Is the slowdown in AI investment purely due to worries? SK Hynix conference call: Leading cloud vendors are still investing heavily, and the probability of mid- to long-term oversu

Is the slowdown in AI investment purely due to worries? SK Hynix conference call: Leading cloud vendors are still investing heavily, and the probability of mid- to long-term oversupply is low.

Early this Wednesday morning, Korean storage giant SK Hynix held a conference call with results for the second quarter of fiscal year 2026. Although the company's revenue and profits hit record highs, they were still lower than consensus market expectations. On Wednesday morning, SK Hynix plunged over 11% again in the Korean stock market. During the earnings conference call, the company's management made clear optimistic judgments regarding core market concerns such as global AI demand, long-term cooperative production capacity, and HBM iteration. Song Hyun-jong, president of SK Group, said at the conference call: "We judge that the mid- to long-term investment in AI by leading cloud vendors will continue, and the storage demand data connected with customers also confirms this trend...In the context of industry competition among cloud vendors and the continued expansion of AI services, infrastructure investment will remain strong starting next year." 1. AI demand remains strong, and there is no need to worry about a slowdown in capital expenditures. SK hynix management denies that AI investment is on the wane. They pointed out that efficient large models and data center leasing are means to optimize computing power utilization, not to reduce investment; agents and inference scenarios continue to generate increases in HBM, server DRAM, and enterprise SSD, and investment in AI infrastructure will remain stable after 2027. 2. 5-year long-term cooperative locking of forward orders to avoid the risk of overcapacity SK hynix has signed 5-year long-term supply agreements with 10 core customers, and the contracts are equipped with deposits and flexible price adjustment mechanisms to hedge against cyclical fluctuations. They emphasized that the production capacity expansion fully matches the actual purchase orders of customers and is put into production in stages. The probability of oversupply in the medium and long term is low. 3. The product iteration route is clear and shipments are accelerated throughout the year. The company's HBM4 will be mass-produced in the second quarter and will ramp up on a large scale in the second half of the year; samples of the next-generation HBM4E have been delivered to customers and will be officially mass-produced in 2027. The company's self-developed IHBM packaging technology optimizes heat dissipation and consolidates technical barriers. In the third quarter, DRAM bit shipments increased by 10% quarter-on-quarter, NAND saw low-single-digit growth, and the proportion of high value-added products increased in the second half of the year, driving up the average price. A full range of SLC, TLC, and QLC enterprise SSDs have been launched for KV cache, data lake and other scenarios to adapt to TCO requirements of different computing power. AI inference will continue to drive the expansion of demand for large-capacity SSDs. 4. Capital and shareholder return planning Capital expenditures in 2026 will remain at a high level of 40 trillion won. Domestic new wafer and packaging production lines will be built, and there are no confirmed plans for overseas factory construction. Currently, cash flow is abundant, and the company is evaluating incremental shareholder feedback plans such as dividends and buybacks, and will disclose them to the public after they are implemented within the year. 5. Main reasons for performance falling short of expectations In the second quarter, the shipment pace of high value-added HBM was delayed and the long-term contract price lock dragged down the comprehensive ASP. This is a short-term structural disturbance. In the second half of the year, the product structure improved and the profit elasticity was restored. Complete transcript of SK hynix’s fiscal 2026 second quarter results conference call (AI-assisted translation, some content has been deleted) Park Seong-hwan (Head of Investor Relations): Good morning, everyone. Thank you all for participating in the SK hynix earnings conference call. Today's meeting started with a performance presentation by the company's management, and then entered a question and answer session. Let me introduce to you the management who attended: SK Hynix President Song Hyun-jong, Chief Financial Officer Kim Woo-hyun, DRAM Marketing Manager Park Joon-deok, NAND Marketing Manager Song Chang-seop, and HBM Sales and Marketing Manager Kim Ki-tae. Next, we officially entered the performance reporting session. First, the group president Song Xianzhong introduced the second quarter operating results, the company's future plans, and industry prospects, and then opened the management to collective Q&A. Ladies and gentlemen, my guest is President Song Xianzhong. Management performance and industry outlook speech Song Xianzhong (Group President): Hello everyone, I am Song Xianzhong. First, I will introduce to you SK hynix’s operating performance in the second quarter of 2026. Benefiting from the strong demand brought by the continued expansion of AI infrastructure and the continued tight supply, prices of DRAM and NAND products continue to rise, and AI-related products such as server DRAM and enterprise-level SSD have become the core driving force for growth.
