A failing component can stall an entire AI training job, Nvidia says. The company’s AI operations software, Minerva, launched on Thursday, October 22, 2026, aims to catch those faults early by monitoring hardware and workloads together across large data centers.
Key takeaways
- Minerva connects hardware monitoring with AI workload data.
- Related alerts become incidents with likely causes.
- IREN already uses Minerva; five other companies plan deployments.
Nvidia’s AI operations software brings monitoring into one view
Minerva combines information from GPUs, networks, storage, job schedulers and AI workloads into a unified map of a data center. Nvidia introduced the software at its GTC Berlin conference, days after signing an AI partnership with Germany.
The software uses machine learning to establish normal operating patterns and predict failures. Rather than passing every alert to operators separately, it collects connected problems into a single incident with a likely cause.
The launch extends Nvidia’s push to sell AI data center products beyond chips. Nvidia did not disclose pricing for Minerva.
Tracing slow training jobs to the faulty component
Nvidia says Minerva can help operators identify the component behind a slow training job, even in systems spanning hundreds of thousands of GPUs. Its combined monitoring addresses problems that can go unnoticed when networking, computing and machine learning teams each track only their own part of the system.
A GPU running unusually hot or a storage drive slowing under load can hold up a training job, Nvidia says. Minerva brings those hardware signals together with information about the work running on the cluster.
A built-in AI assistant lets staff ask plain-language questions about cluster health and receive suggested fixes. Operators can make the changes manually or connect Nvidia’s AI operations software to their existing automation tools.
IREN is using Minerva, with more deployments planned
IREN already uses Minerva, while Lambda, IBM, Tesla, Bitdeer AI and Jio plan to deploy it, Nvidia said. Dell, Lenovo and Supermicro will also offer the software to their customers.
“NVIDIA Minerva gives our operations team one view across the network, the GPUs and the jobs running on them,” said Denis Skrinnikoff, IREN’s chief technology officer.
Nvidia’s relationship with IREN also includes a deal struck in May under which it can invest up to $2.1 billion in the former Bitcoin miner.
Article produced with the assistance of artificial intelligence and reviewed by the editorial team.