NVIDIA has launched Minerva, a new software platform designed to give operators of massive AI data centers a single, unified view of GPU, network and storage health.
The goal is to catch hardware problems before they slow down AI training jobs.
The company unveiled the platform on October 22 as part of its GTC Berlin announcements, describing it in a company blog post as an AIOps platform built to help teams find problems earlier and keep AI factories productive.
What Minerva does
It brings together telemetry from infrastructure and AI workloads, then applies purpose-built AI models to turn that data into timely, actionable insight.
Minerva connects those signals across NVIDIA GPUs, networking, storage, schedulers and workloads.
The platform uses those signals to build a dynamic knowledge graph of the infrastructure, including GPUs, SuperNICs, switches and cables, and the jobs running on it.
This unified topology allows operators to trace a slow training job back to the specific component causing the issue, even in clusters with vast GPU counts.
It analyzes billions of data points per hour using supervised and unsupervised machine learning. It learns normal behavior, detects previously unseen anomalies, predicts emerging failures and correlates related events based on the factory’s topology.
Rather than overwhelming operators with disconnected warnings, it can group an event storm into a focused incident with a likely root cause, according to the company’s announcement.
Minerva also ships with a native AI assistant that accepts natural language queries. Operators can ask questions about cluster health in plain language rather than writing queries against raw logs. It can also connect to external agentic AI tools and automation.
Why uptime matters at this scale
The platform focuses on AI factory yield, a measure of the useful work infrastructure produces through GPU hours, goodput, cluster uptime and token production.
AI factories rank among the most complex operations in the world, NVIDIA says, with huge fleets of GPUs training models and serving AI applications across high-speed networks.
A single degraded component at that scale, a GPU running hotter than its peers or an NVMe drive losing performance under load, can slow an entire job, and such warning signs are easy to miss.
NVIDIA says the fragmentation of existing monitoring tools is part of the problem it is solving.
Traditional monitoring tools often divide the data center into separate views. Networking, compute, site reliability and machine learning teams each see a different part of the operation. Minerva is pitched as a way to merge those views into one.
Early adopters and hardware partners
Cloud and compute provider IREN is already running the platform. Its chief technology officer, Denis Skrinnikoff, said “As we scale our AI factories, the key to providing the best service for our customers is maximizing the productivity of our GPUs,” according to NVIDIA.
He added that 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.”
Joining IREN, Aolani, Bitdeer AI, IBM, Jio, Lambda, Quake AI and Tesla plan to deploy Minerva.
In addition to Dell Technologies, Lenovo and Supermicro will offer Minerva as part of their AI factory operations portfolios.
Industry coverage of the launch also placed the product within NVIDIA’s broader hardware roadmap. According to Lambda, one of the early adopters, Minerva connects telemetry across GPUs, networking, storage, schedulers, and AI workloads, and its dynamic knowledge graph maps infrastructure components to running jobs, while specialized AI models detect anomalies, predict emerging failures, and correlate related alerts to identify likely root causes.
Why it matters
As AI labs and cloud providers race to build ever-larger GPU clusters, unplanned downtime translates directly into lost compute and wasted power, costs that scale with cluster size.
NVIDIA’s bet with Minerva is that software-level visibility across an entire fleet, rather than siloed monitoring by separate teams, can meaningfully lift the output those expensive GPU investments actually deliver.
That concern echoes the broader infrastructure race also visible in projects like linking data centers into single machines and in efforts to expand access to supercomputing resources such as the Barcelona supercomputer’s AI tools.
