Hyundai Electronics’ Q2 earnings blew expectations out of the water, and Wall Street froze: the surprise gap is the core issue

SK hynix released what is arguably its strongest Q2 earnings ever, but the market reaction fell short of expectations—so the expectation gap ended up becoming the core of the trading.
(Background: Crazy! SK hynix’s ADR premium surged to 51% within three days of listing; Barclays called for a target price of $330—memory shortage has only just begun)
(Additional background: Samsung surged 20% in memory! UBS raised its estimates: DRAM Q3 up 32%, NAND up 30%!)

Table of Contents

Toggle

• SK hynix Q2 earnings: stock price freezes due to missed expectations
• Performance details: revenue hits a new high, margins jump to 76%
• AI demand outlook: Agentic AI drives a structural surge
• Supply bottleneck: LTA agreements strengthen long-term cooperation
• Product roadmap: HBM4 mass production, NAND technology upgrade
• Capital expenditure: 2026 expansion plan and shareholder returns

SK hynix just posted its strongest-ever Q2 earnings, but because market expectations were too strong, the market reaction was weak. Operating profit grew sharply, yet remained below market expectations. Revenue also missed expectations, and the stock fell after hours. The impressive performance did not lead to more valuation premium—instead, the expectation gap became the trading’s focal point.

In fact, the market had already been anticipating this expectation gap issue not long ago. Korea Investment & Securities (KIS) cut SK hynix’s profit outlook two weeks earlier, but they did not expect how fragile the market was—more severe than we imagined.

There is chatter in the market about the phone conference after SK hynix’s earnings. The CEO’s comments themselves did not contain much that was truly new; the questions from Wall Street institutions afterward were more worth watching. Because they asked what capital is most focused on.

In Q2, strong demand driven by expanded AI infrastructure investment, alongside a tight supply environment, kept prices trending upward. DRAM and NAND recorded further clear price increases again compared with the previous quarter. Server DRAM and enterprise SSDs—key AI-related products—were the main drivers of this round of growth. Q2 revenue was 79.3 trillion won, up 51% quarter-over-quarter and up 257% year-over-year, setting a new historical high again after the prior quarter.

With limited supply capacity, the company expanded sales centered on HBM and AI server DRAM products, and shipment volume achieved high single-digit quarter-over-quarter growth, in line with earlier guidance. Shipments of server LPDDR products (including SOCAMM) grew significantly; driven by continued strength in traditional DRAM prices, DRAM ASP rose about 30% quarter-over-quarter.

On top of the low base created by shipment contraction in Q1, combined with expanded enterprise SSD sales, NAND shipment volume recorded mid-double-digit quarter-over-quarter growth, also in line with guidance. Enterprise SSD revenue doubled quarter-over-quarter. Enterprise SSDs of 30TB and above increased more than twofold versus the prior quarter. Driven by broad-based strength across product prices, synchronized price hikes at both DRAM and NAND, and improved cost structure, NAND ASP rose at a level in the mid-50% range quarter-over-quarter (up about 50%+).

Q2 operating profit was 60.5 trillion won, up 61% quarter-over-quarter and up 557% year-over-year. The operating margin rose by 5 percentage points quarter-over-quarter to 76%. Both operating profit and operating margin hit historical highs. Q2 depreciation and amortization was 4 trillion won, and EBITDA was 64.6 trillion won, with an EBITDA margin of 81%. Net non-operating gains were 62.2 trillion won, including net FX gains of 1.1 trillion won from exchange rate movements, and gains from disposal and valuation of investment assets totaling 63.3 trillion won. On that basis, profit before tax was 122.7 trillion won, net profit was 93.9 trillion won, and net margin was 118%.

As of end-Q2, cash and cash equivalents (including short-term investments) were 88 trillion won, up 33.6 trillion won from end of the prior quarter. Interest-bearing debt decreased by 0.7 trillion won to 18.6 trillion won. Net cash expanded to 69.4 trillion won, and the debt-to-equity ratio improved by 5 percentage points from end of the prior quarter to 7%.

AI technology is evolving into an Agentic form that can replace users to carry out complex tasks over long periods. As AI penetrates various services such as search, programming, and productivity tools, the scope of demand continues to expand. From a storage perspective, in addition to high-performance storage like HBM that is necessary to boost AI server performance and expand system scale, demand for server DRAM that supports Agentic services is also increasing. To process the outputs continuously generated by AI more efficiently, high-performance enterprise SSDs are also playing a growing role. What this forms is a structural shift in which AI storage and traditional storage grow in parallel.

