SK Hynix Inc

KRX:000660

Earnings Call Analysis

SK Hynix Achieves Record Revenue with Strong AI Memory Demand

In its latest earnings call, SK Hynix reported record quarterly revenue of KRW 17.57 trillion, up 94% year-over-year, driven by robust demand for AI memory products. Notable growth was seen in HBM3E, which increased by over 70% sequentially, now constituting about 40% of DRAM sales. Operating profit hit a high of KRW 7.03 trillion, reflecting a 40% profit margin. For Q4, DRAM bit shipments are expected to grow by mid-single digits, with NAND bit shipments increasing by low-teen percentages. The company maintains a strong focus on high-margin products despite inventory corrections for conventional offerings.

Strong Quarterly Performance Amid Shifting Demand

SK Hynix reported a record quarterly revenue of KRW 17.57 trillion in Q3 2024, marking a 7% increase from the previous quarter and an impressive 94% year-on-year growth. This surge was largely driven by robust demand for AI memory products, particularly in data centers, even as traditional markets like PCs and smartphones showed weaker demand.

DRAM Segment: Mixed Signals with High Margins

In the DRAM sector, HBM (High Bandwidth Memory) products saw sales volume increase significantly, with revenue rising by more than 70% sequentially and 330% year-on-year. Although conventional DRAM shipments decreased due to declining mobile and PC demand, the average selling price (ASP) for DRAM rose in the mid-teen percent range, mainly supported by strong pricing for HBM products. The blended ASP improvements indicate effective pricing strategies despite lower volumes.

NAND Sector: Strategic Focus on High-Margin Products

The NAND segment faced a mid-teen percent sequential decline in bit growth due to reduced demand from PC and mobile customers. However, enterprise SSDs bucked the trend with revenues up almost 20% quarter-on-quarter and more than 430% year-on-year. The company’s focus on high-margin enterprise SSDs contributed to a higher ASP, indicating an effective shift in product strategy amidst market corrections.

Operational Efficiency and Record Profits

Operating profit rose by KRW 1.56 trillion sequentially, reaching a record KRW 7.03 trillion. The operating margin improved to 40%, benefiting from pricing increases rather than shipment growth, showcasing the company’s ability to enhance profitability through cost control and premium product sales.

Future Outlook: Demand for AI Drives Growth

SK Hynix maintains an optimistic outlook, projecting ongoing growth in the memory market, particularly from AI applications. The company anticipates that as AI technology advances, demand for AI servers and associated memory products like HBM and enterprise SSDs will continue to grow, further stabilizing the market.

Financial Strength and Capital Management

As of Q3, the company holds KRW 10.9 trillion in cash and cash equivalents, a KRW 1.2 trillion increase from the previous quarter, along with a reduction in interest-bearing debt by KRW 3.4 trillion. This solid cash position and improved debt ratios (debt-to-equity of 33% and net debt-to-equity of 17%) enhance financial flexibility for future investments and strategies.

Guidance and Projections for Upcoming Quarters

Looking forward, SK Hynix expects DRAM bit shipments to grow by a mid-single-digit percentage sequentially in Q4, driven by demand from servers. Additionally, the HBM sales share is projected to reach about 40% by the end of Q4, reflecting a successful pivot towards high-performance memory solutions. The company also indicated plans to begin shipments of HBM3E 12Hi products, further positioning itself strongly in the premium memory market.

Challenges Ahead: Navigating Market Volatility

Despite the positive outlook, the company is also cautious about prevailing market conditions, including competition from new entrants, particularly in legacy DRAM products. The entry of Chinese suppliers into the market increases pressure on pricing and could impact future growth trajectories.

Innovative Momentum and Product Development

SK Hynix continues to invest in R&D and new technologies, planning substantial investments in HBM and other advanced memory solutions. The company is poised to maintain its market leadership by transitioning to next-generation products and optimizing its product mix to meet evolving customer needs.

Earnings Call Transcript 2024-Q3

[Operator]

[Operator Instructions] Please note that the presentation will be translated simultaneously and the Q&A session will be translated consecutively. With that, we will now begin with SK Hynix's presentation.

