AUO Chairman and CEO Paul Peng stated that the company is tailoring its glass core substrate specifications specifically to semiconductor process requirements rather than competing with Innolux on panel dimensions. AUO aims to leverage its glass-processing expertise to supply critical components and technologies to outsourced semiconductor assembly and test (OSAT) leaders such as ASE Technology Holding and Powertech Technology. This reflects the company's stated plan; the source provides no specific capacity figures, investment amounts, or timelines.
SEMICON West 2026 wrapped up in San Francisco (the source states only 'last week', with no specific date given). At the event, US officials, industry executives, equipment and materials suppliers, and workforce-development partners repeatedly highlighted Taiwan as a critical partner in rebuilding US semiconductor manufacturing. Taiwan's role spans investment, process technology, supply-chain expertise, and talent development. No specific prices, capacity figures, or transactions are mentioned.
South Korean substrate maker Daeduck Electronics is accelerating development of ceramic core substrates (CCS) to address warpage and thermal-management challenges as AI semiconductor packages grow larger. The company is currently conducting performance and internal reliability assessments, plans to advance samples and customer evaluations in 2027 and 2028, and is targeting mass production in 2029; ceramic cores are positioned as a material alternative to glass cores for next-generation large-area packages. These remain plans/targets, with mass production not yet achieved.
Hygon Information Technology reported a 57.58% increase in revenue for the first nine months of 2026, reflecting sustained demand for domestic high-performance processors and AI computing chips. However, slower quarterly growth, mounting R&D expenditure and a steep share-price correction (a 51% slump) have highlighted the gap between the company's operating performance and market expectations. The source does not state the cause of the share-price decline.
domestic high-performance processorsAI computing chipsHygon Information TechnologyDigiTimes Asia
Gaonchips said on Oct. 19 that it will participate in the development of an artificial intelligence (AI) semiconductor for next-generation unmanned systems operated by Korea Aerospace Industries (KAI). Gaonchips will serve as a joint research and development partner in an aerial unmanned platform project overseen by KAI under the K-On-Device AI Semiconductor Technology Development Project, supported by the Ministry of Trade, Industry and Resources (MOTIR), with Supergate as the lead developer. The project aims to develop an AI system-on-chip (SoC) to serve as the brain of next-generation expendable unmanned aerial vehicles, enabling drones to run AI and carry out missions independently through on-device processing even where communications are disrupted or limited. Gaonchips will implement a domestically developed neural processing unit (NPU) and radar signal processing technology using Samsung Electronics' 8nm foundry process, and will handle physical design and final design verification of the entire chip; KAI will be responsible for system integration and verification of the unmanned platform, and Supergate will lead AI chip architecture design and software development. Gaonchips has experience developing and mass-producing 5nm on-device AI semiconductors and has worked on a 2nm GAA process and 2.5D packaging AI accelerator design project for a Japanese customer. The announcement describes a development project; no production or commercialization results are reported.
AI SoC (AI system-on-chip)neural processing unit (NPU)Gaonchips, Korea Aerospace Industries (KAI) and 4 totalThe ElecPartnership/M&A
Related manufacturers
Gaonchips, Korea Aerospace Industries (KAI), Supergate, Samsung Electronics
Samsung Electro-Mechanics said in a regulatory filing on Oct. 19 that it signed a 360.5 billion won contract to supply package substrates for memory semiconductors to a major global company, with the contract running from Jan. 1 to Dec. 31, 2027 and representing 3.2% of the company's consolidated revenue last year. This is the company's first disclosed long-term agreement (LTA) for package substrates, bringing its total disclosed LTAs this year to seven; contracts previously concentrated in the components business involving silicon capacitors (Si-Caps), MLCCs and inductors have now expanded to the Package Solution division, with cumulative contract value of approximately 4.3 trillion won. According to the source, growing demand for high-value products such as GPU and CPU substrates for AI servers and heterogeneous integration chiplets is believed to have led to the LTA. Separately, the company last month disclosed plans to invest 4.27 trillion won in its FC-BGA production facilities in Sejong and 2.51 trillion won in package substrate manufacturing equipment in Vietnam; including building expansion in Vietnam, total spending is expected to exceed 8 trillion won. The company previously announced plans to invest 8 trillion won at its Sejong facility and 15 trillion won at its Busan facility by 2040 as part of a government-led megaproject.
DeepSeek is reportedly working closely with Huawei to influence the design of Huawei's future AI chips, a link reportedly closer than typical supplier-customer relationships. The collaboration shows the growing ties between two of China's largest technology firms to sidestep Nvidia. This is a reported collaboration whose status is unconfirmed; the source provides no specific product models, timeline, or volume figures.
