NVIDIA unveils Minerva AIOps platform for AI factories

At GTC Berlin NVIDIA announced Minerva, an AIOps platform that pulls telemetry from GPUs, networking, storage and schedulers and uses machine learning to spot degrading components before they slow training jobs. Dell, Lenovo and Supermicro will offer it in their AI factory portfolios, IREN is already running it, and pricing or GA date were not disclosed.

• Minerva builds a knowledge graph of GPUs, SuperNICs, switches and jobs, analyzing billions of data points per hour
• Output is grouped incidents with a likely root cause instead of a flood of alerts
• A built-in assistant answers natural-language questions against live telemetry and recommends fixes
• No pricing or GA date; IREN is already using Minerva in production