Nvidia has introduced Minerva, a new AIOps platform designed to provide end-to-end observability and monitoring for large-scale AI factories. The system integrates telemetry data from infrastructure and workloads, utilizing purpose-built machine learning models to generate actionable insights and detect anomalies early. By building a dynamic knowledge graph of GPUs, networking components, and active jobs, Minerva helps operators trace performance bottlenecks and root causes across massive data center deployments.
The platform features a native AI assistant and supports external agentic AI tools, allowing operators to investigate cluster health using natural-language conversations. Designed to maximize AI factory yield and GPU uptime, Minerva correlates events to reduce alert fatigue and shifts maintenance practices from reactive to proactive. Partners such as IREN, Tesla, IBM, and Dell Technologies plan to adopt or offer the platform within their operations.



