend in 2019 when the FCF was in the negative, and since then, as FCF improved every year with performance pickup, dividend rose steadily to KRW 1,170 per share in 2020, and KRW 1,540 per share in 2021.
[Non-English content]
In light of the dividend performance and business changes since then, a new shareholder policy has been drafted to be applicable for the next three business years from 2022 to 2024.
Non-English content]
First, the policy of paying a fixed dividend per share with the additional dividend of 5% of the annual FCF will be maintained, but the fixed dividend will be raised by 20% from KRW 1,000 per share to KRW 1,200. KRW 1,200 per share is equivalent to the three-year average dividend per share, which is now used as the minimum dividend payment for the next policy period.
[Non-English content]
In addition, we will pay quarterly dividends starting this year. By paying a fixed dividend of KRW 1,200 per share spread evenly each quarter, we hope to provide a more stable shareholder value. We will amend the Articles of Association at the upcoming general shareholders meeting so that quarterly dividends can be paid from the first quarter.
[Foreign language].
Finally, we will use 50% of the cumulative FCF for the three-year period from 2022 to 2024 as a resource for shareholder return. Depending on the circumstances, we might also utilize this fund to conduct share buyback.
[Non-English content]
As we continue to generate and increase free cash flow, we will try to level up the return to our shareholders. SK hynix will never cease to challenge and develop so that we can create a new history in our company's value. Thank you.
[Non-English content]
With that, we are ready to take your questions.
[Non-English content]
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've pressed the number star one. For cancellation, please press star two, that is star and two on your phone. In order to allow as many Q&A chances as possible within the restricted time, we would appreciate only two questions per each participant.
[Non-English content]
The first question will be provided by Ricky Seo from HSBC. Please go ahead.
[Non-English content]
Well, first, thank you very much for announcing a much more positive and improved shareholder return policy. Now, you did explain a bit about Solidigm, but could you give us a bit more explanation or more detailed explanation about the expected synergy? We see that between the two businesses, you also have technological differences between the FG versus CTF. What would be the technology roadmap as well as the business plan for the business?
[Non-English content]
Now, with Solidigm becoming a member of the SK hynix family, what we can see first is the increase in the combined market share both from the NAND market share as well as the enterprise solutions. Now for the NAND market share based on the simple arithmetic, this will bring us to the number two position in the market. Now, of course, there have been some cases where the consolidation between two memory companies did not necessarily end up as one plus one being two. In our case, fortunately, there is a few redundancy in our product portfolio, and we were able to thus minimize any losses in the course of the consolidation. We are actually hoping that we would be able to end up with market share that is bigger than the simple sum.
Now of course, there could be some variability or volatility in the beginning, but we will try our best to defend the synergy as much as possible. I would say that the biggest advantage that SK hynix can gain out of this is in terms of the enterprise field, especially in the SSD area. Solidigm has very good technological understanding and also deep understanding in the computing architecture. Based on this understanding and know-how, we are now able to come up with our own product roadmap, which I would say is a big benefit to SK hynix, especially given the demerits that we had to experience as a late entrant so far.
[Non-English content]
It is true that, as you have observed, the two businesses had different technologies. In terms of the NAND, there is the charge trap-based 3D NAND, and then there was also the floating gate technology. The two technologies are different, but we also realized that the FG technology might be a bit difficult to be applied in areas other than the mobile. Let's say especially for SSD, in terms of the product reliability, I would say that the FG would be far superior than the CTS. Now other companies are launching the QLC lineup these days, but Intel has long been the leader in the QLC.
I would say that now Solidigm has the only technological base that would allow the launch of the PLC lineup down the road. I had also emphasized earlier that the PLC-based SSD lineup will be able to reach the kind of cost structure that would allow it to replace the HDD. We believe that, in addition to the cost synergy, we also will be able to have the synergy on our top line. Of course, on one hand, it is possible that maintaining the two technologies in parallel might not be entirely beneficial in terms of the cost implications. But given the potential in the top-line synergy effect, we are currently planning to maintain the two technologies in parallel. Let me add further to that.
