➀ OpenAI collaborates with Broadcom to develop its first custom AI chip ("XPU"), targeting mass production by 2026;

➁ The move aims to reduce reliance on NVIDIA's GPUs and strengthen control over AI infrastructure;

➂ The chip design combines Broadcom's semiconductor expertise with OpenAI's AI optimization requirements.

➀ Lenovo unveiled three OLED gaming monitors at IFA 2025, including 4K 240Hz Legion Pro 32UD-10/27UD-10 and QHD 280Hz 27Q-10;

➁ The flagship Legion Pro 7 gaming laptop features AMD Ryzen 9 9955HX3D CPU, NVIDIA RTX 5080 GPU, and a 16" 240Hz OLED display;

➂ Priced from $699 (monitors) to $2,399 (laptop), products will launch between October-November 2025.

➀ Quantinuum, formed by Cambridge Quantum and Honeywell's merger, secured $600 million in funding with investors including NVIDIA's NVentures and JPMorganChase;

➁ Funds will support the launch of its trapped-ion quantum system Helios and advance fault-tolerant quantum computing;

➂ Collaborates with NVIDIA Accelerated Quantum Research Centre, RIKEN, SoftBank, and Infineon for quantum tech development.

① U.S. Senate proposes GAIN AI Act to prioritize domestic sales of advanced AI GPUs, potentially banning exports of high-end chips;

② NVIDIA criticizes the bill as 'self-defeating' and warns it could undermine U.S. global AI leadership;

③ The act sets strict technical thresholds (e.g., 2,400 TPP score) that would restrict even mid-tier GPUs like H20 from being exported.

➀ NVIDIA and Cadence collaborate to address critical pre-silicon power estimation challenges for AI chips through advanced emulation and DPA technology;

➁ Traditional power estimation methods face scalability and accuracy limitations with AI's billion-gate designs and complex workloads, necessitating gate-level analysis with full benchmark coverage;

➂ Cadence's new DPA App on Palladium Z3 achieves 97% post-silicon power correlation by executing billion-cycle emulations in hours, enabling precise optimization for energy-efficient AI hardware design.

➀ Global semiconductor market surged to $180 billion in 2Q 2025, up 19.6% YoY, led by NVIDIA's $45B revenue and robust memory sector growth;

➁ Memory giants SK Hynix (26% QoQ) and Micron (16%) drove gains, while tariffs and AI demand shaped market dynamics;

➂ 2025 full-year growth forecasts revised to 13-16%, with uncertainties around U.S.-China trade tensions and smartphone market declines.

➀ STH explains its choice of AMD EPYC over Arm in 2025 infrastructure, citing Arm's challenges in enterprise adoption, including installed base compatibility, hardware limitations, and insufficient software support;

➁ Power efficiency gains from Arm are now minimal compared to AMD/Intel x86 CPUs, especially with AI GPU servers dominating power consumption (e.g., AMD Instinct MI325X at 1kW/accelerator);

➂ Enterprise Arm adoption faces hardware scarcity and vendor neglect: NVIDIA Grace CPUs are outdated for general workloads, while major vendors like Dell/HPE offer limited Arm options (e.g., HPE ProLiant RL300 Gen11's market failure).

➀ Intel's XeSS 2.1 SDK enables cross-platform frame generation for AMD/Nvidia GPUs including GTX 10-series and RX 5000 series, though RTX 30/RX 6000+ is recommended;

➁ XeLL low-latency mode now works on non-Intel GPUs but requires active frame generation, with developers needing to adopt the updated SDK for implementation;

➂ Despite the openness, adoption is limited with only 22 games supporting XeSS 2, highlighting challenges against entrenched solutions like FSR and DLSS.

➀ Battlefield 6 (BF6) requires surprisingly modest hardware, needing a 10th-gen Intel or Ryzen 3000 CPU and RTX 3060 Ti/RX 6700 XT GPU for optimal performance;

➁ EA confirms the game won’t run on Steam Deck at launch due to discrete GPU requirements but may work on handhelds with external GPUs or AMD Strix Halo chips;

➂ The free-to-play approach drives lower specs, contrasting with recent AAA titles demanding cutting-edge hardware.

➀ NVIDIA has announced that this October's Game Ready driver release will be the last one for a number of its vintage cards, coinciding with the end of Microsoft's support for Windows 10.

➁ The affected cards include Maxwell, Pascal, and Volta families, with the Pascal series excluding 16xx-series cards.

➂ NVIDIA will still release quarterly security updates for the affected cards for the next three years and extend support for RTX-series graphics cards until October 2026.