SK hynix Q2 earnings: expectation gap triggers a rapid stock-price freeze

At the same time, as AI model improvements and software optimization progress, compute consumption and costs per task continue to decline. The company believes efficiency improvements will not suppress overall infrastructure demand; instead, they will lower the pricing and usage barriers of AI services, thereby expanding the user base and application scope.

Major large technology customers are expanding infrastructure investment due to increased usage of AI services and compute shortages. Supported by revenue and profit growth from AI services, their storage procurement still appears to be continuing to expand. In fact, major customers are still asking for more storage supply. In PC and mobile application areas, due to difficulty obtaining storage, there have been some short-term sales adjustments; however, as supply tightness eases and AI services accelerate in adoption, these two segments are expected to gradually regain growth momentum.

Under supply constraints, the company expects DRAM demand growth in the mid-20% range and NAND demand growth in the high single-digit range (around 10% at the high end). If future supply constraints ease and suppressed potential demand is satisfied, the market’s growth trajectory could move even higher.

On the supply side, a material improvement in the supply-demand balance is unlikely in the near term. The reasons are that the complexity of advanced processes used in HBM and AI server storage continues to rise, and the construction cycle required to build new capacity. The company expects the tight supply-demand situation to persist for a considerable period of time.

To support long-term supply stability, the company is discussing multi-year contracts (LTA) with customers as part of its long-term approach. To date, it has completed LTA negotiations with around 10 customers, including key customers, and continues discussions with other major industry participants. These LTAs are not just about quantity arrangements; they are strategic partnerships: used both to ensure stable long-term supply and to support customers’ technology roadmaps to develop next-generation storage products. Specific pricing structures vary by customer and product characteristics. The design is intended to respond to price volatility. At the same time, financial mechanisms such as deposits are introduced to secure contract performance and improve visibility and reliability of customers’ long-term demand plans. Based on this, the company will enhance investment and production operating efficiency, strengthening the foundation for stable long-term operations and sustainable growth.

Q3 DRAM shipment volume is expected to grow by about 10% quarter-over-quarter from Q2. The company will focus on server products and actively respond to demand; NAND bit shipments are expected to grow by low single digits quarter-over-quarter.

As AI models become increasingly complex and requirements for storage performance continue to rise, the competitive landscape has expanded beyond the design of single storage products into system architectures and packaging technologies. The company will lead storage innovation at the system level, leveraging its DRAM and NAND product portfolio including HBM, as well as joint development capabilities with customers.

HBM4: Through continuous product optimization, it achieves industry-leading energy efficiency and cost competitiveness while meeting the data processing speeds customers need—demonstrating differentiated technology capabilities. The company began mass production shipments in Q2 and plans full ramp-up in the second half of the year.

HBM4E: In the first half of the year, samples were sent to a major customer. The product uses an optimized manufacturing process that has matured and passed validation for mass-production stability. The subsequent development schedule is expected to proceed smoothly. The company will maintain its leading position in HBM by building on stable supply capabilities and cost competitiveness supported by high yields and excellent quality, as well as industry-leading performance.

Performance details: revenue hits a new high, and margins soar to 76%

Traditional DRAM: In Q2, it began supplying SOCAMM2 products based on a 1c nm process across the board. Going forward, the company will optimize its product line according to customers’ development schedules and prepare samples to expand its customer base.

NAND: Accelerating the transition to advanced processes, and strengthening the product mix with a focus on high-capacity, high-performance products to match market demand. In the previous quarter, the 321-layer products accounted for the highest share in NAND production, and the company plans to raise their domestic capacity share to around 50% by the end of the year according to the original timeline.

In a market environment where supply-demand imbalances persist, stable supply capability—delivering the required quantity at the time customers request—alongside technical strength is becoming a core competitive operating advantage. To respond to strong customer demand and long-term growth opportunities, in the near term the company will continue to push expansion investment to improve supply responsiveness: advancing the mass-production timeline of M15X and increasing investment to quickly expand capacity. The Yongin Fab 1 is expected to complete cleanroom line opening in early 2027. Due to earlier progress and a larger investment scale, 2026 capital expenditures are expected to reach the high end of the 40 trillion won range.