In the second quarter, the company's total revenue reached 79.3 trillion won, a 51% month-on-month increase and a year-on-year increase of 257%, setting a new record high after the previous quarter. The supply capacity of the DRAM business is limited. The company focuses on HBM3E and AI server DRAM products. Bit shipments achieved high single-digit month-on-month growth, in line with the previous guidance; the sales of LPDDR products such as server SO-DIMM 2 increased significantly. Traditional DRAM prices continue to strengthen, driving the overall DRAM average price (ASP) to increase by approximately 30% month-on-month. The NAND business's shipment base in the first quarter was low. Coupled with the expansion of enterprise-level SSD sales, bit shipments achieved double-digit quarter-on-quarter growth in this quarter, fully in line with the guidance target. Enterprise-level SSD revenue doubled month-on-month, and its Solidigm brand's 30TB and above large-capacity enterprise-grade SSD revenue more than tripled month-on-month. The price increase of all products drove the average NAND price to increase by about 50% month-on-month. DRAM and NAND price increases across the board coupled with inventory cost optimization, operating profit in the second quarter reached 60.5 trillion won, a month-on-month increase of 61% and a year-on-year surge of 557%; the operating profit margin increased by 5 percentage points month-on-month to 76%, and both operating profit and profit margin set historical records. The total depreciation and amortization in the second quarter was 4 trillion won, the profit before interest, taxes, depreciation and amortization (EBITDA) was 64.6 trillion won, and the EBITDA profit margin was 81%. Non-operating net income was 62.2 trillion won, of which the upward exchange rate brought net exchange income of 1.1 trillion, and the total income from investment asset disposal and valuation appreciation was 63.3 trillion. Pre-tax profit was 122.7 trillion won, net profit was 93.9 trillion won, and net profit margin was 118%. As of the end of the second quarter, the company's cash and short-term investments totaled 88 trillion won, an increase of 33.6 trillion from the end of the previous quarter; interest-paying debt decreased by 0.7 trillion to 18.6 trillion, the net cash scale expanded to 69.4 trillion, and the asset-liability ratio dropped by 5 percentage points from the previous quarter to 7%. Although AI model iteration and software optimization will reduce the computing power consumption of a single task, efficiency improvements will not weaken the overall demand for infrastructure. Instead, they will lower the threshold for using AI services and expand the scale of users and applications. The number of AI service users of leading technology companies continues to grow and computing resources are in short supply. Therefore, they continue to invest in infrastructure and simultaneously increase storage purchases. Currently, leading customers have asked the company to increase supply requirements. PC and mobile storage sales have only experienced phased adjustments. As the supply gap eases and AI services become more popular, the terminal market will return to the growth track. Against the background of continued tightening of supply, DRAM and NAND bit demand are expected to achieve annual growth of around 20% and close to 20% respectively; if subsequent supply constraints are eased, the release of potential demand will further increase market growth. However, it is difficult to significantly improve the short-term industry supply and demand pattern. The core constraints are the increasing complexity of HBM and AI-specific memory advanced process manufacturing, and the long construction period of new wafer fabs. The tight balance between supply and demand will be maintained for a long time, and the company is negotiating multi-year long-term supply agreements (LTA) with many customers to stabilize mid- to long-term supply. At present, we have completed the signing of long-term agreements with about 10 core customers, and continue to promote negotiations with other leading customers in the industry. This type of long-term cooperation is not just a simple supply contract, but also a strategic partnership that can simultaneously match the next-generation chip technology routes of both parties and ensure stable mid- and long-term supply. The agreement pricing mechanism is designed based on customer and product differentiation to hedge the risk of market price fluctuations; performance guarantee clauses such as deposits are added to the contract to enhance the certainty of customers' mid- to long-term purchasing plans. Relying on long-term cooperation, the company will optimize investment and production and operation efficiency, and lay a solid foundation for medium- and long-term stable growth. The third quarter business plan is announced below: DRAM bit shipments in the third quarter increased by approximately 10% quarter-on-quarter, with priority being given to ensuring the delivery of server products; NAND bit shipments achieved low single-digit quarter-on-quarter growth. As the complexity of AI models continues to increase, and memory performance requirements increase simultaneously, industry competition is no longer limited to single chip design, but extends to system architecture and packaging technology. Relying on a complete product matrix of DRAM, NAND, and HBM, as well as joint R&D capabilities with customers, SK hynix will lead storage system-wide innovation:
