NVIDIA Introduces New AI System for Agent Inference

NVIDIA has launched its Vera Rubin NVL72 rack-scale system, now in full production with Groq 3 LPX, designed to enhance AI agent inference. This new system aims to deliver significantly higher throughput per megawatt and lower token costs, boosting performance for agentic AI workloads requiring fast token generation and large context windows. Industry partners like SpaceXAI and CoreWeave are adopting the platform.