NVIDIA Minerva
NVIDIA's AIOps platform that correlates GPU, network and job telemetry in a knowledge graph to find and fix AI factory faults.
Overview
NVIDIA Minerva, launched October 22, 2026 alongside GTC Berlin, is an AIOps platform for AI factories that combines infrastructure and AI workload telemetry to detect problems earlier and raise AI factory yield, which NVIDIA defines as useful output such as GPU hours, goodput, cluster uptime and token production. It builds a knowledge graph of GPUs, SuperNICs, switches, cables and the jobs running on them, so an operator can trace a slow job to the failing component. IREN already uses it, and Dell Technologies, Lenovo and Supermicro will offer it in their AI factory portfolios.
This is a sourced directory entry. Features and capabilities are reported in the linked sources; this entry does not document independent Aymorphic testing.
Key features
- Dynamic knowledge graph linking GPUs, SuperNICs, switches, cables and running jobs
- Supervised and unsupervised machine learning for anomaly detection and failure prediction
- Groups event floods into focused incidents with a likely root cause, correlated by topology
- Native AI assistant with natural-language remediation, plus connectors to external agentic AI tools
- Human-in-the-loop or automated fixes through infrastructure APIs
- Analyzes billions of data points per hour; designed for sites with hundreds of thousands of GPUs
Use cases
- Tracing a degraded training job to a faulty cable, switch port or GPU
- Predicting hardware failures before they interrupt long training runs
- Reducing alert noise for operators of large GPU clusters
Pricing
Unknown; the launch post gives no pricing. Dell Technologies, Lenovo and Supermicro will offer it as part of their AI factory portfolios.
Pricing reflects the source record and may change. Confirm current terms with the provider.
Dates and sources
- Released
- 22 Oct 2026
- Record updated
- 27 Oct 2026
- Record last checked
- 27 Oct 2026