China's electronic design automation (EDA) industry is consolidating: Shanghai-based S2C and Wuxi-based HyperSilicon are merging into a unified digital front-end EDA platform. The deal is backed by Empyrean Technology and other investors, and its stated aim is to strengthen domestic capabilities in hardware emulation, chip verification, and simulation. The consolidation targets the hardware emulation gap in domestic EDA.
hardware emulationchip verificationsimulationS2C, HyperSilicon and 3 totalDigiTimes AsiaPartnership/M&A
TPK Holding has completed a pilot production line for glass-substrate through-glass via (TGV) technology at its Zhongli plant and started joint development and validation of early-stage samples with semiconductor packaging companies. Under its stated plan, the company targets core technology validation by the end of 2027 and mass production in 2028; the title cites an investment of NT$6 billion. This is a planned/expansion event, and mass production has not yet been realized.
Cumulative shipments of chips incorporating Alibaba DAMO Academy's RISC-V processor IP, branded XuanTie, have surpassed 6 billion, as the company expands beyond embedded applications into AI computing, servers, autonomous driving, and robotics.
SK Hynix, South Korea's industry ministry, the state-run Korea Trade Insurance Corp. (K-SURE), and four major commercial banks launched a KRW2 trillion (US$1.5 billion) loan guarantee program for semiconductor materials, parts, and equipment (MPE) suppliers on October 19, 2026. The program is the largest single-sector package under the government's push to strengthen export supply chains.
semiconductor materials, parts, and equipment (MPE)SK HynixDigiTimes AsiaPolicy/Regulation
According to commentary, Samsung Electronics' 2nm mass-production capability at its Taylor, Texas fab has become critical to its efforts to close the gap with TSMC, while Tesla's AI5 chip delivery schedule and Broadcom's potential 2nm orders hang in the balance. The company's ability to achieve competitive yields and stable output at the US facility will determine whether it can turn major customer orders into a broader foundry comeback. The source does not state the cause, and these statements are commentary rather than realized yield or order outcomes.
2nm processSamsung Electronics, TSMC and 4 totalDigiTimes AsiaCapacity Change
Hanmi Semiconductor has established a US subsidiary in Silicon Valley. According to the source, the purpose is to capitalize on growing demand for advanced semiconductor packaging equipment driven by the expansion of AI chip production in the US. The source provides no specific date, investment amount, or capacity details.
advanced semiconductor packaging equipmentHanmi SemiconductorDigiTimes Asia
China is proposing to extend AI technology export controls to chip design, memory technologies, and large-scale computing clusters, introducing specific technical thresholds for each. The source does not state the cause. The measure remains at the proposal stage and has not taken effect; no implementation timeline or approval status is mentioned.
According to an analysis, China's proposed AI export controls would target top-ranked foundation models and advanced agents. The proposed regulations cover foundation models, AI agents, chip design, and large-scale computing clusters. The source does not state the cause. The controls remain proposed and are not yet in effect.
foundation modelsAI agentschip designDigiTimes AsiaPolicy/Regulation
Market reports suggest that MediaTek's tensor processing unit (TPU) v9 project with Google could face delays in volume production, reportedly because Google is revising the chip's design; the title also cites glass substrate yield issues. According to the title, 2026 orders reportedly remain strong. This is an unconfirmed market report/rumor, and the volume production delay has not been verified. Industry observers expect MediaTek may address some of these concerns during its earnings conference in late October.
Citing Unimicron Chairman S.C. Chien at the TPCA Show on October 20, Wealth Magazine and Economic Daily News report that AI accelerator demand driving advanced packaging such as CoWoS and EMIB is spreading supply bottlenecks to critical materials and equipment: lead times for some key equipment have stretched to 30 months, and critical materials could face a 20% supply shortfall in 2027. On the materials side, low-CTE glass fiber cloth, copper-clad laminates (CCL) and ABF are particularly tight; lead times have lengthened for more than 80% of equipment, with over 20% seeing delays of more than double their original lead times. Chien projected CoWoS capacity to grow 6.8-fold by 2028 from 2024 levels, alongside a 6.3-fold increase in hyperscaler capex, while IC substrate capacity is expected to merely double and DRAM capacity to expand 1.8-fold; he cautioned these multiples do not necessarily signal shortages given different capacity bases. Wealth Magazine notes DRAM and substrate shortages are already emerging in 2025 and 2026, and prolonged equipment lead times and tight material supply could make it harder to synchronize capacity expansion. Per MoneyDJ, substrate sizes have grown from roughly 1x reticle size in early designs to 3x in 2025, with 5x targeted for 2026 and 5–9x by 2027–2028, potentially reaching 12–14x; per Central News Agency, AI chips can draw currents exceeding 1,000 amperes, raising demands on power integrity, high-speed signal transmission, thermal management and thermomechanical reliability, and fine-line requirements approaching 12–14x reticle size could exceed individual substrate makers' capabilities, complicating yield improvement. In Intel's EMIB supply chain, substrate yields have recently improved to around 45%, with suppliers reportedly targeting 50% by 4Q26 and 60% by 1Q27.
