Up to 30x More Work Per Watt: NVIDIA Vera Rubin NVL72 Sets a New Efficiency Standard for AI Agents
NVIDIA announces the Vera Rubin NVL72 platform, which sets a new efficiency standard for AI agents. According to OpenRouter data, agentic AI workloads consume significantly more tokens than standard chat requests—up to 15x more—due to the complex operations involved in tasks such as researching companies, querying financial databases, and running peer comparisons via sub-agents. The Vera Rubin NVL72 platform targets this efficiency gap, aiming to deliver up to 30x more work per watt compared to previous generations, thereby improving the cost-performance profile of agentic AI deployments.