Global semi news — Korea, China, Taiwan, the US, and Japan. Government policy, export controls, capex moves, supply-chain shifts, and macro events. AI-classified and tagged with affected tickers. All headlines link back to the originating publisher.
Original: 中 '대중 반도체장비 수출통제 확대' 美법안에 "무역질서 훼손" - 연합뉴스
China publicly condemned a US legislative proposal to broaden export controls on semiconductor manufacturing equipment bound for China, calling it a violation of international trade norms. The escalating US-China tech rivalry raises the risk of tighter restrictions that could disrupt Chinese fab buildouts and reshape demand for Korean equipment, materials, and memory suppliers exposed to Chinese customers.
Why it matters: A potential expansion of US semiconductor equipment export controls to China is a direct, near-term policy risk for Korean memory makers with China fabs (SK Hynix, Samsung) and for equipment/materials suppliers reliant on Chinese demand.
Original: AI 비서 시대… '데이터센터 두뇌' CPU도 품귀
Surging AI assistant workloads are driving acute shortages of server CPUs alongside the existing GPU/HBM supply crunch, with hyperscalers reportedly unable to secure enough Xeon and EPYC parts for data center buildouts. The constraint extends the AI infrastructure bottleneck beyond accelerators into general-purpose compute, benefiting Intel and AMD while reinforcing tight conditions for memory suppliers serving the same systems.
Why it matters: Sector-wide AI infra supply-constraint theme extending from GPUs/HBM to server CPUs, directly relevant to Intel, AMD and memory suppliers but no single-name catalyst.
Original: SK하이닉스, SOCAMM2 양산 돌입…엔비디아 '베라 루빈' 메모리 협력 확대
SK Hynix has started mass production of SOCAMM2, a next-generation low-power memory module, and is expanding its memory supply collaboration with Nvidia for the upcoming 'Vera Rubin' AI platform. The move strengthens SK Hynix's position as a key memory partner in Nvidia's AI accelerator roadmap beyond HBM.
Why it matters: Direct new product mass production milestone tied to Nvidia's next-gen AI platform supply chain.
Open source articleOriginal: 데이터센터의 '조용한 반란'…주인공은 다시 CPU로
ZDNet Korea reports a shift in data center architecture where CPUs are reclaiming relevance alongside GPU-dominated AI workloads, driven by efficiency, inference cost, and general-purpose compute needs. The piece points to renewed momentum for x86 and Arm-based server CPUs as hyperscalers rebalance their compute mix.
Why it matters: Sector-wide theme on CPU resurgence in AI data centers affecting Intel, AMD, and Arm without a specific named event.
Open source articleOriginal: CPU, AI 에이전트 타고 부활…GPU·메모리 이어 품귀에 가격 급등
CPUs are making a comeback as AI agent workloads drive demand, leading to shortages and price hikes similar to GPUs and HBM memory. Server CPU suppliers Intel and AMD stand to benefit, while the broader tightness reinforces ongoing memory and AI infrastructure cycles.
Why it matters: Sector-wide pricing/demand theme covering CPU tightness alongside GPU and memory, without a single-company catalyst.
Original: SK하이닉스, 1c D램 기반 'SOCAMM2' 양산 개시…엔비디아 '베라 루빈' 최적화
SK Hynix has begun mass production of SOCAMM2 modules built on its 1c (6th-gen 10nm) DRAM process, optimized for Nvidia's next-generation Vera Rubin AI platform. The product targets low-power, high-bandwidth memory needs for AI servers, reinforcing SK Hynix's lead position in Nvidia's supply chain alongside HBM.
Why it matters: Confirmed mass production of a new Nvidia-optimized memory product directly impacts SK Hynix revenue mix and Nvidia AI roadmap supply chain.
Open source articleOriginal: HBM 다음은 'SOCAMM2'…SK하이닉스, '베라 루빈'용 모듈 양산 돌입
SK Hynix has begun mass production of SOCAMM2 memory modules destined for NVIDIA's next-gen Vera Rubin AI platform, positioning the product as the next major growth driver after HBM. The move extends SK Hynix's AI memory leadership beyond HBM into low-power compression-attached modules tailored for AI servers, with NVIDIA as the anchor customer.
Why it matters: SK Hynix entering mass production of a new AI memory module for NVIDIA's Vera Rubin is a direct, named product-launch event for a major semi name.
Open source articleOriginal: HBM 이어 '소캠2' 격전…SK하이닉스, 베라루빈용 본격 양산
SK Hynix is moving beyond HBM dominance to commence full mass production of SOCAMM2 modules destined for NVIDIA's next-gen Vera Rubin AI platform. The win extends Hynix's AI memory leadership into the low-power LPDDR-based module segment, setting up a fresh competitive front against Samsung and Micron.
Why it matters: SK Hynix beginning mass production of SOCAMM2 for NVIDIA's Vera Rubin is a concrete supply-chain win with direct revenue and competitive implications for major memory names.
Open source articleOriginal: 엔비디아 깐부 SK하이닉스, '베라 루빈'용 HBM4 메모리 양산 개시
SK hynix has reportedly started mass production of next-generation HBM4 memory for NVIDIA's upcoming Vera Rubin AI accelerator platform, cementing its position as NVIDIA's primary HBM partner. The move extends SK hynix's lead in the HBM market ahead of Samsung and Micron, with Vera Rubin succeeding the Blackwell generation in NVIDIA's roadmap.
Why it matters: Direct confirmation of SK hynix as NVIDIA's HBM4 supplier for the Vera Rubin platform is a major product/supply event for the HBM leader and NVIDIA's AI roadmap.
Open source articleOriginal: 더 강해진 미 대중 반도체 수출통제 법안…동맹국에 ‘수출통제 일치’ 요구 - 경향신문
A strengthened US legislative push on semiconductor export controls to China is pressing allies — Korea and Japan in particular — to harmonize their own restrictions. If enacted, Korean memory and equipment makers selling into China-based fabs would face tighter compliance burdens and potential revenue exposure, with Samsung's Xi'an NAND fab and SK Hynix's Wuxi DRAM/Dalian NAND operations most directly in the crosshairs.
Why it matters: Direct US export control legislation targeting China semis with explicit ally-alignment demands hits Samsung and SK Hynix China fab operations head-on.
Open source articleJul 14, 2026 close · day-over-day
MRVL
$218
-7.75%