NVIDIA launches Minerva, software that turns failures into automated tests. The tool groups alerts into incidents and suggests fixes via AI. Announced during GTC Berlin, the product aims to simplify the management of AI data centers.
Minerva turns failures into automated tests, simplifying the management of AI data centers. The tool combines data from GPUs, network, storage and workloads to identify failures and group alerts into single incidents with a probable cause. This allows operators to fix problems more efficiently, without having to analyze each alert individually.
AI-suggested fixes reduce response time and minimize human error, making the maintenance of AI systems more reliable. The availability of the PowerEdge R8822, scheduled for October 2026, is essential for deploying the software, since it supports AI agents. The PowerEdge M9882, with 256 NVIDIA Vera processors, will be available in the first quarter of 2027, enabling scalability and superior performance in large-scale environments. These servers, combined with Minerva, ensure that the maintenance of AI systems happens in a faster and more precise way, adapting to the growing demands of the sector.
NVIDIA Minerva offers a unified view for operations teams, integrating network, GPU and workload data in real time. Denis Skrinnikoff, chief technology officer at IREN, highlighted that the tool allows professionals to assess cluster health in plain language and receive suggestions for fixes.
"NVIDIA Minerva gives the operations team a single view across the network, the GPUs and the tasks running on them," he said (in a loose translation).
The tool also allows teams to apply the fixes manually or integrate Minerva with their own automation tools. The solution can reduce the complexity of monitoring AI data centers, simplifying failure management and improving operational efficiency. Minerva changes the way AI data centers handle failures, turning alerts into automated tests and fix suggestions. This reduces the complexity of diagnostics in environments with hundreds of thousands of GPUs. Operators can trace a slow task back to its origin, even in distributed systems.
The tool uses an AI assistant to suggest solutions, speeding up incident response without requiring manual intervention. Dell validated the model by listening to Reflection's Beam and announced the integration of Red Hat AI Factory with the NVIDIA OpenShell architecture.
Specifications
| Item | Specification |
|---|---|
| Feature | AI operations software for monitoring large AI data centers |
| Alert grouping | Single incidents with probable cause |
| PowerEdge R8822 availability | October 2026 |
| PowerEdge M9882 processors | 256 NVIDIA Vera processors |
| PowerEdge M9882 form factor | 48U rack |
| PowerEdge M9882 availability | First quarter of 2027 |
| Dell Pro Precision GB10 availability | October 23, 2026 |
With information from thenextweb.com, zdnet.fr and developer.nvidia.com.
Sources
- Nvidia minerva aiops ai factory monitoringthenextweb.com
- Dell et nvidia elargissent leur offre dinfrastructures pour agents dia des postes de travail aux datacenterszdnet.fr
- Use nvidia nemo to turn agent failures into tests and harness fixesdeveloper.nvidia.com
- Nvidia details sentry for runtime ai agent security a software framework designed to monitor agentmedium.com
Text written with AI from the sources listed and checked by the newsroom's automated rules. How we work
Want to see resumer first in Google searches?
Add to Google's preferred sources