1. HBM4: Continuous product optimization to meet customers’ computing power needs while achieving industry-leading power consumption and cost advantages. >Start batch shipments in the second quarter and fully ramp up production in the second half of the year; 2. HBM4E: Samples have been sent to leading customers in the first half of the year, using mature and stable optimal processes, and the pace of R&D and implementation is smooth; relying on stable mass production and high yields to bring cost advantages and industry-leading performance, we will continue to consolidate HBM’s leading position; 3. DRAM side: 1z nanometer process SOCAMM2 products will be fully supplied in the second quarter, and will subsequently optimize the product portfolio in line with customer R&D pace, expand customer groups, and promote sample delivery; 4. NAND side: Accelerate the iteration of advanced processes, focusing on large-capacity and high-performance products to meet market demand. Last quarter, 321-layer products accounted for the highest proportion of NAND production capacity. It plans to increase the proportion of 321-layer production capacity in domestic factories to 50% by the end of the year. In an environment where supply and demand continue to be imbalanced, stable delivery of sufficient production capacity has become a core competitiveness alongside technical strength. In order to respond to strong customer demand and seize medium- and long-term growth opportunities, the company continues to promote investment in production capacity expansion: the mass production time of the M15X factory will be advanced in the short term, and production will be rapidly expanded after the Longren Phase I clean plant is put into operation in early 2027. The early implementation of production capacity and increased investment will push capital expenditures in 2026 to a high range of 40 trillion won. Mid- to long-term production capacity investment is planned based on customer demand forecasts. Recently, the P&T7 advanced packaging expansion and M17 new NAND production line investment plans were announced. At the same time, the mid- to long-term plan for a new domestic semiconductor industrial park in addition to the Longren project was announced. Subsequent wafer fab construction, equipment procurement, and production capacity expansion will be implemented in stages based on customer demand visibility and investment efficiency. On the premise of strictly adhering to capital expenditure discipline, mid- to long-term production capacity growth will be planned in advance, and supply responsiveness and financial stability will be taken into consideration. Next, we will introduce the progress of the issuance of American Depository Receipts (ADRs): On July 10, the company's ADRs were successfully listed on Nasdaq, making it the largest ADR listing project for a foreign company in the history of the U.S. stock market. This listing is not only an expansion of financing channels, but also the global market’s recognition of the company’s technical strength and growth potential, while also broadening channels for cooperation with the next generation of computing power ecosystem. Relying on the opportunity of ADR's listing, the company will deepen strategic cooperation with leading customers and partners, explore new business opportunities, continue technological innovation, and help the semiconductor industry and AI industry grow together. Finally, the financial stability goal and shareholder return plan are explained: record high profits have brought abundant cash flow, and the company's financial strength has been greatly enhanced; structural growth opportunities in the AI era have simultaneously pushed up the demand for long-term investment funds. The company prioritizes investments in growth tracks with high returns and strategic value, while building a robust financial structure that can withstand market fluctuations, and continues to return operating income to shareholders. Although the scale of investment will expand in the future, abundant cash flow can simultaneously meet the triple goals of investment in production expansion, maintaining financial indicators, and increasing shareholder returns. We are currently evaluating various shareholder return implementation plans from multiple perspectives. Park Seong-hwan (Head of Investor Relations): The Q&A session has officially started. I have a question from Jay Kwon of JPMorgan Chase. Jay Kwon (JPMorgan analyst): Hi everyone, thank you management for taking the questions. Recently, there have been two major concerns in the market: many large technology companies are investigating data center leasing plans and efficient AI models continue to be implemented, and the market is worried that capital expenditures on AI infrastructure will slow down or even shrink. Based on the communication between the company and customers, how to predict the AI investment pace of leading cloud vendors (CSP)? What impact will this trend have on the demand for HBM, general DRAM, and NAND categories? Company Management: Thanks for asking. We are fully aware of market concerns, and the market interprets the rental of computer rooms and the implementation of high-efficiency large models by leading companies as signals for cooling down AI investment. However, the company has a different judgment on this: the above measures are not to reduce investment in AI, but to revitalize the existing large-scale computing infrastructure and accelerate the commercialization of AI business.