On a medium-to-long-term basis, the company will proactively secure infrastructure foundations for future capacity, based on discussions with customers and market demand forecasts. Recently, it announced new investment plans to strengthen advanced packaging capabilities and a new NAND production base, M17. It also announced medium-to-long-term plans to build a new semiconductor cluster within South Korea to address long-term demand after Yongin. Subsequent actual construction, equipment installation, and capacity expansion will be implemented in phases, considering factors such as customer demand visibility and investment efficiency. While maintaining capital expenditure discipline, the company will not miss long-term growth opportunities, and will strengthen both supply responsiveness and financial soundness.

On July 10, its ADR successfully listed on the Nasdaq market in the United States, marking the largest issuance in size among IPOs in the U.S. by a foreign company. The significance of this listing is not only about raising funds, but also about confirming the global market’s trust in the company’s technological competitiveness and growth potential, while also widening connectivity points between the company and the next-generation computing ecosystem. Based on this, the company will strengthen strategic cooperation with major customers and partners, uncover new business opportunities, and—through continuous technological innovation—contribute to the development of the semiconductor industry and the growth of AI systems.

With the expansion of its profit and cash-generation capabilities to historical highs, the company’s financial capabilities have further strengthened. At the same time, structural growth opportunities in the AI era continue to expand, and the investment scale needed to realize these opportunities is significantly higher than in the past. In this environment, the company will prioritize investments into growth opportunities that can generate high profitability and strategic value, while also building a financial structure that can ensure stable operations even amid market volatility. The results generated will continue to be shared with shareholders.

Even though future investment demand is expected to rise, the company believes its significantly strengthened cash-generation capability is meaningful enough to expand shareholder returns meaningfully while achieving future growth investment objectives and maintaining financial soundness goals. Currently, the company is reviewing various additional execution options for shareholder returns from multiple angles.

Question: Some large technology companies are considering leasing data centers recently, and as more efficient AI models keep appearing, the market is concerned that AI infrastructure investment may slow down or even decline. Based on the company’s communication with customers, how do you view the direction of major CSPs’ AI infrastructure investment? What does that imply for demand for HBM, DRAM, and NAND?

Answer: The company understands the origin of these concerns. But the company does not view these moves as signals of AI investment slowing down; rather, it sees them as a shift toward increasing utilization of already built AI infrastructure and accelerating monetization.

AI demand outlook: Agentic AI drives a structural surge

For major CSPs, AI competitiveness is tightly linked to their core competitiveness such as search, advertising, cloud services, and software. Therefore, investments aimed at strengthening AI capabilities are very likely to remain steady.

Likewise, the company does not believe that higher-efficiency AI models will reduce infrastructure demand. When model and system efficiency improve, the same set of infrastructure can support more users and services, which in turn expands AI accessibility and adoption. The recent emergence of high-efficiency AI models and the resulting surge in demand is evidence: what efficiency improvements bring is broader AI adoption and usage, not a decline in infrastructure demand.

This assessment is also supported by the medium-to-long-term demand outlook discussed with major customers. Although the timing of individual projects may differ due to physical conditions such as power supply and data center construction, under the ongoing support of AI competition among CSPs and continued expansion of AI services, the company believes AI infrastructure investment will remain steady even after next year.

Therefore, overall storage demand will continue to expand: not only HBM for AI compute, but also server DRAM supporting Agentic AI, and high-performance, high-capacity NAND that will handle the expansion of AI services and data growth.

Question: The company recently proposed a major capacity expansion plan over the medium to long term. What are the basis for the long-term storage demand outlook supporting this strategy? Does it include demand locked in through long-term agreements? Also, with the increase in capacity, the market is worried about potential oversupply—how does the company view this?

Answer: The company’s medium-to-long-term capacity strategy is built on structural growth in the storage demand driven by AI expansion, along with ongoing discussions with core customers about longer-cycle demand.

Recently, cooperation with customers is evolving from a transactional relationship into a more strategic long-term partnership. The stronger willingness from customers to sign long-term agreements and build partnerships is itself evidence of sustainable AI ecosystem demand.

The capacity expansion plans currently set by SK hynix are based on the market demand visibility obtained through customer partnerships. The actual capital expenditures and capacity ramp-up will be implemented in phases after comprehensively considering demand visibility and investment efficiency, with expansions flexibly aligned to already confirmed customer demand. The company believes that medium-to-long-term investment plans will not immediately lead to oversupply.

