NVIDIA Launches Minerva Platform for Enhanced AI Factory Observability

Image: NVIDIA · Source

NVIDIA introduces Minerva, an AIOps platform integrating telemetry and AI to proactively manage AI factories, improving productivity by detecting and resolving issues across large GPU infrastructures.

NVIDIA has launched Minerva, a new AIOps platform designed to provide comprehensive observability and operational insight for large-scale AI factories. These AI factories, which involve extensive networks of GPUs working across high-speed connections to train models and serve applications, face challenges due to the complexity and scale of their infrastructure. A single malfunctioning component, such as an overheating GPU or a degrading NVMe drive, can significantly hinder overall performance.

Minerva addresses these challenges by consolidating telemetry data from various sources including NVIDIA GPUs, networking equipment, storage, schedulers, and AI workloads into a single dynamic knowledge graph. This graph maps the relationships between infrastructure elements—GPUs, SuperNICs, switches, cables—and the jobs running on them, enabling operators to trace performance issues to their root causes even in environments with hundreds of thousands of GPUs.

The platform processes billions of data points every hour using both supervised and unsupervised machine learning models that learn normal operational behavior, identify novel anomalies, predict potential failures, and correlate related events based on the infrastructure topology. By aggregating symptoms into focused incidents with probable causes, Minerva reduces alert noise and streamlines troubleshooting.

Minerva also features a native AI assistant and supports integration with external agentic AI tools, offering a natural-language interface for operators to query cluster health and receive remediation recommendations. This keeps human teams engaged while allowing automation systems to implement fixes through APIs.

Key benefits of Minerva include extending mean time between failures by early problem detection, accelerating repair times by precise fault identification, and providing clearer operational insights by filtering irrelevant data. These outcomes help AI factories maximize GPU uptime and overall yield metrics such as goodput, cluster availability, and token production.

Several partners, including IREN, Aolani, Bitdeer AI, IBM, Jio, Lambda, Quake AI, and Tesla, plan to deploy Minerva. Hardware providers Dell Technologies, Lenovo, and Supermicro will integrate Minerva into their AI infrastructure offerings.

Minerva represents a significant step toward proactive AI factory management, enabling operators to maintain productivity in increasingly complex AI development environments.