[Seong Hwan Park, executive]

[Interpreted] Good morning, afternoon and evening. This is Park Seong Hwan, Head of IR at SK Hynix. Welcome to the SK Hynix 2024 Third Quarter Earnings Release Conference Call. Allow me to introduce the executives present here today. We are joined by CFO, Kim Woo-Hyun, Head of DRAM Marketing [ Kim Kyu Hyun ] and Head of NAND Marketing [ Kim Seok ].

Let me issue a disclaimer that all outlooks presented by the company are subject to change, depending on the macroeconomic and market circumstances. With that, we will now begin the SK Hynix earnings release conference call for the third quarter of 2024.

Mr. Kim, CFO, will first present the earnings followed by the company's future plans and market outlook and a Q&A session with the attending executives.

[Woo-Hyun Kim, executive]

[Interpreted] Good morning, everyone. Allow me to first introduce the company's performance for the third quarter of 2024. In the third quarter, while the recovery of demand in conventional applications, such as PCs and smartphones was somewhat delayed, the demand for AI memory products for data centers remained strong.

We expanded sales of value-added products, such as HBM and enterprise SSDs, achieving record-high quarterly revenue of KRW 17.57 trillion, an increase of 7% compared to the previous quarter and 94% compared to the same period last year.

First, for DRAM, while sales volume of HBM3E products increased significantly, sales volume of conventional DRAM products were reduced due to weakness in mobile and PC demand, resulting in a slight decrease in shipment volume compared to previous quarter. However, our sales volume of HBM increased and as the upward trend in conventional DRAM prices continued, blended ASP for the quarter rose by mid-teen percent sequentially which is a similar level of improvement that we saw in the previous quarter.

Notably, HBM revenue increased by more than 70% sequentially and by more than 330% year-on-year, driving the company's revenue growth.

For NAND, despite solid demand for enterprise SSDs, sales bit growth was down by mid-teen percent sequentially due to weak procurement demand from PC and mobile customers. We are now undergoing inventory corrections. Meanwhile, blended ASP saw another steep increase in the quarter, which rose by mid-teen percent sequentially.

Price increased across all products, except for discrete NAND and MCP products, while sales portion of higher-priced enterprise SSDs expanded. Enterprise SSD revenue in the third quarter increased by almost 20% quarter-on-quarter and by more than 430% year-on-year, driving net revenue growth this year.

In the third quarter, operating profit increased by KRW 1.56 trillion sequentially, reaching KRW 7.03 trillion, which marks a record high quarterly operating profit. The record high profit was achieved as the sales increase was driven by price increase rather than volume increase and also due to the higher sales contribution from high-margin products. Operating profit margin recorded 40%, up 7 percent points from the previous quarter.

Capitalizing on our industry-leading product technology, we were able to expand sales of value-added products, such as HBM and enterprise SSDs meaningfully in the quarter. Therefore, DRAM and NAND profitability both improved sequentially, resulting in a profit that surpassed prior peak operating profit achieved during the super cycle of 2018.

The depreciation and amortization expenses for the third quarter was KRW 3.07 trillion, continuing a downward trend due to last year's CapEx reductions, while EBITDA reached KRW 10.1 trillion with an EBITDA margin of 57%.

Non-operating expenses net of gain for the third quarter was KRW 0.15 trillion, which includes net interest expense of KRW 0.23 trillion, and net foreign currency related gain of KRW 18 billion. As a result, net profit before tax was KRW 6.88 trillion, and net profit for the quarter was KRW 5.75 trillion with net profit margin of 33%.

Total consolidated cash and cash equivalents, including short-term investments, amounted to KRW 10.9 trillion at the end of the third quarter, a KRW 1.2 trillion increase from a quarter ago. Interest-bearing debt was KRW 21.8 trillion, down KRW 3.4 trillion from the end of previous quarter, and net debt was KRW 11 trillion, down KRW 4.6 trillion sequentially.

Consequently, our debt-to-equity and net debt-to-equity ratios at the end of the third quarter was 33% and 17%, respectively, improving meaningfully compared to the that of previous quarter.

Growth momentum of the memory market is continuing this year, driven by ongoing demand for AI memory products. Memory products adopted in AI servers such as HBM and enterprise SSDs are showing particularly high demand growth. Although the recovery of demand from major applications such as PCs and smartphones, is somewhat delayed compared to earlier expectations. Strong demand for servers and AI memory products is offsetting the weakness in demand from traditional end applications.

