➀ Researchers achieved record-high electronic quality in graphene using "proximity screening" with hexagonal boron nitride (hBN) layers, reducing charge fluctuations and achieving carrier mobility up to 5.7×10⁷ cm²/Vs;

➀ The ultrathin hBN-graphene heterostructure suppressed noise and enabled quantum phenomena like the quantum Hall effect, with potential applications in high-speed transistors and quantum technologies;

➂ The scalable fabrication method paves the way for disorder-free 2D material devices compatible with existing semiconductor processes.

➀ Researchers developed electro-optical Mott neurons that produce synchronized electrical and optical signals, enabling computation and communication in one element.

➀ The neurons use niobium dioxide (NbO₂) to emit visible light during electrical switching, eliminating energy-intensive signal conversion.

➂ Potential applications include real-time metrology, in-sensor computer vision, and efficient neuromorphic systems for optical communication.

➀ Qualcomm launches Dragonwing Q-6690, the first enterprise mobile processor with integrated UHF RFID, 5G, Wi-Fi 7, Bluetooth 6.0, and UWB;

➁ Enables proximity-aware applications in retail, logistics, and industrial sectors, with adoption by OEMs like Zebra and Honeywell;

➂ Offers software-configurable feature packs and over-the-air updates, reducing hardware redesign needs and extending product lifecycles.

➀ Rebellions unveiled its REBEL-Quad AI accelerator at Hot Chips 2025, featuring four ASICs with 144GB HBM3E memory and UCIe chiplet interconnects.

➁ The PCIe Gen5 accelerator demonstrated live inference of Llama 3.3 70B models at 35.5ms/token using Samsung SF4X and CoWoS-S packaging.

➂ This marks a significant commercial implementation of UCIe technology, showcasing inter-chip communication advancements in AI hardware design.

➀ NVIDIA CEO Jensen Huang and TSMC CEO C.C. Wei shared a lighthearted exchange about billionaire status and wafer pricing during dinner meetings in Taipei;

➁ Discussions reportedly revolve around China-specific B30 Blackwell AI chips, with Nvidia seeking TSMC's production support amid export restrictions;

➂ Huang confirmed collaborating with TSMC on six revolutionary chips including Rubin-architecture photonics processors, emphasizing the foundry's critical role in NVIDIA's roadmap.

➀ Dutch edge AI chip developer Axelera AI is seeking €150 million to advance its AI chiplet technology, targeting applications in robotics, automotive, and datacenters by 2028;

➁ The EU has supported Axelera with grants (€68 million) and investments, emphasizing its role in building Europe’s sovereign AI industry and reducing dependency on foreign chips;

➂ Axelera’s upcoming Titania chiplet utilizes proprietary Digital In-Memory Computing (D-IMC) and RISC-V architecture, with CEO Del Maffeo highlighting talent acquisition and competitive company culture as key drivers.

➀ Ed questions the commercial viability of specialized humanoid robots developed by ScunnyAI, despite global interest in the technology;

➁ ScunnyAI's robots can perform specific tasks (e.g., making tea, cleaning cars) but lack general-purpose adaptability, limiting their immediate practical use;

➂ Ed proposes spinning off ScunnyAI as an independent company, leveraging hype and government-backed narratives to attract investors, while negotiating a 25% stake in the venture.

➀ Agentic AI's integration into semiconductor workflows accelerates design tasks but introduces risks of opaque decision-making and unintended behaviors;

➀ Security concerns arise from compromised hardware and lack of transparency tools for engineers, necessitating rigorous output validation;

➂ EDA vendors are implementing containment strategies like restricted access and predictable AI modules to mitigate risks in advanced chip designs.

➀ The U.S. White House allows NVIDIA and AMD to export AI chips to China under a licensing agreement requiring 15% of revenue to be paid to the U.S. government, expected to generate $2 billion in 2025;

➁ NVIDIA plans to sell over $15 billion of its H20 chip (a downgraded product for compliance) to China, while AMD aims for $800 million in M1308 chip sales, amid debates over export restrictions and technology competitiveness;

➂ NVIDIA emphasized the need for flexible export controls to maintain U.S. leadership in AI, with potential expansion of the licensing model to semiconductor manufacturing equipment.