➀ Samsung reports a 94% profit plunge in its chip division, driven by U.S. export restrictions and inventory corrections;

➁ The company pins recovery hopes on HBM3E production cost reductions to secure Nvidia's AI GPU supply chain, challenging SK Hynix and Micron;

➂ A $16.5 billion Tesla partnership and geopolitical risks, including U.S. tariffs, add complexity to Samsung's AI-driven turnaround strategy.

➀ NVIDIA confirms discontinuation of Game Ready drivers for Maxwell and Pascal GPUs post-October 2025, transitioning to quarterly security updates until 2028;

➁ Legacy GPUs like GTX 1080 Ti will lose DLSS/ray-tracing optimization but remain functional for non-RT titles;

➂ Windows 10 users receive extended driver support through October 2026 despite Microsoft ending OS updates in late 2025.

➀ China's Cyberspace Administration questioned Nvidia's H20 GPU over potential data security risks and hidden tracking functions linked to U.S. legislative proposals.

➁ The H20 is a modified export-compliant GPU for China, lacking tracking features, yet concerns may stem from efforts to boost domestic AI chip adoption (e.g., Cambricon, Huawei).

➂ Analysts view the scrutiny as symbolic, unlikely to escalate to a ban, as China still relies heavily on Nvidia's CUDA software ecosystem.

➀ Asus will launch its Ascend GX10 mini-PC powered by Nvidia's GB10 Grace Blackwell SiP on July 22, targeting AI development with data center-level performance.

➁ The GB10 integrates an Arm-based Grace CPU and Blackwell GPU, delivering 1 PetaFLOPS FP4 AI performance and 128GB unified memory, comparable to Apple's M4 Pro.

➂ Priced around $3,000, the system positions itself as a compact AI workstation, though general-purpose benchmarks show performance similar to mid-tier mobile chips like Snapdragon X Elite.

➀ The U.S. plans to require export licenses for AI GPUs to Malaysia and Thailand to prevent smuggling to China;

➁ Malaysia and Thailand are emerging as suspected transit hubs despite lacking official revenue significance for NVIDIA;

➂ Questions persist about Singapore's role as NVIDIA's key market and potential re-export routes.

➀ NVIDIA's GeForce RTX 5070 leads GPU sales on Amazon with 5,450 units sold in June 2025;

➁ NVIDIA's mid-range GPUs dominate the top selling GPUs, with GeForce RTX 50 series accounting for half of the top 10;

➂ NVIDIA holds a 70.51% market share in discrete GPUs, followed by AMD at 27.57% and Intel at 1.92%.

➀ NVIDIA股价飙升，市值一度达到3.92万亿美元，成为有史以来最有价值的公司之一；

➁ 由于美国独立日假期，NVIDIA未能打破苹果创下的3.915万亿美元的市值记录；

➂ NVIDIA数据中心业务是其最大的收入来源，最新财报显示数据中心收入超过390亿美元，而游戏业务收入仅为37.6亿美元。

➀ PCIe bandwidth significantly impacts content creation performance: RTX 5090 faces up to 25% slowdown in DaVinci Resolve with PCIe 4 x4 configurations;

➁ Multi-card setups and NVMe SSD installations may force PCIe lane sharing, cutting bandwidth for professional applications;

➂ High-end GPUs like RTX 5090 are most affected, while mainstream GPUs and AI workloads show less sensitivity to PCIe limitations.

➀ Elon Musk's xAI is acquiring an overseas power plant to ship to the U.S. for a data center housing 1 million AI GPUs, requiring up to 2 GW of power (equivalent to 1.9 million homes).

➁ Current xAI Colossus supercomputer uses 200,000 NVIDIA H100 GPUs and 300 MW of power, relying on gas turbines and Tesla Megapacks for energy stability.

➂ The AI industry's energy demands are escalating rapidly, with companies prioritizing dedicated power infrastructure over grid reliance, despite environmental and logistical challenges.

➀ Dell and CoreWeave unveiled the first NVIDIA GB300 NVL72 rack, utilizing Dell PowerEdge XE9712 servers and Vertiv cooling systems with a capacity of up to 2MW per rack;

➁ The GB300 NVL72 integrates NVIDIA's Grace Blackwell Ultra architecture, delivering 21TB of memory and leveraging InfiniBand Quantum-X networking for high-speed AI workloads;

➂ This ultra-high-density rack represents a significant advancement over previous GB200 systems, featuring enhanced liquid cooling and accelerators, with costs potentially exceeding $3.7 million per rack.