As the application of AI technology widens, the application of memory is also expected to broaden, leading to even more stable market growth for memory going forward.

The impact of AI PCs and AI smartphones released this year on memory demand has been rather limited. But from next year, enhanced AI features are expected to provide a growing momentum for memory demand.

First, the PC shipments this year are expected to remain similar to last year's levels but are expected to grow by low to mid-single-digit range next year. The replacement demand for corporate PCs due to the end of Windows 10 support in the second half of next year and full-fledged AI PCs are expected to lead to a gradual recovery in memory demand for PCs. The share of LPDDR5 within PC DRAM has already reached a meaningful level and as sales of AI PCs rise, demand for high-performance and low-power products such as LPCAMM2 is expected to increase.

The smartphone market is also expected to see an improvement next year, growing by around low to mid-single digits. The smartphone market having reached maturity faced stagnation due to lack of innovative technologies and expansion of the refurbished market. However, potential launch of innovative AI features is expected to stimulate consumer purchasing demand.

This year, AI features were limited to a few flagship models, but are expected to be adopted more widely in high-end product lines from next year, significantly increasing the sales proportion of AI smartphones. Even that running small LLM AI models in smartphones requires at least an additional 3 to 4 gigabyte DRAM, the growth of AI smartphones is expected to not only increase DRAM content, but also the need for high-performance memory such as LPDDR5X and LPDDR5T.

The server market is expected to see significant demand growth especially for AI servers, as big tech companies are continuously investing in order to capture the AI market. The scale of investments from major big tech companies this year has increased compared to initial forecast. And despite some concerns over delays in monetization, the likelihood of reducing spending in AI servers appears low.

Meanwhile, due to upcoming replacement cycles for general servers with growing customer needs for energy and space efficiency, the overall server market is expected to grow by mid- to high single digits next year.

As generative AI develops into multimodal forms and workloads continue to increase, the demand momentum for AI servers is expected to proceed in the mid- to long term. For example, as with the recent release of OpenAIs' o1, longer inference times lead to increased computing power needs, thereby enhancing performance. This is because more complex and in-depth models necessitate greater computational resources for more accurate and refined results.

Various companies are currently competitively undergoing research to develop AGI and expanding training infrastructure in the long term to support these developments.

Since large model sizes are essential for AGI, hardware requirements are considerable and are expected to become a key driver of server demand. Additionally, for continuous updates of AI models and the provision of personalized services, real-time data processing and fast inference speed will be required, which is expected to increase the demand for cloud infrastructure and edge computing servers simultaneously.

To meet the increasing demand and performance improvements of AI servers, growth in performance and density of HBM that supports computation operations is essential. Besides expanding bandwidth, the development of higher stacked packaged products for increased density and low power characteristics is also crucial.

Furthermore, as customer demands diversify, the need for custom HBM is also expected to gradually grow. Considering such demand environment, DRAM demand growth is expected to improve from mid- to high teen percent this year to high-teen percent next year, while NAND demand is projected to grow by mid-teen percent both this year and next.

Now I will discuss our plan.

Fourth quarter DRAM Bit shipments are expected to grow by mid-single-digit percent sequentially, driven by further growth in sales of HBM and server DRAM that are showing strong demand. NAND Bit shipments are expected to increase by low-teen percent sequentially, primarily due to higher sales of enterprise SSD products.

We will now -- we will continue our focus on improving profitability by expanding sales of value-added products in the fourth quarter as well.

The share of HBM sales which is driving our DRAM revenue growth, expanded to 30% of DRAM sales in the third quarter and is expected to reach around 40% in the fourth quarter.

In the third quarter, the shipment volume of HBM3E was greater than HBM3, and we plan to begin shipments of HBM3E 12Hi products in Q4 as scheduled.

In the first half of next year, sales volume of 12Hi HBM3E products is expected to surpass that of 8Hi HBM3E. We plan to respond swiftly to the rapidly increasing demand for HBM3E based on our proven high performance and stable 1b nanometer process and advanced MR-MUF packaging technology.

Through our differentiated AI product competitiveness, we have achieved revenue stability as well as expansion of our profit drivers. As a result, we have become able to manage our product and sales strategies flexibly according to market demand. Going forward, we will continue to focus on improving our profitability.

