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: 华为解决‘陶’法则最大弱点发热问题，3纳米级芯片即将发布 - 亚洲日报
Huawei/HiSilicon claims to have solved thermal-dissipation challenges for 3nm-class chips with imminent launch planned. If verified, this signals accelerated Chinese indigenous advanced-chip capability, directly threatening TSMC's foundry dominance and Samsung's competitive position in leading-edge processes.
Why it matters: Huawei's claimed 3nm chip breakthrough directly threatens TSMC's foundry leadership and Samsung's advanced-process competitiveness, but unverified claims require cautious interpretation before assessing as high-impact.
Original: DeepSeek 计划部署 16 万颗华为昇腾 950DT，建设超大规模国产推理算力集群
DeepSeek plans to deploy 160,000 Huawei Ascend 950DT chips to build a large-scale domestic AI inference cluster, representing direct substitution of Chinese semiconductor solutions for what would typically be Nvidia GPU demand. The deployment signals China's accelerating technological progress in homegrown AI chips and reduced dependence on US-origin semiconductors for critical AI infrastructure. This marks a significant milestone in China's domestic-substitution strategy for AI compute.
Why it matters: Huawei Ascend chip deployment at 160,000-unit scale by major Chinese hyperscaler DeepSeek exemplifies direct competitive threat to Nvidia's AI inference chip market, validating high-impact geopolitical competitive dynamic flagged as core coverage priority.
Original: 【电报解读】七部门发文探索算力设施800V高压直流供电架构，机构称800V直流架构有望带来直流保护、备电、电容、配电连接等多环节增量需求，这家公司的产品可匹配英伟达GB300及rubin架构
Seven Chinese government departments jointly promote 800V HVDC power architecture for AI computing infrastructure, signaling major capex deployment. The initiative targets Nvidia-compatible systems (GB300, Rubin), suggesting significant GPU demand from Chinese data center expansion. Domestic component suppliers are developing solutions for power protection, backup systems, and distribution networks.
Why it matters: China's coordinated 800V HVDC power standardization signals major AI infrastructure capex that would drive Nvidia GPU demand, though impact is indirect and depends on deployment scale.
Original: 【金牌纪要库】英伟达或使用OCS交换机，全光交换或带动产业链更多分支稀缺
Nvidia may adopt optical circuit switch (OCS) technology for data center infrastructure, which offers compact size and fast switching speeds. A Chinese OCS manufacturer is already ramping to volume production, and the shift will increase demand for optical fiber connectors, collimators, and polarization-maintaining fiber from Chinese suppliers gaining bargaining power.
Why it matters: Speculative report on Nvidia adopting optical circuit switches for AI infrastructure; highlights Chinese supplier opportunity but lacks specific company names and concrete impact on tracked Korean/Taiwanese component suppliers.
Original: 16万颗！DeepSeek被曝将采购大量华为芯片，国产AI算力要变天？ - Sohu
DeepSeek, a major Chinese AI company, is reported procuring 160,000 Huawei chips to build out domestic AI compute infrastructure. The purchase signals China's strategic shift away from US-restricted Nvidia GPUs toward domestic semiconductor alternatives under ongoing export controls. Nvidia's China-market access is directly threatened by this domestic-substitution acceleration.
Why it matters: Huawei chip procurement by DeepSeek demonstrates Chinese AI infrastructure's pivot from US-restricted Nvidia GPUs to domestic semiconductors, directly curtailing Nvidia's China market access amid export controls.
Original: 【财联社早知道】七部门探索算力设施800V高压直流供电架构，推动超百千瓦单机柜部署，分析师看好AIDC供电架构变革的机遇；内蒙古首笔“算力Token贷”发放
Seven government departments are advancing 800V high-voltage DC power infrastructure for AI data centers to enable ultra-high-power deployments, signaling China's commitment to scaling AI compute capacity. Inner Mongolia issued its first 'computing power token loan' product, enabling data center operators to monetize computing resources. Chinese companies are advancing CPO (coherent optical) chip testing capabilities, strengthening their position in the optical networking supply chain.
Why it matters: China's AI data center infrastructure buildout will increase semiconductor demand (memory, optical chips, power management), but the article lacks specific company names or quantifiable impact on tracked stocks.
Open source articleOriginal: 【狙击龙虎榜】市场冲高回落风险提前释放 美股算力硬件逆势走强 A股科技修复预期升温
Chinese market report highlights a semiconductor company's strategic positioning around structural supply gaps in optical module production as 1.6T mSAP process adoption accelerates for AI infrastructure—the firm secured partnerships with embodied AI leaders and aggressively expanded capacity this year. Concurrent diamond-substrate manufacturing signals advanced heat dissipation becomes critical as physical AI scales, pressuring global packaging supply chains.
Why it matters: Signals critical AI infrastructure demand and process technology adoption driving equipment capex, affecting US hyperscalers and semiconductor equipment vendors, though specific company identification is absent from the excerpt.
Open source articleOriginal: 【明日主题前瞻】AI玩具赛道成为资本追逐的新风口，市场规模未来有望突破千亿
Nvidia is launching RTX Spark-enabled Windows PCs next month, expanding AI GPU reach to consumer and SMB segments. Texas Instruments raised prices for the third time in 2026, reflecting strong demand and pricing power in analog chips. China's AI toys market is emerging as a capital investment target with projected 100+ billion yuan potential.
Why it matters: Nvidia and TI product/pricing data is directly material to tracked US holdings, while AI toys market is domestic China trend with no semiconductor supply-chain impact.
Open source articleOriginal: 华为“韬定律”细节再公开：麒麟2026密度暴涨55% 功耗反而大降
Huawei disclosed Kirin 2026 achieving 55% transistor density gain (to 238M/mm²) while cutting NPU power by 66% and GPU by 58%, demonstrating sustained progress in indigenous chip design. The advancement threatens competitive positions of TSMC/Samsung (foundry) and Nvidia (AI accelerators) amid China's push for semiconductor self-sufficiency.
Why it matters: Huawei's Kirin 2026 advancement signals sustained progress in China's indigenous SoC capability despite export controls, raising competitive threats for TSMC/Samsung foundry positions and Nvidia's AI accelerator dominance.
Open source articleOriginal: 订单排到年底、产线两班倒 芯片“太热”液冷迈入“必选”时代
Nvidia's confirmed Vera Rubin architecture mandates full liquid cooling, pushing penetration to 53% and creating supply constraints through year-end; Chinese quick-disconnect components have entered the ecosystem, but Chinese suppliers still lag in coolants and microchannel coolers.
Why it matters: Vera Rubin's confirmed liquid cooling drives AI infrastructure demand signal; Chinese supply chain entry noted but no specific companies named, limiting direct competitive impact assessment.