Nvidia launched the Minerva AIOps platform on October 22, 2026, during the GTC Berlin event. The system is designed to monitor large-scale GPU clusters, which the company refers to as AI factories. It is intended for deployments that utilize hundreds of thousands of GPUs.
The platform aggregates telemetry data from GPUs, network components, storage, and job scheduling into a single interface. It utilizes machine learning models to process billions of data points per hour, aiming to predict hardware failures and identify anomalies across the infrastructure.
Minerva features a topology-aware knowledge graph that maps physical and logical connections between hardware components. It also includes an AI assistant that allows operators to query cluster health status using natural language, replacing the need for manual log analysis.
The platform focuses on optimizing performance metrics such as GPU hours, cluster uptime, and token production efficiency. By shifting from reactive troubleshooting to proactive maintenance, Nvidia intends to reduce repair times and improve overall system reliability.
Early adopters and partners include IREN, IBM, Tesla, and Jio. Additionally, Dell Technologies and Lenovo have committed to integrating Minerva into their own AI factory hardware offerings, alongside the release of new DSX POD reference designs.
