SK hynix reported a 557% surge in second-quarter operating profit, though it fell short of market expectations. The company's capital expenditure for the year is at least $31 billion, amid growing concerns about overinvestment in artificial intelligence capacity.
Following the earnings release, Seong Hwan Park, head of investor relations, and Hyun-Jong Song, head of the corporate center, among other executives, provided a detailed review and answered analyst questions. The following is a summary of the key points from the Q&A session of the conference call.
Regarding AI Infrastructure Investment Trends
An analyst from JPMorgan queried whether the trend of large tech companies considering self-built leased data centers and the emergence of efficient, lightweight AI models might lead to a slowdown or decline in AI infrastructure capital expenditure. The analyst asked how SK hynix assesses the long-term AI investment trajectory of major cloud service providers and its impact on demand for HBM, general server DRAM, and enterprise NAND flash.
SK hynix responded that these changes, including the rise of efficient models, are not a signal of shrinking AI investment but rather reflect two industry transformations: improved utilization of existing large-scale AI computing infrastructure and accelerated commercialization of AI by enterprises. For leading cloud service providers, AI capabilities are a core competitive advantage, so the long-term logic for increasing AI investment remains intact. Furthermore, lightweight models will lower the barrier to AI adoption, broadening its user base and applications, which in turn will increase infrastructure demand rather than reduce it. This is supported by communication with core customers. While project timelines may vary due to constraints like power supply and facility construction, intensifying competition and expanding AI services will sustain robust AI infrastructure investment at least beyond next year. Consequently, overall memory chip demand will continue to grow, driven by rising HBM demand for AI computing and increased demand for server DRAM and high-capacity enterprise NAND flash.
Explanation of Long-Term Expansion Plans
An analyst from Hana Financial Investment inquired about the rationale behind SK hynix's recently announced large-scale, long-term capacity expansion plan. They asked whether the planned capacity includes orders locked in by long-term supply agreements and how the company addresses market concerns about future oversupply.
SK hynix explained that the core basis for its long-term capacity plan is the structural long-term growth of memory chips driven by AI industry expansion, combined with long-term demand expectations from continuous dialogue with core customers. The company's relationship with clients has evolved into a deep-strategic, long-term partnership, with customers proactively pursuing long-term supply agreements. The planned capacity expansion is entirely based on the demand certainty locked in by these strategic collaborations. The actual capital expenditure and capacity ramp-up will be executed in stages, considering factors like demand visibility and investment efficiency. The expansion pace is fully flexible to match confirmed customer orders, so the company does not believe this plan will directly lead to industry oversupply.
Details on Long-Term Supply Agreements
An analyst from Meritz Securities asked for more details on the structure of SK hynix's long-term supply agreements, including contract duration and pricing mechanisms, noting that competitors have recently announced such agreements.
SK hynix stated that there is no standard template for its long-term agreements. Terms are customized based on customer type and product category. The typical contract duration is around five years, and pricing mechanisms are also negotiated on a case-by-case basis to mitigate volatility from short-term market price fluctuations and ensure stability for both parties. These agreements include long-term purchase commitments and mechanisms like deposits to enhance predictability. The company declined to disclose the specific proportion of total revenue covered by such agreements, but it maintains a reasonable ratio depending on market conditions and customer needs. This model strengthens resilience against downturns while retaining flexibility to capture incremental orders during upcycles. Through long-term cooperation with key AI clients like NVIDIA, the company has established a stable HBM-based profit base.
DRAM Average Selling Price Dynamics
An analyst from Daiwa Capital Markets congratulated the company on its performance and asked why the second-quarter DRAM blended average selling price increase was below market expectations and what the outlook is for the second half of the year.
SK hynix attributed the lower-than-expected ASP to a dynamic adjustment of its HBM and general DRAM shipment mix in the second quarter, with some high-value product shipments deferred to the second half. This negative impact is expected to diminish in the second half. With the volume ramp-up of HBM4 and increased shipments of 1c-nanometer general DRAM, bit shipment growth in the second half will be higher than in the first half. The growing share of high-value products like HBM4 will positively contribute to the blended ASP. The company's sales strategy prioritizes demand certainty, long-term customer relationships, and supply-demand dynamics over solely chasing short-term price fluctuations.
Maintaining HBM Market Leadership
An analyst from SK Securities asked about the core competitiveness of SK hynix's HBM4, given the accelerated development pace of competitors, and what differentiates the company in maintaining its leading position.
SK hynix emphasized that HBM competitiveness is not just about meeting performance standards but also about stable yield rates and mass production capabilities. Its accumulated experience since the HBM2E era, product development cycles, performance, yields, quality, and customer trust are core advantages competitors cannot easily replicate. The company has already begun mass delivery of HBM4 to core customers in the second quarter, with yields and quality levels close to those of mature HBM3E. The immediate focus is on scaling capacity. HBM4E samples have been delivered, and development is on track for 2027 mass production. The company is also investing in next-generation technologies like its proprietary iHBM, which integrates heat dissipation into the package, reducing thermal resistance by over 30%. As AI scales, customers will increasingly value partners with proven mass production capabilities, stable quality control, and reliable supply. SK hynix's early joint development with customers and long-term cooperation will allow it to continue leading technological iterations and maintaining its HBM market leadership.
