Samsung Electronics estimated operating profit for Q4 2025 at 20 trillion KRW, which is about 13.8 billion USD. Reuters reported this as a record quarterly operating profit and said it was nearly triple the year-earlier level, citing 208% YoY growth in operating profit. Revenue was estimated at 93 trillion KRW, which Reuters described as about 23% YoY growth.

The drivers point strongly toward memory and data center demand. Reuters also noted that sharply higher memory prices can raise costs for other product categories, including smartphones, as component inflation spreads through supply chains. Samsung is scheduled to publish full results with divisional detail on 2026-01-29, which will clarify how much of the upside is coming from memory versus other businesses.

DRAM Pricing Is Surging and TrendForce Is Already Projecting Another Jump

The shift is visible in market pricing. Reuters reported that contract prices for a certain DRAM category in Q4 2025 were up 313% YoY, and it said further gains were expected. Moves of that magnitude usually reflect a tight balance between supply and demand, especially when capacity is being pulled toward server and AI uses, while other segments compete for what remains.

TrendForce added concrete near-term projections for Q1 2026. It forecast that conventional DRAM contract prices could rise 55% to 60% QoQ. It projected NAND flash pricing increases of 33% to 38% QoQ. It also projected server DRAM prices rising by more than 60% QoQ and client SSD contract prices rising at least 40% QoQ. TrendForce described US hyperscalers as actively locking in supply, which can intensify pricing pressure elsewhere in the market.

HBM4 Is Slipping, Requirements Are Tightening, and Nvidia Is Pressuring the Supply Chain

The most sensitive part of the stack is high-bandwidth memory for AI accelerators. TrendForce stated that Nvidia revised the HBM4 specifications for the Rubin platform in Q3 2025, increasing the required per-pin speed to above 11 Gbps. TrendForce said this forced the three major HBM suppliers to modify their designs. It also stated that strong demand for Blackwell products prompted Nvidia to adjust Rubin’s mass-production timeline, pushing back the HBM4 ramp, with mass production expected to begin no earlier than the end of Q1 2026, and most likely between late Q1 2026 and early Q2 2026.

TrendForce wrote that Samsung has taken an early lead by adopting what it described as a 1c-nm process for HBM4 and by using advanced in-house foundry technology for the base die. Over the same period, TrendForce said Nvidia raised shipment targets for B300 and GB300 and increased orders for HBM3e, while revising the Rubin production schedule. In parallel, industry reporting around next-generation memory often frames node progress in terms such as SK hynix’s 1b-nm DRAM and Samsung’s 1c-nm DRAM, rather than a literal nanometer measurement.

Conclusion

Memory has moved to the center of the AI infrastructure story, and the recent numbers show why. Samsung estimated 20 trillion KRW in operating profit on 93 trillion KRW in revenue, while Reuters reported a 313% YoY jump in contract prices for the DRAM category in Q4 2025. TrendForce is projecting another sharp step up in Q1 2026, including 55% to 60% QoQ gains in conventional DRAM and more than 60% QoQ gains in server DRAM. At the high end, TrendForce described tighter HBM4 targets above 11 Gbps per-pin and a shift of mass production timing to late Q1 2026. Nvidia’s Rubin messaging also points to continued infrastructure buildouts, including a claim of up to a 10x reduction in inference token cost versus Blackwell in its official communications.