Now for the floating gate technology, now it is specialized in for SSD, especially for the data centers. Also given that, so as you can see from its like 4-bit per cell or 5-bit per cell, now this is also the type of technology that can have better reliability than CTF in cases where we have to have more capacity per cell. That is why, SK hynix is planning to maintain this technology as well.
[Non-English content]
[Non-Engslish content]
The following question will be presented by Peter Lee from Citigroup. Please go ahead with your question.
[Non-English content]
I have question about the Intel NAND business as well. It was reported earlier that you have completed the first phase of the acquisition, and thus you have acquired Intel's SSD business and Dalian fab. What are the changes on the revenue as well as the assets for SK hynix? Also when you combine the earnings, what are the changes?
[Non-English content]
Now, as of 2021, the SSD company, Solidigm and the Dalian fab company have been incorporated as subsidiaries, and they are immediately reflected on our consolidated balance sheet. For the revenue to be gained or the sales to be gained from 2022, they will be booked into our consolidated profit and loss statement.
In terms of the changes to the balance sheet, then there is an addition of KRW 8.6 trillion in inventory assets as well as PPE and KRW 1.7 trillion in intangible assets. The cash and cash equivalent is lower by KRW 7.5 trillion. Then about KRW 2 billion for the second phase of the acquisition is set aside as debt.
[Non-English content]
[Non-English content]
The following question will be presented by Sunwoo Kim from Meritz Securities. Please go ahead.
[Non-English content]
Now, I do understand that the investment for the year is going to be slightly up from last year. We see that there is a trend of the unit cost of investment going up in the memory industry overall. There are also some concerns among investors about this absolute increase in investment. My question is about the cost curve. For both DRAM and NAND, what has been the bit cost for the year and also the year before then? Have there been any changes to the bit cost? This year, do you expect the bit cost to fall? If yes, then what do you believe will be the reasons?
[Non-English content]
The question is pertaining to the CapEx intensity and also the concerns related to this coming from the market, which is about the increase in the unit CapEx per bit. Now of course, that has been an area of interest and concern for the company as well. Now for both DRAM and NAND, what we have been seeing at least up until the early 2010, is that thanks to tech migration, there was almost automatic cost reduction that was made possible almost every year. Simply from tech migration we were able to reduce the cost. Now in the past few years that has been made much more difficult. For DRAM, we see that, tech migration does not automatically lead to cost reduction. At least there are now limitations to that.
Also with the NAND technology, the 3D stacking is becoming more and more difficult. That means that just through tech migration, it is no longer so easy to reduce the cost or for example like fab cost or COO. As a result, what we have been doing is to increase CapEx in order to have capacity to ramp up in order to meet the demand coming from the market, which has resulted in increasing CapEx so far.
[Non-English content]
Between DRAM and NAND, we see that at least for NAND, the cost reduction situation is a little bit better in terms of the reduction in the COO because for the 3D stacking, at least it is at the moment a little bit easier than the tech migration in the DRAM. That is why I say that for the NAND the situation is a bit better. Now interestingly enough, when we look at the COO reduction rate for the past few years for both SK hynix and also other players in both DRAM and NAND, when we look at the numbers then compared to the COO reduction in 2018, 2019, we see that there actually has been a deeper cost reduction in 2019, 2021, and possibly even in the 2022 period based on the estimation.
That is because unlike the 2018, which was a boom year, now at that time, rather than looking for efficiency, we were investing heavily and also producing heavily to simply meet the rapidly rising demand. Now the situation is that the market growth is a lot more stable, so it seems as if SK hynix and other players are focusing more on getting more efficiency out of the existing assets. Also given the fact that the tech migration, it now has limitations in enabling cost reduction, we have been focusing more on reducing inefficiency in fab or the existing business structure. That is how we were able to maintain the COO reduction rate that is similar to what we have seen in early 2010s.
[Non-English content]
[Non-English content]
The following question will be presented by Jun Jeong Park from JP Morgan. Please go ahead.
