TechForge

23rd October 2026

NVIDIA has launched Minerva, an AIOps platform built to keep AI factories productive by correlating infrastructure and workload telemetry in one place.

The company announced the tool today and it’s aimed at operators whose clusters are large enough that a single degraded component can slow an entire training job. IREN is a named partner deployment.

Correlating telemetry across GPU fabrics

Conventional monitoring splits a data centre into separate views. Networking teams watch one slice, compute engineers another, and site reliability staff work from consoles that don’t share a model with the machine learning group. Minerva merges those signals into a single operating view.

The platform ingests telemetry from GPUs and the network fabric. Storage contributes too, along with the schedulers that decide where jobs land. From that input, it builds a dynamic knowledge graph and maps hardware against the jobs running on it. Operators can trace a slow training job to the component affecting it, even inside a factory with hundreds of thousands of GPUs.

Idle GPU time is the cost sitting underneath all this. When a job stalls on a card running hotter than its peers, or an NVMe drive loses throughput under load, goodput falls across the cluster and token production follows it down. The warning signs are easy to miss at that scale.

Minerva analyses billions of data points per hour, drawing on supervised and unsupervised machine learning. The models learn normal behaviour for a given topology, detect anomalies the platform hasn’t encountered before, and predict failures that are still emerging.

Alert volume is where a lot of this work lands. Without correlation, one marginal cable can leak symptoms into every dashboard in the factory, and the operator ends up triaging the storm instead of the cause. Minerva groups an event storm into a focused incident with a likely root cause attached, which is the difference between chasing alerts and scheduling maintenance.

Agentic remediation in production operations

A native AI assistant gives operators a natural-language interface grounded in live telemetry. They can interrogate cluster health and receive recommended remediation through conversation, then choose whether a person implements the fix or Minerva hands it to automation through infrastructure APIs and external systems. Connections to third-party agentic AI tools are supported as well.

NVIDIA frames the payoff as AI factory yield, measured in GPU hours and goodput. The pitch covers longer intervals between interruptions, quicker repair once something breaks, and fewer alerts reaching an operator’s queue. None of that’s independently verified yet but IREN runs AI factories and has adopted the platform.

Denis Skrinnikoff, CTO of IREN, said: “As we scale our AI factories, the key to providing the best service for our customers is maximising the productivity of our GPUs

“NVIDIA Minerva gives our operations team one view across the network, the GPUs, and the jobs running on them, so we focus more on proactive maintenance rather than troubleshooting. That visibility strengthens our operating model for the infrastructure we’re building next.”

See also: Observability budgets rise as AI adds systems and cost pressure

Banner for Cyber Security Expo by TechEx events.

Want to learn more about cybersecurity from industry leaders? Check out Cyber Security & Cloud Expo taking place in Amsterdam, California, and London. The comprehensive event is part of TechEx and is co-located with other leading technology events including the AI & Big Data Expo. Click here for more information.

Developer is powered by TechForge Media. Explore other upcoming enterprise technology events and webinars here.

About the Author

Senior Editor

Ryan Daws is a senior editor at TechForge Media with over a decade of experience in weaving narratives and dissecting complex topics. His articles and interviews with industry leaders have earned him recognition as a key tech influencer from numerous organisations. Under his leadership, publications have been praised by analyst firms for their excellence and performance. Connect with him on X, Mastodon, Bluesky, Threads, and/or LinkedIn.

Related

5th November 2026

4th November 2026

4th November 2026

3rd November 2026

Join our Community

Subscribe now to get all our premium content and latest tech news delivered straight to your inbox

Popular

20296 view(s)
13518 view(s)
11755 view(s)
10339 view(s)

Subscribe

All our premium content and latest tech news delivered straight to your inbox

This field is for validation purposes and should be left unchanged.