NVIDIA has a new product for the people who keep giant GPU clusters running. On October 22 the company launched Minerva, an AIOps platform meant to spot trouble in what NVIDIA calls AI factories before it slows down training or inference. The pitch is that at this scale one slightly overheating GPU or a storage drive losing speed under load can drag down an entire job, and the warning signs are easy to miss.
The core idea is a single model of the whole facility. NVIDIA says operations teams usually split the data center into separate views for networking, compute, site reliability and machine learning. Minerva ingests telemetry from NVIDIA GPUs, networking, storage, job schedulers and the workloads themselves, then builds a dynamic knowledge graph that ties physical components such as GPUs, SuperNICs, switches and cables to the jobs running across them. That map is what lets an operator trace a sluggish training run back to the part causing it, even in a site with hundreds of thousands of GPUs, according to the company.
Analysis is done with machine learning rather than fixed thresholds. Minerva processes billions of data points an hour, NVIDIA says, using supervised and unsupervised methods to learn what normal looks like, catch anomalies it has not seen before, predict emerging failures and connect related events through the facility's topology. Instead of forwarding every symptom, it is designed to collapse a flood of alerts into one incident with a probable root cause.
There is also a conversational layer. The platform ships with its own AI assistant and can connect to outside agentic tools, so operators can ask about cluster health in natural language and get suggested fixes grounded in live telemetry. Teams can apply those fixes by hand or wire Minerva into automation through infrastructure APIs. NVIDIA frames the outcome around what it calls AI factory yield, a blend of GPU hours, goodput, uptime and token output. In other words, the company wants buyers to measure the software by how much more useful work their existing hardware delivers, not by the number of dashboards it replaces.
Early adoption is concentrated among GPU cloud operators. IREN, whose chief technology officer Denis Skrinnikoff is quoted in the announcement, already uses Minerva. Tesla, IBM, Lambda, Jio, Bitdeer AI, Aolani and Quake AI intend to roll it out, while Dell Technologies, Lenovo and Supermicro will offer it within their AI factory operations offerings.
The move extends NVIDIA's reach from hardware and networking into the software that runs the facilities built on them. NVIDIA did not publish pricing, measured uptime gains or details of the models inside Minerva, and the benefits it lists, longer intervals between interruptions and faster repairs, are not quantified. Whether the platform works as well on mixed hardware as on all-NVIDIA estates is also not addressed.