[Non-English content]
I have two questions. Between the FG and the CT technology, the technology for the NAND is different between the two companies. Let's say, will it be possible to use Intel's controller IC on the NAND that has been produced by SK hynix's CT? That is the first question. The second question is about the EUV. It was reported earlier that the EUV-based 1a nm will take up about 25% by the end of the year. I understand that for hynix, the EUV is applied on only one layer of the 1a nm. Does this still bring about a significant cost reduction or does it increase production? Basically the question is what would be the advantages of the 1α nm when applied with EUV?
[Non-English content]
Now for the NAND business and also the SSD business that utilizes the NAND, now I would say that there are largely three chunks in this business. There would be first the raw NAND, and then the SOC, which would be the controller, and the firmware. Now, Solidigm. So, obviously the Solidigm's SoC or the firmware can be used with the SK hynix-produced NAND. I would say that that is actually at the heart of our synergy, our expected synergy effect. Now, Solidigm has always had business competitiveness, but then Intel not being a memory-focused company, it also had limited capabilities to expand its raw NAND capacity, and that is why its business expansion had been limited.
Now, by utilizing FT on the SK hynix's CTF-based wafer, and also combined with the superior technology of SoC and firmware by Solidigm, we do believe that we will be able to further expand into the market.
[Non-English content]
Now having said that, utilizing the SG-based SoC or firmware on the CTF-based wafer could take some time because of the necessary technological revision. Fortunately, the two companies have worked very closely together and now we are able to utilize Solidigm's SoC and firmware on the company's 128-layer NAND wafer. The prototype has been produced based on this, and we hope to launch this into the market soon. To your second question. Now, for the EUV. Now, basically our position is that there has to be a cost break-even between the EUV and the DUV, which is the ArF. For the...
Now when we look, when we compare the technology between the EUV and the DUV, then for the same patterning for the DUV, it requires multiple shots, whereas for the EUV it requires only one shot. Of course there are obvious technological advantages to the EUV, and it could also improve the yield. There are also some disadvantages with the EUV, meaning that, particularly the technology not being mature enough in the beginning, there could be some additional cost in the beginning. Again, when it comes to the patterning for the company, the position is that, as we adopt different technology, there has to be a cost break-even between the EUV and DUV.
[Non-Engslish content]
We will take the next question.
[Non-English content]
The following question will be presented by Simon Woo from Bank of America Securities. Please go ahead.
[Non-English content]
Now I also have two questions. Actually first, thank you very much for the very detailed IR information and materials. Now my first two questions. First is about the financials and second is about the fab operation. First question about financials is, it was mentioned in the presentation that the company plans to utilize 50% of the free cash flow for the next three years for shareholder return. For the remaining resource, it could be used for share buyback or other forms of shareholder return. It means that the calculation or the settlement period for this would be in 2024. May I understand that this is not going to happen next year or the year after that? A follow-up question is for the Dalian fab.
As a result of acquiring the Dalian fab, the company now has an increase of KRW 1.7 trillion in intangible asset. What is going to be the number of years for amortizing this? Or does the company plan to amortize this as cost in the second quarter and start out with a clean slate? The second question is about the fab operation. The company still has the Yongin land, and there is also demand for more plant space. I wonder when the groundbreaking is going to take place for the Yongin site.
For the M16, now, to keep using the remaining space in the M16 for the next four to five years, I'm not sure whether that is going to suffice because yes, it is a big plant, but when we look at the clean room space, it simply does not appear to be enough to meet the demand for the next four to five years. What is the company's plan for the fab operation?
[Non-English content]
First, about the dividend, as it was explained earlier, we intend to use about 50% level of the cumulative three-year free cash flow to return to shareholders. As I mentioned earlier, the policy is to have KRW 1,200 per share as fixed dividend to be added with 5% of the free cash flow of the year. This is to be paid out evenly on a quarterly basis. Now for SK hynix and other memory players, usually the first quarter and the second quarter are when the CapEx is concentrated, so it's unlikely to have free cash flow in the first and the second quarters. For the first and the second quarters, the dividend payout is likely to be the fixed dividend divided by four, in other words, KRW 300 per share.