At the end of August, we developed the industry's first 16 gigabit DDR5 product using 1c nanometer node, the sixth generation of the 10-nanometer process, continuing our DRAM technology leadership. Our 1c nanometer node is developed based on 1b nanometer process platform that has already proven mass production stability and high product quality, thereby minimizing potential errors during the mass production stage and especially maintaining the advantages of the 1b nanometer process.

We have provided samples for product validation and plan to respond to the growing demand for DDR5 next year with products that have the best-in-class performance and cost competitiveness. The sales of enterprise SSD, which is driving our NAND segment recovery continued to increase, accounting for over 60% of our NAND sales in the third quarter. We are currently the only company in the industry, providing 60-terabyte products in mass and we are in the process of validating our 122-terabyte product to supply in the first half of next year.

Furthermore, we developed the PCle Gen 5 eSSD product, PEB110 in September, applying to 238-layer technology, which will further strengthen our product lineup and solidify our position in the enterprise SSD market.

The changes in demand for memory products due to advancements in AI technology are also bringing significant changes in the supply environment. For the past 2 years, production growth of conventional DRAM products was limited due to production cuts during the downturn and expansion of HBM portion in total production. However, this year, supply growth of legacy products is more marked amid growing supply due to recovering utilization rates and increasing capital spending.

Respond to these market changes, we plan to expand production focused on products with clear competitive advantages. While reducing production of low-demand DDR4 and LPDDR4 earlier than initially planned, we plan to accelerate the transition to advanced processes needed to increase production of HBM, DDR5 and LPDDR5.

Meanwhile, for NAND, we will continue to operate with the focus on technology transitions rather than capacity increases until industry inventories are sufficiently depleted while placing greater emphasis on investment efficiency and optimizing our product mix.

This year, our investments have been concentrated on expanding the mass production of value-added products that are increasing in demand. Next year, we plan to continue investing in tech migration to advanced nodes to ensure stable supply for products where demand growth is evident, such as HBM, DDR5, LPDDR5 and enterprise SSDs.

We will also consider multiple scenarios to adjust investments proactively in the event of a market downturn. Furthermore, infrastructure spending is expected to increase next year due to ongoing construction of M15X and preparation of first fab in Yongin Cluster, which are to be our future growth platform.

M15X which is set to be completed first is expected to contribute to our DRAM production starting in 2026, and we plan to flexibly adjust the timing and scale of mass production in the new fab according to market demand. We achieved record high revenue and operating profit this quarter. On the back of our differentiated AI memory technology leadership, we will continue to support next-generation memory products on time to meet customers' demand so that we can continue to contribute to the growth of the AI industry. Thank you.

With that, we are now ready to take your questions.

[Operator]

[Operator Instructions] The first question will be provided by Dongwon Kim from KB Securities.

[Dongwon Kim, analyst]

[Interpreted] I have a question related to the DRAM. Recently, with the continued slowdown in demand for mobile and PC and also because of the intensifying competition, there has been a fall in the price of the general purpose DRAM such as D4 and LPD4. So can you give us your price outlook for the general purpose DRAM going forward? And also, if legacy DRAM prices go down, then the blended ASPs will fall as well. However, because of the increasing share of HBM products in your sales structure, do you believe your blended ASP will move in a different direction from the actual market prices?

[Unknown Executive]

[Interpreted] Well, thank you very much for your question. As you have mentioned, the DRAM price volatility is increasing due to a number of factors negatively impacting the DRAM supply and demand, including the delayed improvement in PC and mobile demand and also the Chinese suppliers entering the legacy market.

However, there are significant differences in terms of supply and demand of legacy products, such as DDR4 and LPDDR4 with that of premium products such as HBM, DDR5 and LPDDR5, which can -- we believe can lead to different price directions for each product.

As you all know, the volume and price of our HBM products are largely locked in through long-term contracts. And so our company's average HBM price in 2025 is expected to increase Y-o-Y accompanied by a shift towards greater share of the HBM3E products.

In addition, for the high-performance, high-spec DDR5 and LPDDR5, at this point, it will still take some time for new entrants to enter the market. and customers and suppliers' inventory levels are still believed to be low, while future demand for these products is expected to increase, therefore, limiting any downward pressure on prices.