Question: This question is about long-term agreements (LTA). Peers have recently completed and published their LTAs. While the company briefly mentioned it in its earnings briefing, could you provide further explanation of SK hynix’s LTA framework, such as contract duration and pricing structure?

Supply bottleneck: LTA agreements strengthen long-term cooperation relationships

Answer: The LTAs the company is discussing with customers take multiple forms and are designed differently according to each customer and product characteristics. Contract duration is typically around five years, with specific conditions varying by customer and product.

The pricing structure also will not be uniform. The company is exploring with customers various pricing mechanisms that can better respond to price volatility. The goal is to reduce uncertainty caused by short-term market volatility while improving long-term operating stability for both customers and the company.

At the same time, given the impact of demand volatility on the storage cycle, obtaining effective procurement commitments is also important. Therefore, beyond long-term quantity commitments, the agreements also include mechanisms such as deposits to strengthen contract execution and demand visibility. The detailed terms vary by each customer’s requirements and contract structure. This structure allows customers to set more reliable long-term procurement plans, while enabling the company to optimize investment and production planning based on improved demand visibility.

On how large a proportion of sales the LTAs will cover, the company cannot provide a specific number at this time, but it will keep it at an appropriate level based on market conditions and customer demand: improving downside resilience for performance while retaining flexibility to capture incremental demand and growth opportunities when the market turns better.

The company has built a solid base for profitability in HBM through long-term cooperation with major AI customers such as NVIDIA. Going forward, the company will continue to strengthen its leading position in HBM, and achieve a balance between stability and profitability by leveraging the demand visibility and operational flexibility obtained through LTA.

Question: This question is about DRAM. In Q2, the DRAM ASP increase appears to be lower than the market expected—what is the reason? What is the outlook for the second half?

Answer: The company manages the sales mix between HBM and traditional DRAM based on customer demand and the medium-to-long-term product strategy. In Q2, shipments of certain high value-added products were delayed to the second half. The resulting change in product mix affected the mixed ASP; these factors will gradually ease in the second half.

As HBM4 shipments fully ramp up and shipments of traditional DRAM based on 1c nm increase, the bit growth in the second half is expected to be higher than in the first half.

In addition, considering changes in customer demand and product mix, the growth in HBM4 sales and the increasing contribution from high value-added products will also have a positive impact on the mixed ASP. Higher shipment volumes combined with ongoing improvements in product mix will drive ASP and performance higher in the second half.

Product roadmap: HBM4 to mass production, NAND technology upgrade

On the sales strategy side, the company does not center on short-term price volatility or short-term profits. Instead, it comprehensively considers demand visibility, long-term customer relationships, and the supply-demand conditions in each product subdivision. This principle will remain unchanged. While capturing market growth opportunities, it will deliver stable and sustainable earnings growth.

Question: This question is about HBM. Some in the market believe competitors have made fast progress with HBM recently. Where does the competitiveness of SK hynix HBM4 show up? What are the key differentiators to maintain a leading position in the HBM market?

Answer: The competitiveness of HBM4 depends not only on whether it can deliver the required performance, but also on whether it can achieve scale supply with stable yields and consistent quality.

Since the HBM2 generation, SK hynix has continuously validated these capabilities. The competitive strengths accumulated at launch timing, product performance, mass-production yields, quality, and customer trust cannot be replicated in the short term.

Based on this, the company has begun HBM4 mass production for core customers in Q2. Currently, the yield and quality of HBM4 are close to mature-stage HBM3 levels. The focus now is to steadily expand capacity.

As mentioned earlier, the company has completed customer sampling for HBM4E. This product uses an optimized manufacturing process with both proven technical maturity and mass-production stability. Development is progressing smoothly according to the roadmap, with a goal of starting batch production in 2027.

The company is preparing beyond this as well, with a forward-looking layout for the next generation of technology. In addition to hybrid bonding, it is developing heat dissipation technologies for future products such as HBM5. This technology integrates heat-dissipation elements inside the package, and is expected to reduce thermal resistance by more than 30%, thereby improving system stability and execution efficiency in high-performance, high-density AI environments. (This technology abbreviation reads as “IBM” in transcription; here it is described according to its function.)

