South Korean memory manufacturer SK Hynix released its financial report for the third quarter of fiscal year 2025 as of September 30, 2025. The company's revenue in the third quarter of fiscal year 2025 was KRW 24.4489 trillion, an increase of 39% compared with the same period in 2024. Operating profit was KRW 11.3834 trillion, down 62% from the same period in 2024. Net profit was 12.5975 trillion won, an increase of 119% compared with the same period in 2024. Operating profit margin in the third quarter of fiscal 2025 was 47%, and net profit margin was 52%. Overall, the performance was in line with market expectations.
SK Hynix stated that as the prices of DRAM and NAND Flash memory increased across the board, shipments of high-performance products for AI servers increased, and the company achieved its highest performance in a single quarter. In particular, operating profit exceeded KRW 10 trillion for the first time since the company was founded. SK Hynix pointed out that customers have expanded investment in AI infrastructure and overall memory demand has increased sharply. Sales of high value-added product groups such as 12-layer HBM3E and DDR5 for servers increased, thus once again setting a new record for a single quarter set in the previous quarter.
The demand for AI servers has grown, and the shipments of high-capacity DDR5 higher than 128GB have doubled compared with the previous quarter. In the field of NAND Flash memory, enterprise-class solid-state drives (eSSD, enterprise SSD) for AI servers have seen a premium, and their shipment proportion has increased. Therefore, the company predicts that as the AI market rapidly shifts to be inference-centric, in order to disperse the computing burden of AI servers, the market is shifting the computing load to multiple infrastructures such as general-purpose servers, which will in turn drive the continued expansion of overall memory demand such as high-performance DDR5 and eSSD.
In addition, major companies in the AI field have recently signed strategic partners and announced AI data center expansion plans, which is also a positive sign. This is expected to drive balanced demand growth for various products including HBM and general-purpose memory for servers. In this regard, SK Hynix plans to accelerate the conversion to the most advanced sixth-generation 10nm-level (1c) process that has achieved stable mass production. This will provide a full-line up DRAM product lineup for servers, mobile devices, graphics processing, etc., and expand product supply to respond to customer needs. In the NAND Flash business, the company plans to expand the supply of 321-layer TLC and QLC products based on the world's highest stacking layer count and quickly respond to customer requirements.
Currently SK hynix has reached an agreement with major customers on the supply of HBM in 2026. HBM4, which was developed and built into a mass production system in September, is ready to meet all customer performance requirements and achieve the highest speed in the industry. The company plans to start supplying the product in the fourth quarter and will officially expand sales in 2026. At the same time, due to the sharp increase in memory demand for AI, customer demand for all DRAM and NAND Flash products has been ensured until 2026.
SK hynix emphasized that in order to cope with the unexpected customer demand, it has recently opened the M15X clean room early and started to introduce production equipment. It will ensure new production capacity as soon as possible through M15X and accelerate the conversion of advanced processes. Therefore, the company's investment scale plan for 2026 has increased compared with 2025, and it will maintain an optimal investment strategy in line with market conditions. Kim Woo-hyun, vice president of finance (CFO) of SK hynix, said that with the innovation of AI technology, the memory market has transformed into a new model, and demand has begun to spread to all product areas. In the future, we will respond to customer needs based on market-leading products and differentiated technological competitiveness, and consolidate our leading position in the memory field for AI.