HBM Pricing Negotiations for 2027
An analyst from UBS asked about the progress of HBM price negotiations for 2027 and the overall pricing expectations for HBM3, HBM4, and the upcoming HBM4E for the full year.
SK hynix confirmed it is currently in negotiations with core customers for 2027 HBM supply volumes and pricing. Customer demand is strong, and negotiations are progressing smoothly, but the company did not disclose specific details. While the recent sharp rise in general DRAM prices can influence HBM pricing discussions, HBM pricing is not solely driven by general DRAM trends. HBM production requires more wafers, advanced manufacturing processes, TSV technology, and high-end packaging. Pricing negotiations consider multiple factors, including general DRAM market prices, supply-demand dynamics, the resources and opportunity cost of HBM production, and the differentiated value created for customers. The company's pricing goal is to achieve reasonable profits that match its technological value while fostering a healthy AI ecosystem. The HBM business will maintain stable profitability through technology leadership, cost advantages, and trusted customer relationships.
Global Investment Strategy and Overseas Expansion
An analyst from CLSA Korea asked about the company's overseas investment strategy, given the frequent market discussions about building new factories in the US and Japan, in addition to the recently announced large domestic investment in South Korea.
SK hynix stated that delivering capacity on time and in full is a core competitive factor in the AI era. The company's long-term investment strategy is to match AI memory demand with timely capacity investment, strictly adhering to criteria of commercial viability and investment efficiency. The strategy involves maximizing the utilization of existing facilities while building new production bases as needed. In Korea, the Icheon and Yongin campuses will be core hubs for next-generation DRAM and AI memory, while the Cheongju site will enhance NAND flash and advanced packaging. For new overseas facilities, the company will not simply distinguish between domestic and international locations. The decision will be based on a comprehensive evaluation of factors like electricity, water resources, talent, semiconductor ecosystem, and proximity to customers. Beyond previously announced plans, there are no new overseas factory decisions at this time. The company will continue to make location decisions based on customer needs, maximizing existing assets to improve efficiency.
NAND Flash Strategy for the AI Era
An analyst from Daishin Securities asked about SK hynix's strategy for enterprise SSDs, noting the increasing importance of this segment due to AI inference scaling and the growth of external KV cache. They inquired about the company's positioning in different product lines, such as QLC-based SSDs for HDD replacement and SLC-mode high-performance SSDs.
SK hynix explained that the AI industry is shifting from training to inference, making NAND flash a core component in the AI memory hierarchy, with enterprise SSD demand rising rapidly. The strategy is not to choose between SLC, TLC, or QLC, but to offer a customized product portfolio optimized for different customer workloads. For AI data lakes and HDD replacement, cost and capacity are key, making QLC-based enterprise SSDs the optimal solution. For emerging applications like external KV cache and near-GPU memory, new layered AI storage solutions are being developed. The company will leverage firmware optimization to maximize the advantages of each flash memory type, covering high-performance TLC, high-capacity TLC, and SLC-mode enterprise SSDs. This comprehensive product matrix will allow SK hynix to match the evolving needs of the AI memory market, opening new long-term growth opportunities for its NAND business.
ADS Conversion Mechanism and Future Plans
An analyst from NH Investment Securities asked about the company's management of the ADS conversion mechanism and whether it plans to increase the proportion of outstanding ADS.
SK hynix stated that since July 30, ADS can be freely converted into ordinary shares listed on the Korean Stock Exchange. However, the reverse conversion of ordinary shares into ADS has procedural and volume limitations. The total number of outstanding ADS is capped at 17.79 million shares, equal to the total issuance in this offering. Any future increase in the ADS ratio will be evaluated based on the regulatory environment and other external factors, with no decision made at this time.
Capital Allocation Strategy and Shareholder Returns
An analyst from DS Investment Securities inquired about the company's overall capital allocation strategy following the sale of its stake in Kioxia and the ADS listing, specifically asking if there will be any additional shareholder return measures this year.
SK hynix outlined its capital allocation strategy, which balances three objectives: making timely investments to capture AI-driven growth, maintaining a sound financial structure, and enhancing shareholder value through returns. The company acknowledged that the market is highly focused on shareholder returns and is currently evaluating various measures. However, due to regulatory requirements related to the ADS listing, it cannot disclose any material new information beyond the offering documents. The specific form, size, and timing of any shareholder return plan cannot be shared at this time, but the company will announce its final plan within this year. Long-term, SK hynix aims to build a balanced capital allocation system that simultaneously supports business growth with capacity investment, maintains a robust financial structure, and delivers consistent shareholder returns based on sustainable cash flow.
Comments