As the AI market continues to expand and AI accelerators become increasingly complex in both performance and packaging, the company believes customers will value partners with validated manufacturing capabilities, quality, and reliable supply even more. HBM is a high value-added product. If quality issues arise, it imposes huge costs on customers and can have broad impacts across the entire system.

Leveraging early joint development experience with customers and a long-term strategic partnership, the company will continue to reliably deliver the right products at the right time and lead the migration to next-generation technologies, thereby maintaining its leading position in the HBM market.

Capital expenditure: 2026 expansion plan and shareholder returns

Question: How is the current progress on 2027 HBM price negotiations? Could you explain the contract discussions, including those for HBM4E and HBM4, as well as the outlook for actual pricing?

Answer: The company is in discussions with core customers regarding HBM supply volumes and pricing for 2027, and the negotiations are progressing smoothly supported by solid customer demand. Specific contract terms and pricing details for individual customers cannot be disclosed.

In recent months, traditional DRAM prices have surged significantly, and this market environment may have some influence on HBM price discussions. However, HBM prices are not determined solely by traditional DRAM prices.

Compared with traditional DRAM, HBM requires far more resources, including more wafer processing, advanced manufacturing processes, and TSV plus packaging capacity. As each product generation evolves, customers’ requirements for performance and quality continue to rise, and product development and certification become increasingly complex.

Therefore, the company’s price discussions will take into account a range of factors: traditional DRAM prices and market supply-demand, resources and opportunity costs related to HBM production, the degree of technical complexity, and the value products create for customers. The goal is to secure a reasonable profitability level that matches the differentiated value being provided while supporting a healthy, sustainable growth of the AI ecosystem.

Relying on accumulated technical leadership advantages, cost competitiveness, stable manufacturing capability, and the trust and collaboration with customers, the company will maintain solid profitability in its HBM business through successful product-generation transitions and continued customer value creation. It will also reinforce its position as a strategic partner for joint growth with customers in the AI era, focusing on long-term sustainable growth and profitability.

Question: This question is about capacity expansion. In addition to the large-scale investments recently announced within South Korea, the market is also discussing expansion overseas to the U.S., Japan, and elsewhere. Could you elaborate on the company’s investment strategy and direction domestically in South Korea and abroad?

Answer: In the AI era, having technological leadership alone is not enough. Whether you can supply the required quantities at the right time is also a key part of competitiveness. Especially in periods when supply is extremely tight, it is the supplier’s responsibility to provide the storage products the ecosystem needs.

The company’s medium-to-long-term investment direction is: invest in line with AI storage demand at the appropriate time, while executing capital expenditures based on business feasibility and investment efficiency. Over the medium to long term, it will ensure additional manufacturing capacity through an optimal mix of maximizing utilization of existing production bases and building new infrastructure when necessary.

Within South Korea, the company will continue to treat Icheon and Yongin as the core manufacturing hubs for next-generation DRAM and AI storage, while strengthening Cheongju’s manufacturing capabilities in NAND and advanced packaging. (In the transcript, the name of the hub may be missing or misidentified here; described according to the company’s current production base references.) The large-scale investments that have already been announced are also part of this strategy, proactively ensuring the manufacturing bases and infrastructure needed to support future demand.

Regarding future production bases, the company does not distinguish broadly between domestic and overseas locations. The basic direction is to make the best decisions considering factors such as power supply, water resources and manpower, supply chain and semiconductor ecosystem, and customer accessibility.

To clarify, as of now, there are no other decisions besides the investments already announced. Going forward, the company will continue to secure production bases at the right times to respond to customer demand, and improve investment efficiency by leveraging existing assets while also evaluating additional investments.

Question: This question is about NAND. As reasoning demand expands and KV cache uninstallation demand increases rapidly, how is the role of enterprise SSDs changing? Could you explain the company’s strategies across various product segments, including QLC SSDs as replacements for HDDs, and high-performance SSDs based on an SLC mode? Competition in these submarkets also seems to be intensifying.

Answer: As the question observes, the AI market is shifting from a training-centered environment to a reasoning-centered environment, and NAND is quickly becoming a core component in the AI storage layer. Therefore, NAND demand centered on SSDs is rising rapidly, and the company believes this trend is likely to continue.

At the same time, the company believes the AI storage market cannot be covered by a single technology. Requirements for latency, throughput, power consumption, capacity, and TCO vary by customer. What customers require is not necessarily a specific technology or a specific medium—the key is whether it can reliably deliver the performance required for each type of workload and the