In the third quarter and the fourth quarter, it's likely that free cash flow is going to be generated. In such a case then the 5% of the free cash flow for the year is likely to be calculated and paid out in the fourth quarter of the year. Now for the cumulative three-year free cash flow and also the 50% of that.
Now that means, yes, in order to do the calculation then it has to be after 2022, 2023 and 2024. Meaning that the calculation or the settlement will be done in the early 2025. It could happen that way or in another scenario, let's say like in 2018, if there is a huge free cash flow generated in a particular year, then we might decide to have a special dividend payout just so that we can avoid having to pay out a huge sum at the end of the period. In other words, just to even out the financial burden.
[Non-English content]
For the second question on the intangible asset of KRW 1.7 trillion acquired through Solidigm, we are currently not considering amortization.
[Non-Engslish content]
Now, if I may add a bit more explanation. In some other M&A cases, what we have seen is that the transaction value was significantly higher than the book value, so the intangible asset portion was quite large. But in the case of Solidigm, that does not apply. Because of that, for the intangible assets, unless there is damage to the business, currently we do not have any plan for impairment. About the question about the fab space, the question was specifically about the Yongin time frame and then also the available space for DRAM in M16 fab. Actually it's a big question for the company as well.
Yes, it's true that there is a need for us to secure the site in Yongin and start with the fab construction. We believe that we need to have a new fab going to operation by early 2026. It doesn't entirely depend on us. It depends more on the SPC, which is a completely separate entity from SK hynix, which is currently buying the site and also preparing the ground to create an industrial complex there. Once the SPC is done with that, we need to buy the site from them, and then we can go into the groundbreaking for the fab. There are some uncertainties involved here.
It is true that because of the faster than expected ramp up in the M16, there is a need for additional space. Should there be some, let's say, unexpected delay or issues with the planned Yongin fab or the Yongin site, then we would have to look for other space. The company is currently mulling over the different options regarding that as well.
[Non-Engslish content]
[Non-English content]
The following question will be presented by Hyunwoo Doh from NH Investment & Securities. Please go ahead with your question.
[Non-English content]
Now I have questions about the inventory first. Now we see that one thing that was a bit interesting in the market last year was the gap in the inventory between the suppliers and the customers. Since the third quarter until today, does the company see this gap having been reduced or having been increased? Another question is about the demand coming from the market. We see recently that the demand is higher than expected, and that is probably because the customers concerned about the COVID-19 or the Omicron variant have been trying to build up their inventory. Perhaps they have been, let's say, pulled up the demand that might be necessary after the second quarter into the first quarter.
Does this also mean that perhaps, after the second quarter, demand might start to drop very quickly?
[Non-English content]
Now, I would take your question to be pertinent about the DRAM. That is also going to be my response. Now between the third quarter and today, there appears to be a gap in the inventory position between the suppliers and the customers. That was the question. Now for the suppliers. For the company, our inventory level remains largely unchanged. It is between 1.5-2 weeks, and it did dip a little bit at the end of the year. As was explained earlier, for the fourth quarter, we would be trying to respond mostly to the real demand. There's also the need for us to rebuild the inventory that went down a little bit in the fourth quarter of last year.
We are also trying to rebuild the inventory in preparation for the second quarter and the second half. By the end of the first quarter, it's likely that the inventory level will be a bit higher.
[Non-English content]
In relation to demand, what we have to focus on most is the computing side. Looking at the PC industry, so we see that at the end of last year, the PC industry's memory inventory went down a little bit. Also for the server side, there have been earnings releases and also outlook announcements by major players recently. All of them are commonly looking ahead to increased investment in 2022. In terms of the inventory buildup by the customers at the end of the year, we see that as a natural inventory buildup. We see not much, so we don't really see anything out of the ordinary on the mobile side either. For the computing side, we believe that the balance between PC and server is continuing.
[Non-Engslish content]
To the second question, briefly. Now, for the supply chain disruption, it works on both ends, meaning on both demand and supply, so it also affects supply. In the end, it actually balances out. In the second half, what we have to focus on is, of course, the inventory on both the demand and supply side, but also the inventory for the new products and the old products. Now, for companies like us who have to achieve big growth, based on 1anm, the inventory consumption for old tech products will happen very quickly going from the first half to the second half of the year.
