THE APEX TIMES
NVIDIA launches Minerva, an AIOps platform aimed at making “AI factories” easier to run
The new NVIDIA Minerva platform is designed to connect infrastructure and workload telemetry, detect anomalies, and help teams maintain GPU clusters that train and serve artificial intelligence at massive scale.
NVIDIA on Thursday unveiled Minerva, a new AIOps platform positioned to deliver end-to-end observability across the systems behind large-scale artificial intelligence training and inference, which the company calls “AI factories.” The company says Minerva is built to help operators find problems earlier, reduce the need for manual troubleshooting, and keep AI workloads productive as GPU clusters grow larger and more complex.
In NVIDIA’s description, AI factories run vast numbers of GPUs across high-speed networks to both train models and serve AI applications. At that scale, a single degraded component can slow an entire job, and the early warning signs are often scattered across different monitoring tools or hidden inside subtle performance changes. NVIDIA highlights examples such as a GPU running hotter than its peers or an NVMe drive losing performance under load.
Minerva is designed to focus on AI factory “yield,” a set of operational measures intended to represent how much useful work infrastructure produces through GPU hours, goodput, cluster uptime, and token production. NVIDIA says traditional monitoring setups tend to divide a data center into separate views, with networking, compute, reliability, and machine learning teams each seeing only part of the picture.
The platform’s core aim is to connect those indicates across NVIDIA hardware and the software layers that run jobs. NVIDIA says Minerva gathers telemetry from infrastructure components and AI workloads, then applies purpose-built AI models to translate raw data into timely, actionable insight for operators. It also describes the platform as learning the normal operating behavior of a given factory so it can detect anomalies, predict emerging failures, and correlate related events.
To provide the context needed for troubleshooting, Minerva builds what NVIDIA calls a dynamic knowledge graph. NVIDIA says the graph logically maps the factory’s infrastructure, including GPUs, SuperNICs, switches, and cables, and ties that to the jobs running on the same systems. With that context, the company argues operators can trace a slow training job back to the component that is affecting it, even when a factory has hundreds of thousands of GPUs.
NVIDIA says Minerva analyzes billions of data points per hour using supervised and unsupervised machine learning, then groups what it calls “event storms” into focused incidents with likely root causes. The platform includes a native AI assistant and can connect to external agentic AI tools, giving operators a natural-language interface grounded in live telemetry.
The assistant is described as enabling teams to investigate cluster health through conversation and receive recommended remediation steps, while still letting organizations decide whether a person stays in the loop to approve and implement fixes or whether Minerva can trigger automation through infrastructure application programming interfaces and external systems. NVIDIA frames the approach as a way to move from reactive alert handling to more proactive maintenance, improving GPU uptime.
NVIDIA did not disclose pricing, deployment timelines, or technical requirements in the announcement, but it did name multiple customers and partners. The company said IREN, one of its stated partners, is already using Minerva to optimize its AI factories. NVIDIA quoted IREN’s chief technology officer, Denis Skrinnikoff, saying that as IREN scales its AI factories, the key to delivering for customers is maximizing GPU productivity, and that Minerva gives IREN’s operations team one view across network, GPUs, and jobs to shift toward proactive maintenance.
Beyond IREN, NVIDIA said Aolani, Bitdeer AI, IBM, Jio, Lambda, Quake AI, and Tesla plan to deploy Minerva. It also said Dell Technologies, Lenovo, and Supermicro will offer Minerva as part of their AI factory operations portfolios. What remains unclear is the scope of rollout and whether the named deployments represent pilots, broad production use, or specific configurations for different environments.
Why It Matters
- As AI training and serving clusters scale, the operational burden of debugging performance issues can become a major bottleneck, and NVIDIA is positioning Minerva to reduce that load.
- By correlating infrastructure indicates with workload outcomes, Minerva targets a common gap between monitoring tools used by different teams in large data centers.
- If the platform’s incident grouping and anomaly prediction work as described, it could help organizations shift from alert overload to more proactive maintenance and better GPU utilization.
- NVIDIA’s approach also reflects a broader trend toward AI-driven operations, where observability systems incorporate machine learning models and conversational interfaces.
Key Facts
- NVIDIA launched Minerva, an AIOps platform intended to provide end-to-end observability for AI factories.
- Minerva connects telemetry from infrastructure and AI workloads, then applies AI models to produce “timely, actionable insight.”
- The platform targets AI factory “yield,” including GPU hours, goodput, cluster uptime, and token production.
- NVIDIA says Minerva builds a dynamic knowledge graph that maps infrastructure components and the jobs running on them to speed root-cause tracing.
- Minerva analyzes billions of data points per hour and can group related alert events into focused incidents.
- The platform includes a native AI assistant with natural-language investigation and remediation guidance, and it can connect to external agentic AI tools.
- NVIDIA named IREN as already using Minerva, and listed additional companies and hardware partners that plan to deploy or offer it.
Technology Related
Ken Griffin warns investors may be in an “AI bubble,” cites Amazon’s long-run rise
Citadel CEO Ken Griffin said markets appear to be in the middle of an AI bubble, drawing parallels to earlier speculative eras. He pointed to Amazon as an example of how lasting market leaders can emerge from prior booms, even as he cautioned that investors may be overextending now.
Meta’s planned stock split introduces non-voting shares, setting up new AI financing options
A reported share-structure change would preserve Mark Zuckerberg’s control while giving Meta flexibility to pursue major funding for its artificial intelligence expansion.
Meta gives no 2027 cost and capex outlook, Wedbush flags the gap
In the latest read-through from Wall Street, Meta Platforms did not provide forward projections for 2027 spending on costs or capital expenditures, leaving analysts to infer how the company is budgeting for its next phase of investment.
Broadcom’s AI-led momentum fuels a bold call for AVGO’s long-term upside
A new market forecast argues Broadcom’s surge in artificial-intelligence revenue could propel its stock far past the S&P 500 through 2030, even as shares lag the broader index so far in 2026.
3M expands collaboration with Amazon, creating GSEO partnership to scale Expanded Beam Optical R&D and manufacturing
The industrial company said the effort is aimed at expanding and scaling its Expanded Beam Optical product work, with an expected $40 million contribution by end-2026.
Palantir shares rise 7% in a month as Wall Street sets higher price targets
After a roughly month-long rebound, Palantir’s stock has drawn renewed attention from several Wall Street firms, which have outlined price targets ranging from $230 to $265, even as the broader Nasdaq 100 declined.
Waymo, Google and Microsoft alums back Maxinsights with $30 million seed round to build a “data layer” for physical AI
Maxinsights says it is using the fresh funding to accelerate the infrastructure it is building for AI systems that interact with the real world, and pitched the round as part of a broader push toward “physical AI”
Wells Fargo keeps Meta at Overweight, trims its price target to $995
The bank maintained its Overweight rating on Meta but lowered its target after flagging uncertainty around the timing and scale of a potential capital raise tied to new non-voting Class C shares.
Meta shares dragged by record-high “credit risk gauge” as investors weigh AI spending plans, Yahoo Finance reports
A Yahoo Finance market wrap said major U.S. indexes fell late Thursday, with Meta in focus after a credit-risk indicator hit a record high amid investor attention on the company’s artificial intelligence spending plans. In the same roundup, Tesla’s latest-quarter revenue was reported as missing consensus.
Intel Slips as Chip Sector Watches a Deeper AI Supply Warning
A fresh supply-warning reported in the chip supply chain is spreading concern beyond a narrow slice of artificial intelligence hardware, dragging down data-center-focused semiconductor names including Intel, which fell about 7% in the session.