➀ Intel CEO Lip-Bu Tan reaffirms his commitment to U.S. leadership and Intel’s transformative role amid political pressure;

➁ Tan denies allegations about past roles at Walden International and Cadence, emphasizing adherence to ethical standards;

➂ Reports reveal internal board disagreements on strategies like foundry spin-offs, fab investments, and AI acquisitions, despite public statements of support.

➀ Two Chinese nationals face up to 20 years in prison for illegally exporting AI GPUs worth tens of millions to China via shell company ALX Solutions Inc., bypassing U.S. export controls;

➁ Shipments were routed through Singapore and Malaysia, with payments traced to Hong Kong and Chinese entities;

➂ The GPUs, speculated to be NVIDIA’s H100 or B200, were labeled fraudulently to avoid detection.

➀ AMD reports record $7.685B Q2 2025 revenue driven by 69% YoY growth in client CPUs and 73% surge in gaming GPUs;

➁ An $800M write-off from MI308 GPU export restrictions to China slashed gross margins to 40%, but net income still rose 229% YoY;

➂ Projects Q3 revenue of $8.7B±$300M, fueled by MI350 AI accelerators and holiday season demand.

➀ PCI-SIG announces PCIe 8.0 with 256 GT/s per lane, enabling 1 TB/s bi-directional bandwidth;

➁ Challenges in copper-based implementations drive exploration of optical interconnects and advanced packaging;

➂ Targeting AI servers, data centers, and HPC, with consumer adoption expected post-2030.

➀ Neuromorphic computing addresses the limitations of von Neumann architectures by mimicking the brain's parallel processing for energy-efficient AI tasks like computer vision and autonomous systems;

➀ Key innovations include synaptic plasticity integration and in-memory computing using technologies like ReRAM, enabling real-time learning and reduced power consumption;

➂ Emerging quantum tunneling advancements could further enhance computational intelligence in next-gen neuromorphic hardware.

➀ Socionext licenses aiMotive's aiWare IP and tools to develop chiplet-based SoCs for automotive ADAS;

➁ The collaboration aims to accelerate AI adoption in automotive solutions, integrating energy-efficient neural network accelerators;

➂ Partnership focuses on mass production-ready Custom SoCs with SDK support, enhancing road safety and intelligence.

➀ China launches two AI alliances: the Model-Chip Ecosystem Innovation Alliance and the Shanghai AI Committee, aiming to reduce reliance on U.S. tech.

➀ The Model-Chip Alliance focuses on integrating domestic AI hardware (e.g., Huawei, Biren) with software to streamline development amid restricted access to NVIDIA GPUs.

➁ The alliances seek interoperability through standardized APIs and frameworks, potentially accelerating China's AI ecosystem and global competitiveness.

➀ Samsung secures a $16.5B deal to produce Tesla's next-gen A16 AI chips at its Texas fab, aiming for production by 2029;

➁ Elon Musk highlights strategic collaboration, including Tesla engineers assisting in Samsung's manufacturing efficiency;

➂ The contract boosts Samsung's foundry credibility against TSMC, targeting 2nm-class automotive-grade chips for autonomous vehicles.

➀ Researchers at Sejong University developed STAU, a hardware accelerator enabling edge devices to run large AI models like BERT and GPT with 5.18× speedup over CPUs and 97% accuracy;

➁ The design uses a Variable Systolic Array (VSA) and Radix-2 softmax optimization to reduce computational complexity and power consumption, cutting processing time by 68% for long inputs;

➂ Implemented on FPGA with a custom 16-bit floating-point format, STAU supports multiple transformer models via software updates, advancing on-device AI deployment without cloud dependency.

➀ TSMC reported a record $30.07 billion revenue in Q2 2025, driven by AI processor demand;

➁ Plans to build 15+ new fabs worldwide, including 11 in Taiwan and overseas expansions;

➂ N2 node enters mass production in late 2025, with AI-focused A16 node on track for 2026.

➀ The U.S. will allow NVIDIA to resume H20 AI GPU sales to China after a $5.5B export ban loss, with licenses pending approval;

➁ NVIDIA’s China-specific H20 faced prior restrictions under both Biden and Trump administrations, despite CEO Jensen Huang's lobbying against export controls;

➂ The move follows U.S.-China trade talks and rare earth/EDA concessions, though H20 sales remain conditional, pushing NVIDIA to develop compliant alternatives like RTX Pro GPUs.