Sort Newest to Oldest
2026/8/20
Chinese AI startup Moonshot AI has launched Kimi K3, featuring 2.8 trillion parameters, a one-million-token context window, and architectural innovations including KDA, AttnRes, and MoE. Its overall performance is close to the world’s leading AI models while maintaining more competitive inference costs. K3 quickly reached its computing capacity limit after launch, reinforcing the implications of Jevons’ Paradox: greater model efficiency does not necessarily reduce computing demand and may instead accelerate AI Agent adoption, further driving demand for GPUs, HBM, DDR5, and enterprise SSDs.
# USA
# China
# AI
# Semiconductors
2026/8/17
As AI Factories and high-density GPU computing continue to expand rapidly, data center power infrastructure is reaching a critical turning point from traditional AC distribution toward 800VDC. NVIDIA’s latest white paper outlines several 800VDC deployment architectures, including Power Rack, Power Center, and Power Block, while incorporating technologies such as SSTs, SSCBs, and liquid-cooled power racks to lower the barriers to upgrading existing data centers. This article examines NVIDIA’s 800VDC technology roadmap, power architectures, and Rack Power Roadmap to explore the next stage of power infrastructure development for AI data centers.
# Manufacturing Industry
2026/7/28
Growing demand from AI servers, high-voltage power architectures, vehicle electrification, and industrial automation, combined with geopolitical developments and supply chain restructuring, has gradually shifted the Power semiconductors industry toward a seller’s market. This article examines the major power semiconductor product categories, end-use applications, and competitive landscape, while exploring how AI is driving higher demand for components such as MOSFETs, IGBTs, and SiC devices. It also assesses how Taiwanese suppliers, including PANJIT and APEC, may benefit from supply chain realignment.
# Stocks
# Taiwan
# Electronics
# Automotive
2026/7/14
ASML is the world's only company capable of mass-producing EUV lithography systems, making it the global leader in semiconductor manufacturing equipment and a key indicator of worldwide semiconductor capital spending and AI industry trends. This article explores ASML's position in the semiconductor industry, the differences between EUV and DUV technologies, and how demand for its equipment reflects capacity expansion in advanced process nodes and HBM. It also analyzes the company's latest financial performance, equipment orders, key takeaways from its earnings call, and the impact of AI, export controls, and global foundry investment on its future business outlook and the semiconductor industry cycle.
2026/7/8
An antitrust lawsuit against Samsung, SK hynix, and Micron has been accepted, alleging that the three leading memory manufacturers shifted production capacity from conventional DRAM to higher-margin HBM in response to booming AI demand. The plaintiffs claim this strategy has contributed to a roughly 700% increase in DRAM prices over the past four years. This article examines the key issues surrounding the lawsuit, explains how HBM capacity expansion is affecting memory supply and demand, and explores the oligopolistic structure of the DRAM market, the drivers behind soaring memory prices, and the potential implications for the memory industry and the broader AI supply chain.
2026/7/2
As computing power requirements for AI and HPC chips continue to rise, advanced packaging is shifting from wafer-level packaging toward panel-level packaging (PLP). This article explains PLP’s core technologies, the advantages of glass substrates and through-glass via (TGV) processes, and the differences between FOPLP and mainstream packaging technologies such as FOWLP, CoWoS, CoPoS, and EMIB. It also explores how demand for AI GPUs, chiplets, and heterogeneous integration is driving PLP market growth, while analyzing its future opportunities and technical challenges across the advanced packaging supply chain. Finally, this article highlights Taiwan's key companies involved in PLP technology and discusses their future outlook.
As Automate 2026 Chicago highlighted the next phase of smart manufacturing, with Physical AI emerging as the driving force behind industry transformation. From the development paths of wheeled and bipedal humanoid robots, to NVIDIA's Halos for Robotics safety system, as well as advances in simulation-first design, digital twins, AI vision, robotic touch sensing, and collaborative robots, the exhibition demonstrated that industry competition is no longer centered solely on hardware performance. Instead, it has expanded to include AI capabilities, hardware-software integration, and ecosystem development, making the event a key indicator of future trends in smart factories and industrial automation.
2026/6/14
The Trump administration is shifting toward Section 301 as the foundation for a new long-term tariff framework. The Office of the United States Trade Representative (USTR) has proposed additional tariffs ranging from 10% to 12.5% on 60 economies, covering most goods imported into the United States, while maintaining exemptions for semiconductors, energy products, medical goods, and critical minerals. This article summarizes the latest developments regarding the proposed Section 301 tariffs, the scope of exempt products, and the potential implications for Taiwan, China, Vietnam, Japan, and global supply chains. It also examines the future direction of Trump's tariff policy and the challenges businesses may face.
# Consumer Discretionary
# Life Sciences
# Basic Materials
# Basic Industries
# Energy
2026/6/9
As AI models continue to grow in scale, the competitive focus of data centers is shifting from GPU computing power to high-speed connectivity. At Computex 2026, Marvell highlighted that AI infrastructure is gradually entering the era of the “connectivity bottleneck,” where technologies such as optical interconnects, Co-Packaged Optics (CPO), silicon photonics, and high-speed SerDes will become critical enablers of future growth. This article summarizes Marvell’s perspective on the evolution of AI data center architectures and how optical connectivity is expected to drive the next wave of AI infrastructure innovation.
2026/6/2
NVIDIA GTC 2026 showcased a pivotal transition in the AI industry from Generative AI to Agentic AI. From the Vera Rubin platform and DSX AI Factory to the Agent Toolkit and Physical AI ecosystem, NVIDIA is accelerating the development of a comprehensive AI infrastructure spanning data centers, enterprise applications, personal computers, and robotics, offering a glimpse into the next wave of growth in AI computing and the semiconductor industry.
# Power Infrastructure
# Technology
# Communication Services
