Samsung Electronics and SK hynix: Recovery or Sideways Trading? Securities Analysts Weigh In

[Financial News] A forecast has emerged that the memory semiconductor market will enter a strong upcycle from next year, with both prices and shipments rising at the same time. 
Kim Dong-won, a researcher at KB Securities, said in a report on the 2nd, "According to market research firm TrendForce, commodity DRAM prices in the fourth quarter of this year are expected to rise by an additional 10-15% from the previous quarter, while NAND flash prices are projected to rise by an additional 15-20%, continuing a double-digit uptrend that exceeds market expectations."
He also predicted, "Demand will remain strong, particularly for server DRAM and enterprise solid-state drives (SSDs)." Driven by the rapid expansion of AI inference demand, SSD bit demand is expected to increase by more than 80% year on year.
KB Securities noted that HBM prices will enter a full-fledged uptrend next year. Kim said, "Micron Technology has said it is significantly raising HBM supply prices next year," adding, "HBM, which is typically sold under one-year long-term fixed-price contracts, has so far failed to fully reflect the steep rise in commodity DRAM prices. However, beginning with contracts for next year, higher memory prices and tight supply-demand conditions are beginning to be fully reflected in HBM prices."
He further argued, "If HBM drove growth last year and commodity memory is leading it this year, next year is expected to mark a broad-based growth phase in which price increases and shipment growth for both HBM and commodity memory unfold simultaneously."
KB Securities also forecast that memory supply-demand conditions next year and in 2028 will be tighter than this year. Kim predicted, "Recently, frontier AI models such as OpenAI's GPT-6 Astra and Google's Gemini 4 Argon have continued to expand the upper limits of AI performance, centered on agentic AI and artificial general intelligence (AGI). At the same time, China's high-efficiency AI models are expected to rapidly expand the number of new users and usage frequency by lowering barriers to entry for AI services through low inference costs."
He also anticipated, "As the high performance of frontier models and the popularization of efficiency-focused models proceed simultaneously, falling token prices and rising usage will combine to produce a structural surge in overall AI computing and token consumption."
Kim added, "Sanjay Mehrotra, CEO of Micron Technology, forecast during an earnings conference call that memory supply-demand conditions next year and in 2028 will be considerably tighter than this year," saying, "This is because building new cleanrooms and expanding fabs takes more than three years, while the expansion of HBM-focused production capacity is also proceeding in parallel. As a result, the increase in supply available to meet rapidly rising memory demand is expected to remain extremely limited." Finally, he added, "Therefore, next year's memory market is expected to enter an acceleration phase of a powerful upcycle, with prices and shipments rising simultaneously," and recommended "Samsung Electronics and SK hynix as our top picks."

[email protected] Han Young-jun Reporter