NVIDIA and NetBox Labs Partner to Accelerate AI Data Center Deployment and Operations

For the past year, we’ve been investing in building capabilities in NetBox for one of the most demanding environments in infrastructure: the AI datacenter. In November, we launched the NetBox AI Datacenter Acceleration Program to work directly with the operators building at hyperscale. That work shaped what we built next: TurboBulk, which moves data into and out of NetBox more than 100x faster than the REST API, and Asset Lifecycle, which brings procurement into the system of record so the bill of materials follows the design. Customers like CoreWeave, have shown us what’s at stake: deploying a datacenter even a month sooner goes straight to revenue. Today we’re taking the next step.

NetBox is now a native system of record for NVIDIA DSX OS and the open source NVIDIA Configuration Manager, enabling neocloud and AI factory builders to design, deploy, operate, and evolve infrastructure at 100MW+ scale. 

We speak daily to AI factories tasked with meeting the surging demand for their infrastructure and the outcome they are most focused on each time is compressing time to first token, the time between committing to a new build and that capacity serving its first token to a customer. These organizations are deploying more infrastructure in a quarter than most enterprises deploy in a decade, and they can’t afford to do it from spreadsheets. We are talking designs that span thousands of switches across multiple fabrics, hundreds of thousands of interfaces and cables, and hundreds of megawatts of capacity. The issue is these mega datacenters are designed, procured, deployed, and revised on timelines that traditional data center practice was never built for. And the difference between getting to that first token quickly or not increasingly comes down to a question of their system of record.

NetBox is how today’s AI infrastructure leaders answer that question. NetBox Labs customers including CoreWeave run the NetBox Labs platform at the heart of their infrastructure operations, and NetBox is used by the majority of the industry’s top-rated GPU cloud providers. The NetBox Labs infrastructure intelligence platform builds on that foundation to accelerate the full AI data center lifecycle  design, procurement, deployment, and ongoing operations — compressing deployments, reducing time to first token, and accelerating revenue.

NetBox: a native system of record for NVIDIA DSX OS

NVIDIA DSX OS is the open, modular operations software portfolio of the NVIDIA DSX platform — NVIDIA’s playbook for designing, simulating, building, and operating gigawatt-scale AI Factories. DSX OS spans infrastructure lifecycle management, runtime consistency, configuration validation, multi-tenancy, network orchestration, health, and remediation, delivered as open source and designed for incremental adoption alongside existing stacks. Its network orchestration component, the open source NVIDIA Config Manager (NVCM), bootstraps and operates AI Factory network infrastructure — combining event-driven configuration rendering, DHCP, zero-touch provisioning, versioned configuration storage, and durable day-2 workflows across NVIDIA Spectrum-X Ethernet, NVIDIA Quantum InfiniBand, and NVLink fabrics.

NVIDIA and NetBox Labs closely collaborated to make NetBox a native system of record for DSX OS, delivering NetBox native configuration management for NVCM to model AI-fabric intent:

• Fabric intent lives in the system of record. NVCM reads physical inventory and fabric intent — devices, interfaces, cabling, IPAM, overlays, partitions, domains — directly from NetBox, and renders, provisions, validates, and tracks device configurations from that model.
• Change flows automatically. When infrastructure changes in NetBox, NVCM is notified, determines exactly which devices are affected, and reconciles fabric configuration against documented intent — no swivel-chair, no drift between what was designed and what is deployed.
• The whole lifecycle connects. Because the same NetBox model drives design, deployment, and day-2 operations, teams can bring up massive fabrics from a validated blueprint, operate them against documented intent, and evolve them generation over generation.

The integration is open source: the NetBox provider and companion plugin developed by NetBox Labs are available under Apache-2.0, and NVCM is available from NVIDIA on GitHub. NVIDIA will present the integration in a technical session at NetBox Evolve.

We deeply value this partnership with NVIDIA. Making NetBox a native system of record for DSX OS means our joint customers and our community get one authoritative model of their infrastructure, from first rack elevation to production fabric — and that’s how you compress time to first token.