semiconductor

This article deeply analyzes the early career of AMD CEO Lisa Su, focusing on her choices at MIT and IBM to research hardcore technical problems such as SOI and copper interconnects, and actively supplementing her business judgment, emphasizing her judgment to pick hard, specific, and deliverable problems rather than chasing trends.

Introducing the Tiezhu website, a miniature version of Bloomberg that provides basic data, stock prices, and valuation information for 219 core companies in the global semiconductor supply chain (materials, equipment, wafers, memory, CPO, etc.).

ReflectiChain bridges the epistemic gap between LLMs and RL for supply chain resilience using a generative world model and double-loop learning, improving rationale consistency by 33% and maintaining operability under adversarial shocks.

Recommends following @citrini's AI and semiconductor analysis account, detailing the accurate prediction records of two analysts, Jukan and Zephyr, including LG Innotek's stock surge, DRAM pricing power shift, etc., emphasizing the intelligence value in AI chips, memory, and supply chain.

The UK government announces a $1.47 billion plan to build a national AI supercomputer and invest in domestic chip startups, aiming to reduce dependence on US and foreign AI hardware.

C12 unveils Pick & Place, a patented nanoassembly process that precisely positions carbon nanotubes on quantum chips, dramatically increasing manufacturing throughput—50 devices assembled in four weeks versus one year previously—and enabling scalable multi-qubit integration.

Researchers at POSTECH developed a zinc oxide-tellurium transistor that performs multiple circuit functions in a single device, reducing transistor requirements by 75% and increasing processing speeds fourfold, with potential applications in compact AI hardware and wearable electronics.

This tweet shares internal information about Samsung's 2026 profit forecast and tight memory supply at SK Hynix.

Jensen Huang responds to Samsung's strike and wage increase issue, saying that employees' salaries should be as high as possible, and reveals Nvidia's salary levels: 50% of employees earn over $25 million, and 78% earn over $1 million, with a salary slip of Beijing Nvidia employees attached.

LithoGRPO introduces a novel framework that combines flow matching with GRPO-based reinforcement learning for fast and high-quality inverse lithography mask optimization, achieving state-of-the-art performance while maintaining efficient generation.

This article uses high-resolution micrographs to deeply analyze the internal structure, transistor layout, memory interface, and major functional blocks of the Apple M1 chip, showcasing its revolutionary breakthrough in performance and efficiency.

Researchers at the University of Illinois have demonstrated a scalable method to sequentially stack high-performance silicon circuits, achieving monolithic 3D integration within strict thermal budgets, which could extend Moore's Law beyond traditional transistor shrinking.

An experienced chip designer shares insights on the differences between academic and industry chip design, covering goals, risk tolerance, and methodologies, while noting the growing demand for ASIC designers.

TSMC's senior VP says energy efficiency is now the primary constraint in AI chip design, surpassing raw computing power. The shift is driving changes in transistor density, advanced packaging, and chip stacking to reduce power consumption.

imec presents the world's first quantum dot qubit device fabricated using High NA EUV lithography, achieving nanometer-scale gaps between qubit control gates, paving the way for industrial scaling of silicon-based quantum computers.

Proposes the first unsupervised anomaly detection framework for IC latent defect screening using a Diffusion Transformer, achieving state-of-the-art performance on industrial 16nm test data.

FAB-Bench is a benchmark framework for evaluating Retrieval-Augmented Generation (RAG) systems in semiconductor manufacturing, with six diagnostic metrics and analysis across context windows. It provides 200 curated query-answer pairs and reveals context-scaling behaviors and attention dilution issues.

This article explores the open source silicon business model pioneered by AESC Silicon, which offers free core IP and charges for support and customization, similar to Linux. It highlights the potential for verifiable security and lower barriers for custom chips, and introduces IP Forge, a package manager for open-source IP blocks.

Samsung reached a tentative deal with semiconductor employees, offering potential bonuses up to $340,000 for memory chip workers amid an AI-driven boom, averting a strike.

Samsung is awarding its chip division workers an average bonus of $340,000 due to soaring profits driven by the AI boom.