Now, for the new tech, in this case, it is. I mean, there is not going to be much inventory on both the demand and supply side because they are new products. We would also have to look into the inventory position on different aspects as we try to respond to the changes.
[Non-English content]
[NonEnglish content]
The following question will be presented by Nicolas Gaudois from UBS. Please go ahead with your question.
Yes, good morning, and thanks for taking my questions. Both will be related to DRAM. The first one is regarding DDR5. What is your forecast for DDR5 demand as percentage of total DRAM demand in H2 2022 and the first half of 2023? Do you expect supply, mostly 1α nm to be able to keep up with demand? Or could we actually see the premium of DDR5 over DDR4 remaining high for longer as supply does not match demand for a while? As a follow-up question and more longer term, you talked about scaling in DRAM becoming more and more challenging. What is your view on when you could actually move to 3D DRAM or more specifically 4F² vertical channel access transistor for mass production?
Would you still, after this, need more advanced lithography like High-NA EUV to continue scaling? Thank you.
[Non-Engslish content]
[Non-English content]
First about the DDR5, the question was about the percentage or the share of the demand for DDR5 out of the total DRAM demand. As of the end of 2022, the forecast is somewhere between high teens percent to low 20%, and this is out of the server side. By the end of 2023, we expect there to be a crossover. Per your question, which was until the first half of 2023, we believe that by the end of the first half of 2023, the expectation is that it will be along that trend line until the crossover at the end of 2023. Currently the demand for DDR5 is rising, especially from the high-end or high-value PC side.
For the server, now DDR5 offers the advantages of better performance with low energy consumption and also better reliability as this is on the SOC. We believe that there is going to be a growing demand for 1α nm-based 500/600 Mbps into 2023.
[Non-English content]
In terms of the premium for the DDR5, I cannot give you the specific numbers, but it's true that there would be some cost addition or premium added compared to the DDR4. Because even if it is the same lithography, then still the net die would be reduced. Also there would be some other process technologies to be added, and also new BOM costs would be increased. Also we need to have additional ECC die. Also for the high-density modules, we would also have to add a TSV process. Overall, there is going to be increased cost for the DDR5. As I said earlier, it is rising among the high end or the high value.
On the client side, we see that there is a demand coming from the high-value segment, and likewise for the server. Especially for the customers for the high-density products, we are working on the right value proposition, so they will be able to reach a balance between demand and supply.
[Non-English content]
The next question was about the 3D DRAM, which is actually a question that we have been getting recently from the market. Based on the experience of the NAND, which has moved on from 2D NAND to 3D NAND, so I believe that the expectation in the market is that it is going to be natural for DRAM to move from 2D to 3D. Compared to NAND, which is actually technologically relatively simple, for DRAM to move to 3D is by no means easy. For DRAM, as I said earlier, because there are going to be a number of technological limitations to come our way in the next few years, that means that the memory suppliers would have to try to overcome the technological limitations with increased cost.
[Non-English content]
For 3D DRAM, it's not like it's a set course or set path for us and the question is only about when. It's more about addressing the various technological limitations that we are going to experience over DRAM. Among the various options of addressing the technological limitations, moving to 3D DRAM could be one of them. I would say I could put the question into this frame. Of course, there could be many different technologies to be used in order to address the scaling limitation of DRAM. Part of the technological solution could be to have a structural change to, on the technological platform, in other words, moving on to 3D DRAM.
I would say that that is the question given to Hynix as well as other DRAM companies, and I'm sure that the other DRAM players are also mulling over this question. In terms of when this is going to happen, I must say that there is not enough visibility for us to give you a specific timeline. The likely scenario is that it could happen sometime in the late 2020s. I'm sure that, including Hynix, many different DRAM companies are currently looking over the possibility of various component technologies. That concludes the SK hynix 2021 fourth quarter earnings release conference call. Thank you very much.