Strong memory profits have materialized; how long they last depends on product differences and capacity coming online

Profits have materialized, but their sources still need to be broken down

Samsung’s preliminary consolidated guidance for the third quarter puts revenue at approximately KRW 195 trillion and operating profit at approximately KRW 107.40 trillion, both midpoints of the estimated ranges. This supports an assessment of the scale of group earnings, but the announcement did not disclose quarterly results for DS, Memory or HBM, so it cannot establish the contribution of each memory product category. The DS division’s second-quarter operating profit of KRW 89.2 trillion covers a different reporting period. DS also covers the semiconductor business, so that figure cannot serve as evidence of third-quarter HBM profits. The next segment results need to show which businesses generated the profits and whether that mix can persist.

Evidence: [1] [2]

Price expectations remain strong, while consumer customers are already adjusting

On September 30, 2026, TrendForce forecast that fourth-quarter contract prices would rise 10% to 15% quarter over quarter for conventional DRAM and 15% to 20% for NAND. These are forecasts and must not be presented as realized increases. Its research also observed that supply for servers was constrained by manufacturing processes, components, and packaging and testing, with some price increases bounded by ranges in long-term contracts. PC customers, meanwhile, were cutting SSD capacities and purchase volumes. This suggests that upstream suppliers still have scope to raise prices, but customers are already responding to costs by lowering specifications and buying less. Even if industry-average quoted prices continue to rise, that does not establish that demand is strengthening across all products, much less that individual manufacturers’ realized selling prices and profits are improving by the same percentages.

Evidence: [5]

Enterprise SSD demand has concrete support, with limits set by workloads

Micron disclosed that the 7600 and 9650 had shipped to leading customers for KV caching, while the 6600 ION ramped up in the fourth fiscal quarter. This provides specific product evidence for enterprise SSD demand. According to Micron’s engineering explanation, HBM serves the most active generation tasks, main memory handles another layer of workloads, NVMe SSDs retain context across multi-turn workflows, and high-capacity SSDs support data lakes, retrieval-augmented generation and long-term memory. My assessment is that SSDs are gaining new uses within tiered storage, with their value depending on access latency and capacity requirements. This explanation comes from the manufacturer. It does not yet establish that different workloads achieve the same benefits, nor does it imply that SSDs broadly replace HBM.

Evidence: [4] [6]

Long fab construction timelines do not mean supply is frozen

SK hynix announced a capacity expansion plan of approximately KRW 54 trillion. The first cleanrooms at Y2 and M17 are scheduled to open in June 2029 and December 2028, respectively, with the projects focusing on DRAM/HBM and NAND. Cleanroom expansion and equipment installation will proceed in phases according to demand, and opening dates do not mean the entire fabs will be running at full capacity. This supports a limited judgment: these projects cannot deliver their full capacity quickly. But the announcement also mentions expansion plans at other fabs, so the schedules for these two projects cannot define the limits of new supply across the industry. Assessing how long supply constraints can last still requires consideration of other projects, equipment installation and actual output.

Evidence: [3]

Sustained high profits need to be demonstrated by subsequent deliveries

Micron has reported substantial revenue and operating cash flow and provided revenue guidance for the next fiscal quarter. TrendForce, meanwhile, points to both support from enterprise demand and purchasing pressure in consumer markets. The WeChat publication argues that the high-profit cycle can continue. The available evidence supports that possibility, but does not establish its duration. My assessment is that persistence depends on whether enterprise demand translates into actual deliveries, whether price increases can be realized within the constraints of long-term contracts, and how new supply and customer purchasing evolve. If consumer customers keep cutting capacities and orders, enterprise growth will need to provide more support. Revenue guidance is an expectation and cannot be treated in advance as a realized result. High profits alone also do not establish that future prices and profits will necessarily hold up.

Evidence: [4] [5] [7]

What to watch next

• Whether Samsung’s subsequent segment results confirm the memory business’s profit contribution and product mix.
• Whether actual contract prices deliver the increases TrendForce forecast, and whether consumer customers continue cutting capacities and purchase volumes.
• Whether Micron’s subsequent realized revenue and enterprise SSD deliveries support its revenue guidance and demand assessment.
• Whether equipment installation, production starts and actual output at SK hynix and other expansion projects change supply constraints.