South Korea's June Memory Exports Surge 280% YoY; MLCC Adoption Accelerates Across AI Hardware Chain
Claire Weston
Korea's June memory exports surged 280% YoY, the fifth straight month above 200%, with DRAM growth hitting a record in Goldman Sachs' tracking history since 2008; simultaneous acceleration in MLCC and equipment trade signals AI hardware demand is spreading from the storage backbone into passive components and capex.
Five straight months of triple-digit memory export growth — what's actually driving it?
Korea's June memory exports rose 280% YoY. DRAM alone hit $21.8 billion, up 385% — the fastest growth Goldman has tracked since 2008.
NAND grew 301% and SSDs 355%. Enterprise SSDs, AI inference, and vector databases are sustaining NAND demand, lifting it beyond a consumer-electronics inventory-cycle afterthought.
This means → the surge is not just pricing. It is a product-mix upgrade: AI servers demand higher-bandwidth, higher-capacity memory, pushing up the share of premium-priced shipments — gains that flow straight into storage leaders' income statements.
Can Goldman's Samsung forecasts be read off customs data?
Goldman projects Samsung Electronics' 2Q26 memory revenue up 452% YoY — DRAM up 502%, NAND up 355%.
In plain terms = customs-level export growth and sell-side earnings forecasts are pointing the same way — the export numbers are strong enough to enter analysts' revision range directly.
One caveat: Goldman notes June had roughly 1.5 more working days than last year. A low base plus price increases amplify the YoY reading; July exports and DRAM/NAND contract prices need continued tracking.
Is there physical evidence that HBM is ramping?
Goldman uses plastic-film imports from Japan to Samsung's Hwaseong and Pyeongtaek fabs — a proxy for TC-NCF material, a key consumable in HBM packaging — as a leading indicator. May imports rose 76% YoY; the January-to-May run is up 61%.
This reflects a signal rooted in material flow, not management commentary: Samsung's HBM production prep and shipment cadence are advancing in substance.
But the boundary matters: film imports ≠ HBM revenue, and they say nothing about customer qualification outcomes or SK Hynix's share shift. Samsung's valuation discount narrows only if material indicators keep rising and resonate with HBM orders and yield data in Samsung's earnings.
What does MLCC acceleration tell us?
Korea's June MLCC — multilayer ceramic capacitors, the small components on circuit boards that stabilize power delivery and filter noise — exports rose 31% YoY and 16% MoM, a clear step-up from May's 14%.
This means → AI hardware demand has punched through from the storage backbone into passive components. As GPUs, HBM, and high-end CPUs raise power draw and signal complexity, demand for stable power delivery and high-reliability parts rises in step — the breadth of the diffusion is widening.
Goldman expects Samsung Electro-Mechanics' 2Q26 MLCC revenue up 22% YoY, driven by AI-server and automotive MLCC. Once both high-end demand pools absorb capacity simultaneously, room opens for mix improvement and upward pricing.
Equipment imports and exports both surging — is capex real or still on paper?
Korea's June semiconductor equipment (WFE) exports rose 79% YoY; imports rose 54%. For 2Q26, exports and imports were up 69% and 73%, respectively.
In plain terms = hyperscalers' capex can stay on a slide deck, but equipment imports require purchase orders, delivery, and real budget commitment. Sustained import strength shows memory fabs' conviction in demand durability is high enough to back expansion decisions with capital.
The risk sits here too: capex is a double-edged sword. Near-term, it validates tight supply and order visibility. Medium-term, it may unleash supply post-2027. If expansion flows mainly toward HBM and advanced-packaging bottlenecks, the cycle stays healthier; if it reverts to commodity DRAM, future oversupply risk rises.
Content is for reference only, not financial advice.