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.
NVIDIA and HPE are deepening their AI factory partnership with expanded infrastructure offerings, broadening enterprise access to NVIDIA-accelerated compute stacks. The collaboration reinforces NVIDIA's role as the de facto AI infrastructure platform and channels more enterprise AI capex through HPE's systems.
Why it matters: Sector-wide AI infrastructure partnership expansion that channels enterprise AI capex through NVIDIA-accelerated systems but contains no major new financial commitment.
Original: LG이노텍, AI 훈풍 타고 "5년 내 기판사업 영업익 1조원" 목표 제시
LG Innotek announced a goal to reach KRW 1 trillion in operating profit from its substrate business within five years, riding AI-driven demand for FC-BGA and advanced packaging substrates. The Korean component maker is positioning its substrate unit as a core growth driver alongside its camera module business as AI server and HPC chip demand accelerates.
Why it matters: Mid-term guidance from a key Korean substrate supplier signals sustained AI packaging demand but is forward-looking rather than an immediate earnings event.
Open source articleOriginal: LG이노텍 "고객사와 베트남 FC-BGA 라인 증설 논의"
LG Innotek confirmed a ~KRW 1tn first-phase investment in a Haiphong, Vietnam substrate plant (RF-SiP, FC-CSP) and disclosed it is in advanced talks with multiple customers — reportedly US x86 CPU and Taiwan AP/ASIC vendors — for a second-phase FC-BGA capacity build in Vietnam and Gumi. Network AI-server FC-BGA ships 2H26 with training/inference boards targeting 2027, pulled forward from prior 2027-28 guidance, signaling a new Korean entrant taking ABF-substrate share from incumbent Taiwanese suppliers.
Why it matters: LG Innotek itself is outside the tracked universe, but the capex confirms strong AI ABF-substrate demand while signaling new Korean competition that pressures Taiwanese FC-BGA incumbents (Unimicron, Nanya PCB, Kinsus).
Original: Intel starts production of advanced 18A-P chip node - Crypto Briefing
Intel has begun production of its 18A-P process, an enhanced variant of its 18A node aimed at improving performance and yield for foundry and internal products. The ramp is a key milestone in Intel's bid to win external foundry customers and narrow the process gap with TSMC and Samsung Foundry.
Why it matters: Intel's 18A-P production start is a company-specific milestone with direct competitive implications for TSMC and Samsung Foundry in the leading-edge node race.
Original: AI 주도권 연산에서 메모리로… HBM 2027년 '두 배' 뛴다 - 글로벌이코노믹
Industry forecasts indicate the HBM market will double in size by 2027 as AI infrastructure demand pivots from pure compute toward high-bandwidth memory. SK Hynix and Samsung, the dominant HBM suppliers, stand to benefit alongside Micron, with implications for memory pricing and capacity allocation through 2027.
Why it matters: Direct sector-wide demand forecast for HBM that materially affects the core Korean memory duopoly and Micron over the next 18 months.
Open source articleOriginal: 마이크론, 엔비디아 베라루빈 AI 플랫폼 HBM4 공급사로 선정
Micron has secured a supplier slot for HBM4 memory on Nvidia's next-generation Vera Rubin AI platform, joining the HBM4 race alongside SK Hynix and Samsung. The win validates Micron's HBM roadmap and expands its share of the AI memory pie, while intensifying competitive pressure on Korean incumbents who have historically dominated HBM supply to Nvidia.
Why it matters: Direct design-win event reshaping HBM4 supplier mix for Nvidia's flagship AI platform, with clear share implications for MU, SK Hynix and Samsung.
Open source articleOriginal: 마이크론, 엔비디아 베라 루빈 AI 플랫폼 HBM4 공급사 확보
Micron has reportedly secured a supplier role for HBM4 memory on Nvidia's next-gen Vera Rubin AI platform, validating its push into premium HBM. The win positions Micron alongside SK Hynix and Samsung in the HBM4 race and pressures Korean incumbents on share retention.
Why it matters: Direct HBM4 design-win for Micron on Nvidia's flagship AI platform reshapes the HBM competitive landscape for SK Hynix and Samsung.
Open source articleOriginal: 엔비디아, 신형 Vera CPU 중국 판매 추진…수출규제·현지 경쟁이 관건
Nvidia is reportedly eager to sell its new Vera CPU into the Chinese market, but faces hurdles from US export restrictions and rising domestic Chinese chip alternatives. The piece questions whether a downgraded or compliant Vera variant can compete against Huawei and other local CPU/accelerator efforts.
Why it matters: Sector-wide AI infra/geopolitics theme around Nvidia's China access strategy with no fresh hard data point.
Open source articleOriginal: 'CXMT 상장 임박' 중국 메모리는 위협일까 아닐까 [AI칩 인사이드] - v.daum.net
Chinese DRAM maker CXMT is reportedly close to a domestic IPO, reigniting debate over whether China's memory ramp poses a structural threat to Samsung and SK Hynix. While CXMT remains behind on HBM and advanced nodes, an IPO would arm it with capital to accelerate capacity and commodity DRAM pricing pressure.
Why it matters: CXMT IPO is a structural competitive signal for Korean memory makers but the timing and pricing impact remain speculative rather than near-term confirmed.
Open source articleOriginal: Intel begins production of most-advanced chip, inching closer to possible Apple deal - CNBC
Intel has started production of its most advanced 18A process node, a critical milestone as it courts Apple as a potential foundry customer. A win would validate IFS's leading-edge capability and intensify competition with TSMC's N2, while raising stakes for Samsung Foundry's catch-up effort.
Why it matters: Specific milestone (18A production start) plus named potential customer (Apple) directly reshapes the leading-edge foundry competitive landscape vs TSMC and Samsung.
Open source articleSK Hynix
000660
₩1,842,000
-11.53%