On October 29, Samsung Electronics released its 2026 third-quarter results and held an earnings call: powered by strong AI chip demand, both Q3 revenue and profit hit record highs, with operating profit up nearly eightfold year over year and topping the KRW 100 trillion mark for the first time — over 99% of it contributed by the semiconductor business. On the call, memory business head Jaejune Kim said that as competition in AI infrastructure buildouts intensifies and personal AI agents spread, memory demand is expected to keep growing substantially, while new capacity cannot be converted into actual supply in the short term. Synthesizing the views of multiple market research firms, he indicated that memory supply tightness is expected to persist through 2028.
The supply-demand imbalance has already hit end devices: Samsung has notified its supply chain to cut fourth-quarter 2026 production of mid-range Galaxy A phones by about 20% versus the original plan early this year — roughly 10 million fewer units for the full year. On some mid-range and entry-level models, memory now accounts for more than 50% of production cost; from roughly one-tenth of a phone’s bill of materials in the past, memory has climbed to three or four tenths or more, completely rewriting the price structure of low-end devices. Samsung’s response is to shift resources toward more profitable flagship models and AI-server memory. By contrast, the company’s Device eXperience (DX) division recorded a loss this quarter, deepening the profit divergence between its semiconductor and consumer businesses.
From an industry perspective, Samsung’s ‘semiconductors win alone, devices strain’ divergence reveals how AI dividends are being redistributed across the supply chain: upstream memory makers capture the vast majority of incremental profit, while midstream and downstream system vendors face structural BOM inflation and must absorb it through reduced configurations, production cuts, or price hikes. The ‘tightness through 2028’ judgment means this is not a short-term disturbance but a capacity cycle of at least two years — HBM crowding out conventional DRAM capacity and AI servers siphoning enterprise SSDs will keep reshaping how capacity is allocated across memory categories. For the hardware supply chain, the ability to ‘lock in volume and secure supply’ is becoming one of the most critical competencies for system makers over the next two years.
For Junyi Microelectronics, the ‘tightness through 2028’ outlook directly shapes the material strategy of its system business: the DRAM/eMMC memory required for the J-3568/J-3576/J-3588 core boards and K-series edge-AI motherboards calls for longer-horizon price locking and stocking arrangements to hedge against continued price increases. The 5G-EDGE-3399Pro and AI-BOX-3588 edge-computing hosts serve industrial edge-AI scenarios where large local storage is a hard requirement for on-device model deployment; through standardized core-board design, Junyi keeps storage configuration elastic so the same platform can adapt to different capacity tiers on demand. The IPC-i5-1235U industrial mini-PC and touch-panel all-in-one machines adopt tiered storage configurations — differentiated by application scenario — to balance cost and experience. Every swing of the upstream cycle tests the depth of a system maker’s supply-chain management.
Source: https://finance.eastmoney.com/a/202610303916424534.html (2026-10-30)