IC substrateCoWoSEMIBUnimicron, IntelTrendForceInventory Shift
According to a DIGITIMES Insight report, generative AI is pushing chip power consumption and heat flux to unprecedented levels, driving AI packaging thermal management requirements from several hundred watts to the kilowatt (kW) scale. Heat sink specialist Jentech Precision points out that package lids are evolving beyond conventional solid-state heat conduction toward phase-change heat spreading and microchannel liquid cooling, with TIM-less architectures representing the next horizon. The source does not state a timeline or specific implementation plan.
AI chip packaging thermal managementpackage lidsheat sinksJentech PrecisionDigiTimes Asia
As China's semiconductor equipment makers showcase growing capabilities at WESEMIBAY, SiCarrier says, according to National Business Daily, that all its semiconductor equipment products have entered mass production and are already in use at multiple Chinese wafer fabs; the report notes that specific customer names, equipment installation numbers, and production volumes for individual models remain unconfirmed. At this year's exhibition, SiCarrier showcased two major product categories—process equipment and inspection and metrology systems—spanning diffusion, thin-film processing, optical inspection, and optical metrology; according to 21jingji, its equipment brochure revealed two previously unseen products: PMI (Liupan Mountain) for photomask defect inspection and mEBI (Jinggang Mountain) for AI-powered nanoscale wafer defect detection. According to Nikkei, SiCarrier, reportedly linked to Huawei and backed by the Shenzhen government, first gained attention for claiming it had developed critical lithography machines capable of producing 28nm chips. SiCarrier subsidiary Qiyunfang is broadening its EDA portfolio beyond the schematic and PCB design tools introduced last year toward a more comprehensive electronic design toolchain; per National Business Daily, its current EDA tools focus on schematic capture and PCB layout rather than chip-level EDA, with preliminary chip-level EDA research underway but no clear product roadmap yet; per Economic Observer, Qiyunfang's schematic and PCB design software has been adopted by more than 100 enterprise customers, with the number of engineers using the tools rising from 20,000 last year to 50,000. Subsidiary Long Sight's 90GHz ultra-high-speed real-time oscilloscope has entered commercial use, reportedly adopted by Huawei and other Chinese companies and research institutions per Mydrivers; several leading Chinese customers use it for R&D testing, and company representatives said demand is outpacing production capacity. The oscilloscope's 90GHz bandwidth ranks second globally and marks a breakthrough for China in a category subject to Western export restrictions on equipment exceeding 60GHz. Per Shanghai Securities News, Long Sight also introduced a 65GHz optical sampling oscilloscope for testing 1.6T optical modules and single-lane 200G interconnects, alongside a 110GHz spectrum analyzer for 6G research and 67/110GHz vector network analyzers, with the full lineup showcased at this year's WESEMIBAY.
semiconductor process equipmentinspection and metrology systemsphotomask defect inspection tool PMI (Liupan Mountain)SiCarrier, Qiyunfang and 3 totalTrendForce
Agentrys published a case study stating that, using agentic loop engineering and its Agentrys Studio Platform, it autonomously optimized XiangShan, a Linux-class RISC-V processor: in eight hours the workflow found a configuration that, relative to an expert-tuned reference design, improved throughput by 40% while reducing power by 28% and area by 10%, and satisfied a 1.5 GHz timing objective the reference configuration could not meet; another configuration achieved approximately 0.84 W power consumption and 0.134 mm² area at 1.5 GHz. The workflow used compilation, correctness checking, cycle-accurate simulation, synthesis, physical implementation, and timing and power analysis, with approximately 50 physical evaluations; candidates that failed required checks were rejected. The XiangShan design contained approximately 143,000 lines of Chisel code across 528 source files, with 3,284 adjustable parameters organized into 17 microarchitectural subsystems; the primary benchmark was GAP BFS, a graph-processing workload in which approximately 88% of execution cycles were spent stalled on memory. As background, the article notes that at Hot Chips 2026 OpenAI presented Jalapeño, an inference ASIC built with AI in the design loop, reporting measured block-level wins over a human baseline — 56% on a BF16 multiply, alongside a 10% smaller matrix-unit area — within a roughly nine-month path from RTL to tapeout. These AI-driven optimization results are case-study reports with no direct link to chip prices or supply.