NVIDIA Unveils Minerva for End-to-End AI Factory Observability

NVIDIA launched Minerva, an AIOps platform designed to help operations teams identify problems in AI factories before they disrupt workloads.

The platform combines telemetry from GPUs, networking, storage, schedulers and workloads. Minerva analyzes billions of data points per hour, detecting anomalies, predicting failures and correlating events according to the infrastructure’s topology.

Mapping infrastructure and workloads

Minerva builds a dynamic graph linking GPUs, SuperNICs, switches, cables and jobs. This allows operators to trace a slow training run to the component affecting it, including in environments with hundreds of thousands of GPUs.

The platform can consolidate multiple signals into a single incident with a likely root cause, reducing the number of separate alerts operators must handle. NVIDIA says this is intended to increase time between interruptions, speed repairs and improve GPU utilization.

AI assistant and automation

Minerva includes a native AI assistant and can connect to external agentic AI tools. Operators can investigate cluster health through natural-language queries and receive remediation recommendations.

Teams can keep people responsible for implementing fixes or connect Minerva to automation through infrastructure APIs and external systems. The platform targets AI-factory yield, measured through GPU hours, goodput, cluster uptime and token production.

Partners and suppliers

IREN is already using Minerva to optimize its AI factories. IREN Chief Technology Officer Denis Skrinnikoff said a unified view of the network, GPUs and jobs helps the company focus on proactive maintenance.

Aolani, Bitdeer AI, IBM, Jio, Lambda, Quake AI and Tesla plan to deploy Minerva. Dell Technologies, Lenovo and Supermicro will offer it as part of their AI-factory operations portfolios.