Meanwhile, our DRAM business has become much more stabilized with HBM share in the revenue growing, expected to reach 40% by the end of the year. Given HBM's relatively high ASP, we expect our blended DRAM ASP to continue its improvement trend even as prices for some products decline.

Going forward, by strengthening our HBM business, we will strive to maximize our profitability while also securing stable revenue.

[Operator]

The following question will be presented by Simon Woo from Bank of America.

[Simon Woo, analyst]

[Interpreted] I have a question about the third quarter Bit growth, especially for the NAND products. It seems that your Bit growth has been lower than was expected. So can you provide a background as to why this has occurred. But despite this being so, I would like to congratulate you on your good performance.

[Unknown Executive]

[Interpreted] First, let me talk about the DRAM. In the case of DRAM, 3 months ago in our earnings release, our growth guidance in our earnings call was for a low single-digit growth versus Q2, but actually, we delivered a low single-digit decline in the third quarter.

The lower-than-planned sales were mostly PC applications. And as you might expect, the DDR4 sales volumes have declined. However, we sold more DDR5 than planned making up for some of the lost volume.

For NAND, we said our Q3 EBIT growth guidance was for a mid-single digit decline Q-on-Q. And the actual result was a mid-10% decline. Again, most of the decline was in discrete and client SSD products with pressure on discrete pricing as demand for PCs -- from PCs and smartphones slowed.

In line with our profitability-driven sales strategy, we have decided to carry inventory for some of the products where demand is slowing down and prices are falling, even if it is well below our original plan. And we plan to gradually reduce inventory through production adjustments in the future.

[Operator]

The following question will be presented by Sung Kyu Kim from Daiwa Securities.

[S. K. Kim, analyst]

[Interpreted] First of all, let me offer my congratulations for your excellent performance. I have a question about the HBM. Recently with the potential of a possibility of delays in the delivery of the Blackwell platform and the increase in HBM supply from the suppliers, there are concerns about a possible supply -- oversupply situation in 2025 HBM market. So what is the company's view on this?

[Unknown Executive]

[Interpreted] Thank you very much for your questions. So unlike the general DRAM, HBM has a long-term contract structure with the volume and pricing negotiations already completed with most of our customers for 2025. So we have very high visibility on the demand side.

We believe that next year's HBM demand is likely to be higher than what is currently anticipated, given the strong demand for AI chips and also the customers' willingness to expand their AI investments.

The current AI technologies are evolving in a direction that uses more computing time to generate multiple possibilities and consider potential outcomes, and also make inference-based decisions rather than simply generating learn patterns or making predictions that are based on historical data.

So as we expect the computing power requirements to increase further and computational resources to be in greater demand going forward, we believe that it is premature to talk about a slowdown in demand for AI semiconductors or HBMs at this point.

Rather, the increasing difficulty of the technology required to develop new HBM products, the resulting yield losses and also considering the need for customer qualifications. We believe all of these factors will make it difficult for the memory industry to supply enough products of the quality required by the customers in a timely manner.

In other words, the potential for upside on demand and a downside for the supply is high, and we expect demand to remain stronger than supply in the coming year. We believe this is reflected in the fact that unlike the general DRAM, customers are signing long-term contracts for HBM.

[Operator]

The following question will be presented by Giuni Lee from Goldman Sachs.

[Giuni Lee, analyst]

[Interpreted] Hello. My name is Giuni Lee from Goldman Sachs. Once again, I would like to also offer my congratulations for your good performance, and thank you very much for giving me this opportunity to ask my question.

I also have a question on HBM. During your presentation, you noted that your HBM capacity has been largely sold out for 2025. But if there is an additional demand or upside demand for HBM, do you have the additional capacity to respond? And how do you intend to respond? That is my question.

[Unknown Executive]

[Interpreted] Thank you very much for your question. Contrary to some concerns, AI-driven demand continues to exceed expectations this year, resulting in additional supply requests from our customers.

We are on track with our plan to more than double the TSV capacity this year compared to last year and our transition to 1b nanometer and to ramp up, the HBM3E supply is also proceeding very smoothly.

We are also making the necessary investments to meet our customers' volume requirements for next year, which have already been secured. But it is also true that our production capacity is facing limitations in meeting all of the increased demand in excess of our original plan.

