Rumble's Quake AI Turns to Nvidia's Minerva to Keep Its GPU Fleet Running
As AI data centers scale into the tens of thousands of GPUs, the hard problem is no longer buying compute — it's knowing when something is about to break.
The race to build AI infrastructure has a quieter second act: keeping it running. Quake AI, the AI infrastructure arm of RUM Group Inc. (Nasdaq: RUM), said on Tuesday it plans to deploy NVIDIA Minerva, a new operations platform built to spot trouble in AI data centers before it turns into downtime.
Watching the whole factory, not just the GPUs
According to the announcement, Minerva is an AI-native observability platform that pulls together telemetry from across an AI factory — GPUs, networking, storage, job schedulers, and the AI workloads themselves — and correlates those signals to flag anomalies, point to likely root causes, and warn of failures that may be forming. The pitch is a shift from reactive firefighting to proactive maintenance: catch the weak link early, fix it faster, and keep the factory productive.
That matters because the scale of modern AI infrastructure has outgrown manual oversight. A single site can hold tens of thousands of accelerators, and a failing interconnect can idle expensive hardware while engineers hunt for the cause. Minerva is designed to give operations teams a single view of that sprawling stack, so the first sign of trouble arrives with context attached.
Why Quake AI wants it
Quake AI sells bare-metal GPU infrastructure for AI training and inference — customers rent direct access to the hardware, without a virtualization layer in between. That appeals to buyers who want full control, but it raises the bar on reliability: the provider's job is to make sure the underlying infrastructure never becomes the bottleneck. NVIDIA had already named Quake AI among the companies planning to adopt Minerva in an announcement on October 22, and Tuesday's release confirms the intention from Quake AI's side.
RUM Group founder and CEO Chris Pavlovski framed the move as keeping a promise to customers, noting that NVIDIA hardware has powered the company's infrastructure for years, with more than 22,000 GPUs in operation today and deployments of NVIDIA's Vera Rubin platform planned for 2027.
A company straddling video and AI
Quake AI is one half of a company better known for its other business. RUM Group describes itself as an AI infrastructure and video company; its video platform, Rumble, was the original tenant of Quake AI's compute. The AI arm now operates data centers offering GPU and CPU compute, storage, and networking at scale, selling AI compute as a service.
The company pitches itself as building infrastructure for what it calls the agentic-first enterprise — the compute, cloud, and trust layers for an AI-driven future. Whether that framing lands with investors, the Minerva deployment is a concrete step in a familiar direction: as AI factories grow larger and more expensive, the companies running them are investing as much in watching the machines as in buying them.
Written by the BidAsk Newsroom. Based on reporting from the sources below.
