Tag
NVIDIA 和 Cisco 联合推出可插拔光学模块（USFP），首次支持每通道 400Gb/s，带宽密度提升 4 倍，帮助 AI 数据中心在不增加网络层级的情况下连接更多算力并降低单位功耗。
一位拥有12年以上苹果产品开发经验的博主正在测试搭载NVIDIA RTX Spark、128GB内存并支持CUDA的Surface Laptop Ultra，考虑从Mac转向Windows。
在OCP全球峰会上，NVIDIA通过开源MGX架构向OCP贡献了Vera Rubin机架与计算托盘设计，并与Meta、Google、Cisco等合作伙伴发布了AI工厂开放标准、网络互连及800V直流供电等新进展，以推动全球AI工厂的规模化建设。
据报道，NVIDIA RTX 5080 24GB 版本预计将于 2027 年第一季度推出，但板卡合作伙伴仍在等待更多细节。
NVIDIA 的 CUDA 软件生态正推动 RTX Spark 笔记本热销，微软 Surface Laptop Ultra 的 128GB 统一内存顶配版（$5,899.99）已售罄，凸显开发者对本地运行大型 AI 模型的强劲需求。
NVIDIA 提出 Elastic Latent Recurrence (ELR) 训练配方，使循环（looped）语言模型在单次循环时达到与同规模非循环模型相当的精度，同时每次额外循环都能以更多推理计算换取精度提升，最终产出可在 1-4 循环间弹性的单一检查点 Koi-4B。
NVIDIA researchers propose NV-Tesseract-AD, a curriculum-enhanced conditional diffusion framework with a Transformer backbone for multivariate time-series anomaly detection, achieving ~3.7× F1 improvement over non-curriculum baselines and strong zero-shot transfer to EasyTSAD benchmarks.
Nvidia正在与初创公司Reflection AI进行早期谈判，考虑加大投资或提供算力支持，甚至可能提出收购要约。Reflection AI近期发布了其首个开放权重模型Beam，主打性价比，面向企业和开发者的编码与AI智能体任务。
A commentary on how tech billionaires like Jensen Huang, Elon Musk, and Mark Zuckerberg are adopting Trump's preferred term 'super intelligence' over 'AI,' driven by their deep financial entanglement with federal contracts, permits, and regulatory favor.
An animated explainer breaking down where Big Tech's ~$800 billion AI spending actually goes, arguing that GPU depreciation (~$60 of every $100) — not electricity (~$7) — is the real cost of running AI data centers, and questioning whether the AI boom can pay off before a 2000-style fiber bubble repeats.
RTX 5090 gaming GPUs are spiking to around $5,000 as AI companies purchase them in bulk by the pallet, driving consumer card prices skyward.
Gavin Baker argues that open-weight models reducing margins at the model layer are positive for AI infrastructure, noting that since Silicon Data's token index peaked on May 31, B300 GPU prices are up 38% and B200 prices up 10%, with other GPU indices rising even more.
New fake RTX 4090 graphics cards are circulating on the market, featuring GPUs and memory chips that are nearly indistinguishable from genuine ones, posing a risk to buyers.
Ting-Wei 'Willy' Sun, a former contractor charged alongside Super Micro co-founder Yih-Shyan Liaw, pleaded guilty to conspiring to smuggle about $2.5bn of AI servers powered by Nvidia chips to China in violation of US export controls.
Nvidia is reportedly in talks to acquire Reflection AI, a US startup focused on open AI models, in a move that would expand the chipmaker's footprint in the AI model ecosystem.
AI-related stocks including Nvidia, Oracle, and CoreWeave dropped sharply after OpenAI confirmed to investors that its annualized revenue was roughly $50 billion, well below the widely reported $68 billion figure, raising concerns about AI sector valuations ahead of OpenAI's expected IPO.
Nvidia据报已停止GeForce RTX 5090的生产，转而将产能分配给AI数据中心和专业GPU，预计供应短缺将推高价格，RTX 5080 24GB版本或将取代成为新的游戏旗舰。
NVIDIA reportedly discontinuing the RTX 5090, with GB202 GPUs being redirected to the professional RTX PRO series instead of consumer cards.
NVIDIA announces Boro, an open-source tool that uses AI to assist with Linux kernel development, aiming to streamline code review and maintenance workflows for kernel contributors.
Nvidia RTX Spark N1X 腕上型AI筆電（如 ASUS ProArt P16 OLED，128GB RAM、2TB SSD）現已在英國開放預購，售價 £5999，並有較低記憶體的其他配置可選。