Especially recently, demand for HBM3E from customers has been growing at an accelerating pace more so compared to our initial expectations. We plan to transition legacy technologies utilized in HBM3 and DDR4 to advance the production processes as soon as it's possible, focusing on reducing production of products with slowing demand and increasing production of HBM3E to meet customer demand as much as possible.

[Operator]

The following question will be presented by Peter Li from Citigroup.

[Peter Li, analyst]

[Interpreted] I have a question related to inventory. There are inventory issues due to the recent slowdown in mobile and PC demand and there are market concerns about this matter. So what is the current inventory situation for the company and your customers? And do you believe the situation will be different from conventional products compared to the premium products, but when do you think the inventory will become normalized?

[Unknown Executive]

[Interpreted] Thank you very much for your question. As the recovery in PC and mobile demand has been slower than initially anticipated, we have adjusted our production plans and reducing production of some of the legacy products with high share of inventory faster than has been originally planned.

The overall inventories at PC and mobile customers at the end of third quarter is understood to be remaining at similar levels compared to the previous quarter, while inventories at server customers where demand is relatively robust or at healthy levels.

Meanwhile, a high level of demand interest is maintained for DDR5 and LPDDR5 across PC, mobile server and other applications. Our conventional DRAM production capacity is limited by the ramp-up of HBM production. So inventories of DDR5 and LPDDR5 are tight. We are accelerating the transition to advanced process by increasing the scale of production cuts for legacy products to meet HBM demand upside and also to respond to demand for DDR5 and LPDDR5 next year.

As we gradually run through the inventories of legacy products through this flexible production plan to meet demand, we expect industry-wide inventories of legacy products to normalize in the first half of next year.

[Operator]

The following question will be presented by Nicolas Gaudois from UBS.

[Nicolas Gaudois, analyst]

Nick Gaudois from UBS. So going back to HBM, for HBM3 12Hi, you just mentioned that you're starting shipping in Q4 '24. That seems, first of all, earlier than peers, but also considering that NVIDIA may not need to put HBM3E 12Hi in servers before we think Q2 '25 and whatever AMD MI325 seems also to be using 8Hi, which effectively looks like a reasonably early timeline. So how could we explain this discrepancy?

And how do you look at your 12Hi ramp from here? On the back of this, do we think -- what do we think about your likely market share in HBM3E 12Hi compared to your already high market share in HBM3E 8Hi?

[Unknown Executive]

[Interpreted] In the early part of this year, we have announced that we started to supply HBM3E 8Hi for the first time in the industry. In this quarter, we have announced that the share of HBM3E has surpassed HBM3 -- and as I mentioned earlier, demand for HBM3E is going faster than we had originally anticipated.

HBM3E 12Hi started commercial production in September, and we'll start shipping this quarter as has been planned. With the new product launch of customer, the HBM market demand next year is expected to quickly shift from HBM3E 8Hi to 12Hi.

As a result, HBM3E 12Hi sales are expected to surpass HBM3E 8Hi sales in the first half of next year and the majority of sales in the second half of the year is expected to be 12Hi products.

We have made the necessary preparations to proceed with smooth commercial production as the increased number of staff increases the difficulty of process. And we will work closely with our customers to preempt the 12Hi demand.

As with the 8HI, our HBM3E 12Hi product will also take the lead in our market share. And as we have secured the industry's best product characteristics, we will continue to lead the HBM market through product differentiation.

[Operator]

The following question will be presented by Sun Woo Kim from Meritz Securities.

[Sunwoo Kim, analyst]

[Interpreted] I have a question about your investment for NAND. It seems to be that you have a rather conservative strategy when it comes to investments. You are making proactive and aggressive investments when it comes to the DRAM, but your investment style seems to be rather conservative when it comes to the NAND products. So what is the market outlook that you have for the NAND products? And what is your strategy to respond to this market situation?

[Unknown Executive]

[Interpreted] Thank you very much for your question. As we have stated since the beginning of this year, our NAND business is focused on profitability and investment optimization rather than focusing on outward growth.

As part of this strategy, we are focusing on investment efficiency by only spending CapEx on process convergence for products that are guaranteed to be profitable. And we plan to maintain a conservative capacity in investment posture until industry inventories returned to normalized levels, and we see a full-scale improvement in demand.

