NVIDIA launches Minerva AIOps platform for AI factories

NVIDIA says traditional monitoring tools often divide the data center into separate views, while a single degraded component can slow an entire AI-factory job.[1] Minerva is designed to connect signals across GPUs, networking, storage, schedulers and workloads rather than treat those systems as isolated sources.[1] NVIDIA describes AI-factory yield as useful work measured through GPU hours, goodput, cluster uptime and token production.[1] That definition puts operational performance in terms of work delivered, not just the condition of individual hardware components.[1]

Minerva builds a dynamic knowledge graph that maps infrastructure and the jobs running on it.[1] NVIDIA says it analyzes billions of data points per hour using supervised and unsupervised machine learning.[1] The platform learns normal behavior, detects previously unseen anomalies, predicts emerging failures and correlates events using factory topology.[1] NVIDIA says this can help trace a slow training job to an affected component in a factory with hundreds of thousands of GPUs.[1] It can also group an event storm into a focused incident with a likely root cause.[1]

Minerva includes a native AI assistant and can connect to external agentic AI tools.[1] NVIDIA says operators can use natural-language conversation grounded in live telemetry to investigate cluster health and receive remediation recommendations.[1] Teams can implement fixes manually or connect Minerva to automation through infrastructure APIs and external systems.[1] NVIDIA says the platform aggregates related problems into a single actionable item to help AIOps teams move from reactive response toward proactive maintenance.[1] The described workflow combines investigation and recommendations with operator decisions about how to apply fixes.[1]

NVIDIA says IREN is among its partners using Minerva to optimize AI factories.[1] IREN, Lambda, IBM and Tesla plan to use the platform, according to a separate report.[2] Dell Technologies, Lenovo and Supermicro will offer Minerva to their own customers, according to another report.[3] The NVIDIA item is dated October 22, 2026, and names Brian Sparks.[1] NVIDIA says Minerva can help jobs run longer between interruptions by spotting trouble early and can pinpoint problems when incidents occur.[1]