NVIDIA unveiled an AI-native storage architecture that offloads and shares model context across a Spectrum‑X Ethernet fabric, reporting up to 5x gains in tokens-per-second and power efficiency for context-heavy inference.
Siemens and NVIDIA are deepening a multi‑year collaboration to connect Siemens’ Xcelerator and automation portfolio with NVIDIA’s Omniverse, AI, and accelerated computing—aiming to standardize an end‑to‑end stack for digital twins, simulation, and edge AI in factories.
Open‑source models and GPU tooling now let teams run capable LLMs and diffusion models locally on RTX PCs—cutting latency, protecting data, and reducing reliance on the cloud.
Local LLM inference on NVIDIA RTX PCs: Faster, private AI on your desk Read Post »
NVIDIA’s compact DGX Spark taps Grace Blackwell hardware and a tuned AI software stack to accelerate local training, fine‑tuning, and inference—bringing data‑center‑style capabilities to the desktop.
NVIDIA has added BlueField DPUs to its Enterprise AI Factory validated design, offloading security and networking from CPUs to accelerate protected, multi-tenant AI workloads on-premises.
NVIDIA’s BlueField DPU for enterprise AI security arrives in the Enterprise AI Factory Read Post »
Nvidia is pushing beyond standalone GPUs with its Vera Rubin rack-scale platform, integrating six co-designed chips and a new NVFP4 numeric format to boost throughput and lower cost per token. Early systems like NVL72 emphasize modular MGX design, pooled memory, and networking for large-scale AI workloads.
Nvidia Vera Rubin platform: rack-scale design, NVL72 specs, and cost claims Read Post »