Also in the NAND business, we are strengthening our product mix toward higher-margin products such as enterprise SSDs, where we are seeing a clear demand recovery this year and we will further advance our portfolio by expanding our ultra-high capacity enterprise SSD product lineup.

As was shown in our Q3 results, we are focused on stabilizing prices even if it entails our Bit sales volume declining slightly. As a result, we expect the market share of our NAND products to decline this year in terms of volume, but in terms of revenue, our market share will be higher year-over-year.

While NAND demand is improving more slowly than expected this year, the NAND demand environment is expected to improve next year as AI PCs and smartphones are rolled out and replacement demand is generated.

To strengthen our NAND positioning and fundamentals to drive sustained profitability. We plan to maintain our profitability first strategy and investment optimization.

[Operator]

The following question will be presented by Jay Hyun Kwon from JPMorgan.

[H. Kwon, analyst]

[Interpreted] I have a question on HBM. It is said that HBM requires about 3x more wafers than conventional DRAM for the same chip production. But if we transition to HBM4, will the conversion ratio become even larger compared to a conventional DRAM?

[Unknown Executive]

[Interpreted] HBM requires a larger die size than conventional DRAM to achieve high bandwidth and high performance, and due to the nature of the package processing, where multiple layers are stacked, more wafer capacity is required to produce the same number of bits as conventional DRAM products.

HBM3E production requires 3 to -- 2x to 3x more additional wafer capacity compared to DDR products of the same capacity and 12Hi products is even more difficult to package than 8HI with lower maturity yields requiring more capacity.

The transition to HBM4 will require an increase in die size to enable higher bandwidth and higher performance characteristics, which will more than triple the conversion rate to conventional DRAM.

In the medium to long term, DRAM demand is expected to grow at a CAGR of mid- to high-teens, while HBM is expected to grow significantly above the DRAM average.

Given the increasing demand for HBMs and the rise in conversion rate, more wafers are expected to be required to produce the same bit with each new generation of HBM, which will continue to limit the supply of conventional DRAMs.

[Operator]

The following question will be presented by [indiscernible] from [indiscernible] Investment & securities.

[Unknown Analyst]

[Interpreted] First of all, let me offer my congratulations for your excellent performance this quarter. I have a question related to China. Recently, the aggressive capacity expansion by Chinese DRAM suppliers is of concern for the market as more Chinese suppliers are entering into the memory market. So my question is, what is your medium- to long-term response to China's entry into the memory market?

[Unknown Executive]

[Interpreted] As you have mentioned, due to the increased supply from Chinese memory suppliers, for legacy products like DDR4 and LPDDR4, the market is becoming more competitive.

However, the market where DDR4 and LPDDR4 -- DDR5 products are -- were adopted is already making the crossover to -- I'm sorry let me rephrase my answer.

The market where DDR4 and LPDDR4 products were adopted is already making the crossover to DDR5 and LPDDR5 and the newcomers are still far behind incumbents in terms of technology and product capabilities.

Depending on the application, price competitiveness may be more important than performance. But as AI capabilities become more prevalent on PCs and smartphones, the need for high-performance, high-end memory will increase.

Given the changing competitive landscape, we will be rapidly reducing our legacy offerings and selectively focus on higher value-added markets where we expect mid- to long-term growth.

In the case of DDR5 and LPDDR5 as well, we plan to continue to maintain our lead over the late comers by developing products that support even higher speeds.

[Operator]

The following question will be presented by Dong Hee Han from SK Securities.

[Dong Hee Han, analyst]

[Interpreted] I am Han Dong Hee from SK Securities. I have a question related to your capacity investment plans. So may I ask what your investment plans is for capacity related to 2024 and 2025? And if there is a downturn in the memory market that is more serious than what you have expected. How do you plan to reduce such investment plans going forward?

[Unknown Executive]

[Interpreted] Thank you very much for the question. Let me answer your question. Investment for this year is expected to be in the mid- to high KRW 10 trillion range, slightly higher than planned at the beginning of the year reflecting faster-than-expected growth in HBM demand and also the decision to invest in M15X.

And although the exact investment amount for next year 2025 is not yet finalized, 2025 is expected to be slightly higher than this year due to investments for a stable supply of HBM in-short due 1b nanometer transition and securing TSV capacity investments for our products for which demand has already been secured through supply agreements with customers, including back-end process capacity. Transition investments to ramp up production of D5 and LPD5 while reducing legacy products and also continued investment into M15X and the Yongin infrastructure.

