NVIDIA has launched Minerva, a tool that watches huge AI data centers and flags failing parts early, aiming to cut slowdowns and wasted cost.
NVIDIA has launched Minerva, a monitoring platform for the giant computer farms that train and run AI. The idea is simple. Catch a failing part before it drags down the whole operation, and turn a flood of alerts into one clear problem to fix. IREN, IBM, Lambda and Tesla are among the companies that plan to use it.
NVIDIA calls these sites AI factories. They pack vast numbers of GPUs, the chips that do the heavy AI work, connected by fast networks. Minerva collects data from the chips, networks, storage and the jobs running on them. It then uses AI models built for this purpose to turn that data into advice that operators can act on.
At this scale, one weak part can slow an entire job. NVIDIA gives two examples that are easy to miss: a GPU running hotter than its neighbors, or a storage drive losing speed under heavy use. Different teams often see only their own slice of the data center, so nobody spots the pattern.
Minerva builds a map of the hardware and the jobs that run on it. That lets operators trace a slow job back to the specific part causing it, even in a site with hundreds of thousands of GPUs, according to NVIDIA.
The platform analyzes billions of data points per hour. It learns what normal looks like, flags odd behavior it has not seen before and tries to predict failures. Instead of sending every symptom to a person, it can group a storm of alerts into one incident with a likely root cause.
NVIDIA says the goal is to measure and raise what it calls AI factory yield. That means how much useful work the equipment produces, including GPU hours, uptime and the amount of AI output.
IREN, Aolani, Bitdeer AI, IBM, Jio, Lambda, Quake AI and Tesla plan to deploy Minerva. Dell Technologies, Lenovo and Supermicro will offer it as part of their AI factory operations products. IREN's chief technology officer, Denis Skrinnikoff, said it gives his team one view across the network, the GPUs and the jobs, so they can focus on prevention rather than troubleshooting.
The announcement does not give a price or a rollout schedule. It also offers no independent results showing how much downtime or cost Minerva saves. For ordinary users, the hoped-for benefit is indirect: AI services that run more steadily and waste less money behind the scenes.
Minerva is a tool meant to keep AI data centers healthy, but price, timing and real-world results are not yet public.
Otto, a new customer-tracking app for field sales teams, lets reps speak their notes instead of typing them. Price and launch details are not yet known.
Read more: Otto launches a sales app you talk to instead of typingAI researcher Andrej Karpathy built a video site in one weekend with Claude. He says it beats YouTube's own picks, and it hints at how fast small software projects can move.
Read more: Karpathy built a video site in one weekend with Claude's helpVisa tested 22 AI models on realistic shopping tasks. They passed just over half, and often bought the wrong thing anyway.
Read more: Visa tested 22 AI shoppers, and most still make costly mistakes
