AI news for tracking tools, models, and market shifts

NVIDIA highlights performance-per-watt as the critical metric for AI infrastructure efficiency, emphasizing that power limits directly impact the profitability and revenue of large-scale AI deployments.

NVIDIA unveils Spectrum-6 networking infrastructure, engineered for Vera Rubin to power gigascale AI factories. The system supports hundreds of thousands of GPUs and CPUs for frontier model training and agentic AI.

Wistron has officially opened its first U.S. manufacturing facility in Fort Worth, Texas. The 324,000-square-foot greenfield plant produces specialized superchips that serve as the core hardware for NVIDIA’s most advanced artificial intelligence systems.

NVIDIA unveils Jetson Thor-based T3000 and T2000 modules, offering compact, power-efficient computing to deploy foundation models in mainstream robotics and edge AI.

Bristol Myers Squibb (BMS) is deploying a second NVIDIA DGX SuperPOD built on Vera Rubin, expanding its existing life sciences AI cluster dubbed the “SuperDuperPOD.”

NVIDIA showcases agentic and physical AI advancements at SIGGRAPH, highlighting breakthroughs in open models and real-time simulation that are reshaping media, content creation, and robotics industries.

NVIDIA unveils Vera Rubin, designed to maximize intelligence per dollar for post-training workloads in the agentic AI era through extreme codesign.

NVIDIA and Japanese partners showcase full-stack AI and robotics advancements across manufacturing, infrastructure, and gaming industries this week.

NVIDIA launches Nemotron Labs, promoting open-source models to help enterprises build trusted, controllable, and customizable AI solutions tailored to specific business workflows and domain knowledge.

NVIDIA unveils Vera, an AI accelerator optimized for agentic workflows with a focus on high single-threaded CPU performance to boost reasoning and tool-calling speeds.

NVIDIA reports nations are deploying AI to drive economic growth, ensure data sovereignty, and innovate in healthcare and transport, marking a significant shift in global infrastructure investment.

NVIDIA Nemotron 3 Ultra achieves benchmark-leading performance with LangChain's Deep Agents harness, offering higher accuracy and throughput than top closed models at a lower cost.

NVIDIA and Hugging Face partner to advance LeRobot, providing open robotics developers with shared models, datasets, and validation tools to overcome fragmented physical AI development.

NVIDIA reports that open frontier models and infrastructure are now foundational to AI research, highlighted by 74 accepted papers at ICML 2026.

NVIDIA partners with Anthropic to integrate its BioNeMo Agent Toolkit into Claude Science, empowering life sciences researchers with GPU-accelerated AI for advanced biological workflows.

NVIDIA shifts focus to production inference, inviting partners to build scalable 'AI factories' for efficient, multi-tenant token generation.

NVIDIA explores three workflows for enhancing Vision AI agent accuracy using synthetic data and fine-tuning within the Omniverse ecosystem, leveraging OpenUSD for industrial applications.

NVIDIA pivots from peak chip specs to cost-per-token metrics, leveraging a codesigned software stack to optimize inference efficiency across GPUs, CPUs, and networking for lower production costs.

Anthropic's Claude models are now generally available on Microsoft Azure, powered by NVIDIA GB300 Blackwell Ultra GPUs, enabling enterprises to build advanced autonomous AI agents with enhanced performance.

Palantir launches a secure AI engine for US government agencies using NVIDIA Nemotron open models, highlighting the role of open-source innovation in national technology leadership.