Song Xianzhong (Group President): We judge that the mid- to long-term investment in AI by leading cloud vendors will continue, and the storage demand data connected with customers also confirms this trend. Due to physical constraints such as power supply and computer room construction, the investment time of a single project will fluctuate slightly. However, against the background of industry competition among cloud vendors and the continued expansion of AI services, infrastructure investment will remain strong starting next year. The demand side will be fully expanded: in addition to HBM, the core of AI computing power, demand for server DRAM that supports the intelligent agent business and large-capacity, high-performance NAND that handles AI's massive data are simultaneously rising. Thanks for the question, and I’d like to invite the next person. Kim Rok-ho (Hanwha Securities Analyst): Management recently announced a mid- to long-term large-scale production capacity expansion plan. What is the basis for forecasting long-term storage demand to support the expansion? Does planned production capacity include long-term supply agreement lock-in orders? The market is worried that large-scale production expansion will cause excess supply in the future. How does the company view this risk? Song Xianzhong (Group President): Thank you for your question. The company's medium- and long-term production capacity strategy is based on the structural growth of AI-driven storage and the negotiation of medium- and long-term needs with top customers. Recently, the customer cooperation model has shifted from a single transaction to a long-term strategic binding, and major storage purchasers are actively seeking multi-year supply agreements, which is enough to prove the long-term sustainability of AI storage demand. This production capacity expansion is all based on the demand visibility locked in cooperation with customers. Equipment procurement and production line ramping are implemented in stages, and demand changes and investment efficiency are simultaneously considered. Capacity expansion strictly matches confirmed orders, and the probability of medium- and long-term investment plans directly causing oversupply is extremely low. Thanks for asking, next one. Kim Sun-woo (Merrill Lynch Analyst): Storage companies in the same industry have recently disclosed details of long-term supply agreements (LTA). This press conference only briefly mentioned it. Can you introduce in detail the core terms of SK Hynix's LTA contract cycle, pricing mechanism, etc.? Company Management: Thanks for asking. The company's LTA customizes differentiated terms based on customers and products. The contract base period is 5 years, and the specific conditions can be adjusted through negotiation between the two parties. Pricing will not adopt a unified template, and multiple price adjustment mechanisms will be designed to hedge spot price fluctuations. The core goal is to reduce the uncertainty of mid- to long-term operations for both parties. At the same time, the contract sets up performance guarantee clauses such as deposits to lock in the actual purchase volume of customers and improve the clarity of supply and demand expectations. We will not disclose LTA's overall revenue share to the outside world. We will maintain a reasonable agreement size based on the market environment and take into account both performance downside protection and market upside capture. Relying on HBM's leading advantages, we have reached long-term strategic cooperation with core AI customers such as Nvidia; in the.

#Stocks #Nvidia #AI #Semiconductors #Bonds