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: 華邦電漲停飆天價！00403A神級卡位 抱63億元市值成最大贏家
Winbond (2344-TW) locked limit-up at a record NT$184.50 on June 2 amid AI-driven DRAM/NAND price hikes, Samsung orders, and a Micron-led memory rally. Among nine active ETFs holding the stock, Uni-President's 00403A dominates with an NT$6.3B (~US$210M) position built since May 13, far ahead of the No. 2 holder at NT$700M. Analysts see the memory shortage extending from DRAM into NOR and NAND Flash as a multi-year supercycle.
Why it matters: Sector-level memory upcycle story centered on a Taiwan niche DRAM/NOR maker not in the tracked universe; the ETF positioning angle is informational rather than a discrete catalyst for tracked names.
Original: 〈COMPUTEX〉童子賢：AI產業仍處於大西部開拓時代 和碩千人團隊往上衝刺
Pegatron Chairman T.H. Tung said the company has expanded its server team from a few hundred to 1,000 people as it pushes into AI server business, acknowledging a late start versus Quanta (2382), Inventec (2356) and Wistron (3231) who built data center expertise over a decade ago. At Computex 2026, Pegatron showcased NVIDIA's Vera Rubin NVL72 rack-scale AI supercomputer, HGX Rubin NVL8, RTX PRO Servers, and DSX AI Factory reference design, signaling a serious catch-up bid in the AI server ODM race.
Why it matters: Sector/competitive-positioning story on Taiwan AI server ODMs with no specific contract, revenue or capex figures disclosed.
Open source articleOriginal: 〈COMPUTEX〉AI液冷助攻！大井泵浦科技水泵出貨放量 法人看好全年營收創高
Walrus Pump (6982-TW) said its high-margin 'tech pump' line for AI server liquid cooling jumped to 7.3% of revenue through April from ~4% in 2025, with the 10% threshold likely crossed a year ahead of plan; gross margins on these products exceed 40%. Management flagged rising co-development activity with strategic shareholder Delta Electronics (2308) and AI thermal customers, and analysts model FY26 revenue reaching a record NT$1.9B (~US$60M).
Why it matters: Primary subject 6982 is outside the tracked universe; the only in-universe name is Delta (2308), mentioned tangentially as strategic shareholder/AI thermal partner — supply-chain color rather than a stock-moving event for 2308.
Original: AI時代誰最賺？三星超車輝達奪全球獲利王！SK海力士衝前三、台積電排名曝
Bloomberg's 2027 global operating profit forecast puts Samsung first at $331B (edging Nvidia's $329B), SK Hynix third at $243B, and TSMC ninth at $109B, reflecting an AI profit pool shifting toward HBM memory suppliers. Samsung is also reportedly the exclusive display supplier for Apple's first foldable iPhone over a three-year window, adding a second earnings engine alongside the HBM super-cycle.
Why it matters: Sector-level profit forecast and roadmap narrative reinforcing the HBM bull case for Samsung and SK Hynix rather than a discrete stock-moving catalyst.
Open source articleOriginal: 台積電成最大贏家！AI封裝重構半導體格局、大摩點名CoWoS、CPO、WoW三大技術
Morgan Stanley projects the global AI semiconductor market will reach $753B by 2030, with advanced packaging now the key determinant of AI chip performance and cost. TSMC's CoWoS monthly capacity is set to expand from 23K wafers in 2023 to 165K by 2027, with Nvidia consuming ~60% (875K wafers in 2026), Broadcom ~20%, and AMD ~8%; AI-related revenue CAGR is forecast at 60% through 2029. ASE (3711) and Amkor are stepping in for mid/low-end CoWoS overflow, while CPO and WoW markets are projected to grow at 172% and 257% CAGRs respectively.
Why it matters: Brokerage roadmap/market-data report with specific capacity and share numbers for TSMC and ASE, but no new contract, capex commitment, or earnings catalyst.
Open source articleOriginal: [컴퓨텍스 2026] 타이베이를 물들인 '베라 루빈'…엔비디아 생태계 총출동
Computex 2026 in Taipei is saturated with Nvidia's next-gen Vera Rubin platform, with the broader Taiwanese supply chain ecosystem aligning around the new architecture. Server, networking and ODM partners are showcasing Rubin-ready designs, reinforcing Nvidia's central role in AI infrastructure buildout.
Why it matters: Sector-wide AI infrastructure theme reaffirming Nvidia/Taiwan ecosystem alignment around Vera Rubin, but no single hard catalyst.
Original: [GTC 타이베이 2026] 엔비디아 베라 루빈 시대 개막…젠슨 황 "본격 양산 돌입"
At GTC Taipei 2026, Jensen Huang declared Nvidia's next-gen Vera Rubin platform has entered full-scale mass production, signaling a generational shift from Blackwell. The announcement reinforces TSMC's CoWoS ramp and HBM4 demand for Korean memory suppliers, with Taiwan ODM/networking partners positioned as primary beneficiaries.
Why it matters: Nvidia confirming Vera Rubin volume production is a direct, major product-launch event that drives capex and order flow across TSMC, HBM suppliers, and AI server ODMs.
Original: [컴퓨텍스 2026] 전시장 휩쓴 '베라 루빈'…엔비디아 생태계로 물든 타이베이
Computex 2026 in Taipei is saturated with Nvidia's next-gen Vera Rubin platform, with the broader Taiwanese supply chain showcasing products built around Nvidia's ecosystem. Server ODMs, networking partners, and component suppliers are aligning roadmaps to Rubin, reinforcing Nvidia's gravitational pull on Taiwan's AI hardware stack.
Why it matters: Computex showcases confirm Taiwan supply chain alignment with Nvidia's Rubin roadmap, a sector-wide AI infra theme but no single hard catalyst.
Original: [컴퓨텍스 2026] 전시장 곳곳 '베라 루빈'…엔비디아 생태계로 물든 타이베이
At Computex 2026, Nvidia's next-gen Vera Rubin platform took center stage across the Taipei show floor, with Taiwanese ODMs, foundries, and component suppliers showcasing Rubin-based servers, networking, and cooling solutions. The event underscores Nvidia's tightening grip on the AI infrastructure value chain and Taiwan's pivotal role as the primary manufacturing hub for the next AI capex cycle.
Why it matters: Industry-wide AI infrastructure showcase highlighting Nvidia's Rubin platform and Taiwan's ecosystem positioning, without a single discrete earnings or product launch event.
Original: 〈COMPUTEX〉晟銘電羅志吉：「生意還不錯」 未來2至3年每年資本支出年增20至30%
Sunteng Electronics (3013-TW) GM Lo Chih-chi said business is solid at Computex 2026, guiding annual capex increases of 20-30% over the next 2-3 years to expand capacity in Thailand and potentially the US. The company ships AI racks with GB200 (1 IT rack + 2 liquid-cooling racks) and GB300 (2-3 liquid-cooling racks) configurations, and expects Nvidia's Vera Rubin platform launching Q4 2026 with volume ramp in 2027 to materially boost revenue.
Why it matters: Capex guide and AI rack volume commentary are meaningful for the liquid-cooling/server-rack supply chain, but Sunteng (3013) is outside our tracked universe and the read-across to tracked names is indirect.
Open source articleRealtek Semiconductor
2379
NT$762
-9.29%