NVIDIA launches Minerva for end-to-end observability in AI factories
🎧 Listen to this story
NVIDIA launched Minerva, an AIOps platform that brings together signals from GPUs, networking, storage, schedulers and AI workloads. The company says the platform uses machine-learning models to analyze billions of data points per hour, identify anomalies, predict failures and correlate events with a likely root cause, including across factories with hundreds of thousands of GPUs. An AI assistant lets operators investigate cluster health in natural language and receive remediation recommendations, while keeping a person in the loop or connecting to automation systems.
NVIDIA says IREN is already using Minerva to monitor networks, GPUs and jobs in a single view. Aolani, Bitdeer AI, IBM, Jio, Lambda, Quake AI and Tesla plan to deploy it; Dell Technologies, Lenovo and Supermicro will offer the platform in their AI factory operations portfolios.
Why it matters · editorial interpretation
By unifying telemetry from hardware, networking, storage and AI workloads, the platform could reduce fragmentation in large-cluster operations. Combining automated recommendations with human oversight also frames how teams may balance faster response with operational control.