However, the impact of this on short-term production growth is limited given that most of the increase in investment is in infrastructure, R&D and back-end processing.

We are considering multiple scenarios for next year's investments and we'll continue to select and continuously review the signpost that may influence our investment decisions.

Our company continues to maintain the principle of only investing in products that are profitable and we will be flexible in our investment decisions as the market evolves, and this is how we're going to respond to the market situation.

[Operator]

The following question will be presented by Ricky Seo from HSBC.

[Ricky Seo, analyst]

[Interpreted] If you look at Bit growth and ASP performance, it appears that it will be at a level that is similar to the previous quarter -- the NAND sales. And I believe there is also some improvement in your margins. This may be due to the product mix improvement that you have mentioned during your presentation?

And -- or is there any one-off expenses that has been reflected in the third quarter such as the write-back of the losses on valuation of inventory and other such one-off expenses?

And also going forward, in the fourth quarter and onwards, do you have any remaining one-off expenses that needs to be further reflected?

[Unknown Executive]

[Interpreted] In the third quarter, NAND revenue declined very slightly Q-o-Q. Shipments declined due to a focus on eSSD products where demand continued to be strong, and sales responses were driven by real demand. But higher ASPs nearly offset the decline in shipments.

Cost of goods sold decreased due to lower sales volumes, while revenue was maintained due to higher prices, resulting in improved profitability, and the share of highly profitable eSSDs and the revenue exceeded 60%, contributing significantly to improved profitability.

Regarding the one-off expenses such as inventory valuation losses or reversals, most of the inventory valuation provisioning recognized during the downturn has already been reversed, so the impact of inventory valuation profit or loss was minimal this quarter. And other than this factor, there was no significant increase in one-off expenses that affected profitability.

[Seong Hwan Park, executive]

[Interpreted] We will now receive the last question.

[Operator]

The last question will be presented by Young Ho Ryu from NH Investment & Securities.

[Young Ho Ryu, analyst]

[Interpreted] First of all, I would like to congratulate you on your record high quarterly performance this quarter. And I have a question on HBM4. So what is the HBM4 development status? And do you believe you can maintain the technical leadership in HBM4 as well. And one of the biggest differences that we see forthcoming related to HBM4 is, of course, the logic die and so with regard to logic die, it seems that the cooperation with the customers and the foundries will be even more important going forward. So would this mean that there would be any changes in the way that you cooperate with your customers and foundries going forward?

[Unknown Executive]

[Interpreted] Thank you very much for your questions. In the case of our company, in order to develop HBM products with the best characteristics and quality, we have been identifying technical issues from the very pre-planning stage of HBM products with our customers. And using the customer feedback, we have been improving our internal baseline, increasing consistency and enhancing the completeness of our products.

HBM4 will see a lot of technical changes, including the doubling of the I/O, new schemes for using less power and utilizing our logic foundry for the first time.

This requires a much more deeper technical exchange going beyond the scope of traditional testing, and this is the reason why we and our foundry partners and the teams involved in the project are working together in a "One Team" structure.

We believe that timely delivery of next-generation products with improved performance in line with the rapidly evolving technology pace of semiconductors for AI is a key factor in securing leadership in the HBM market going forward.

Unlike general DRAMs, which are by nature commodity products, the scale of production for HBM is determined based on customer demand. So it is more important to reliably supply quality products that meet customer requirements in a timely manner rather than to use advanced processes to reduce cost by reducing chip size.

In addition, since the chip structure of HBM has a small area for the cell, the benefit of reducing the chip size by applying advanced processes is limited. So we are preparing to supply HBM4 products with stable quality in a timely manner by applying the 1b nanometer and advanced MR-MUF technology, the stability and mass production capabilities, which have been already proven. And our aim is to ship to customers the HBM4 products in the second half of 2025 as was planned.

This will ensure not only timely supply but also investment efficiency so that we can continue to defend our market leadership and profitability.

[Seong Hwan Park, executive]

[Interpreted] Thank you very much. With this, we'd like to conclude the SK Hynix 2024 Q3 Earnings Release Conference Call. Thank you very much.

[Portions of this transcript that are marked [Interpreted] were spoken by an interpreter present on the live call.]