As AI factories grow from megawatts to gigawatts, building them takes more than advanced computing. At the Open Compute Project (OCP) Global Summit in San Jose, NVIDIA announced new partnerships and contributions meant to create a common foundation for next-generation AI infrastructure, with open standards helping the industry scale AI infrastructure faster, more efficiently and more reliably. NVIDIA contributed key NVIDIA Vera Rubin rack and compute tray designs to OCP through the open NVIDIA MGX architecture, giving system builders a shared base for developing next-generation AI factory infrastructure. NVIDIA says these efforts span the AI factory itself, the utility grid that powers it and the rack-scale architecture at the heart of modern accelerated computing.

What Was Announced

At the event, NVIDIA and its partners made several announcements. NVIDIA and hyperscalers published open standards for next-generation AI factories, including AI factory design standards and grid resilience. NVIDIA also launched new open source software components for the NVIDIA DSX AI Factory platform. Meta and NVIDIA advanced open AI networking with NVIDIA Spectrum-X Ethernet SuperNIC and Spectrum-6 innovations. Visionbay introduced an 800V DC AI factory design for the Vera Rubin era. NVIDIA and Cisco announced the Ultra-Dense Small Form Factor Pluggable (USFP) standard, which the companies describe as the industry's first module for 400Gb/s-based networking fabrics. Finally, NVIDIA and Google announced Falcon Transport support on future NVIDIA ConnectX-10 SuperNICs and BlueField DPUs.

Ian Buck, vice president of hyperscale and high-performance computing at NVIDIA, discussed before a packed audience how open infrastructure is becoming a critical ingredient for scaling AI factories worldwide. NVIDIA frames the news as evidence that its “extreme codesign” combined with open standards is accelerating deployment, improving power efficiency and driving tokens-per-watt performance.

Third-Generation MGX and Open Data Center Design

NVIDIA joined industry leaders in contributing to the third-generation MGX AI data center reference design through OCP's Open DC Initiative. Combined with open liquid-cooling technologies such as MQD and MQDB, the effort is intended to turn proven AI factory designs into industry standards that others can adopt, manufacture and scale.

For Vera Rubin AI factories, the updated MGX 1.2 specification extends the open architecture with rack-scale blind-mate liquid cooling, interoperable rack-scale interconnect infrastructure and a liquid-cooled 5,000A busbar designed for higher-power AI systems. It adds native 19-inch equipment support and higher-density rack designs. MGX 1.2 also expands modularity with configurable I/O bays, fanless liquid-cooled compute trays, standardized power and cooling interfaces, and a common infrastructure that helps ecosystem partners build interoperable accelerated computing platforms. According to NVIDIA, it improves assembly, test and service times to streamline manufacturing and field service, enabling faster and more efficient infrastructure buildout. The architecture supports rack deployments up to 250kW while simplifying serviceability, deployment and infrastructure reuse.

The Vera CPU rack design specification extends MGX beyond accelerated computing into the agentic AI era. While Vera Rubin NVL72 serves as the foundation for large-scale training and inference, the NVIDIA Vera CPU rack is designed to support tool calling, orchestration and sandboxed execution, helping AI agents interact with applications, data and services in real time. Built on the third-generation MGX rack architecture, it shares the same rack-scale design principles that unite compute, networking, power and cooling into a common modular platform, letting partners deploy specialized infrastructure on a shared ecosystem.

NVIDIA DSX AI Factory and Open Source Software

The NVIDIA DSX AI factory platform applies extreme codesign to unify computing, networking, power, cooling, facilities and software, so AI factories operate as a single optimized system that maximizes AI output within available infrastructure constraints. At its center is DSX OS, the open operating layer that turns the DSX architecture into coordinated, repeatable operations at scale.

At OCP, NVIDIA expanded DSX OS with open source components for configuring networks, validating clusters, managing racks and provisioning bare metal:

NVIDIA Config Manager: Automates switch configurations at scale, allowing customers to deploy, validate and maintain consistent AI factory network fabrics with minimal to zero manual intervention.
NVIDIA Cluster Readiness Engine: Helps eliminate costly job crashes by stress-testing GPU clusters with real workload simulations to guarantee hardware health before training begins.
NVIDIA Rack Management Service: Offers a rack hardware abstraction and operations solution that treats racks as first-class units for improved AI factory scale firmware management.
Cluster API Provider for NVIDIA Infrastructure Controller: Accelerates bare-metal deployment and tenant onboarding, turning raw AI hardware into production-ready clusters in hours instead of days.

NVIDIA says these components were developed from its experience running NVIDIA DGX Cloud, and they let cloud providers and software partners integrate AI factory capabilities directly into their multi-tenant platforms. Developers can explore and contribute to both new and existing components through the DSX AI Factory GitHub page.

Open Data Center and Grid Resilience Standards

At OCP, NVIDIA joined AMD, Google and Meta to publish the Open Data Center Specification v0.7, which defines a unified blueprint for next-generation AI factory facility design, from power architecture to cooling infrastructure to physical layout. Separately, the OCP Grid Resilience Specification, published by Google, Meta, Microsoft and NVIDIA, sets out how large-scale AI infrastructure should behave during utility grid disturbances: staying connected, responding automatically and restoring to near-full load within less than two seconds.

With these standards open, manufacturers have a common target to build for. Equipment makers, colocation providers and data center operators can design, fabricate and deliver the infrastructure layer for AI factories under construction without waiting for proprietary specifications from individual customers. Because the specifications are compatible with the NVIDIA DSX reference design, NVIDIA says they provide a validated, deployable implementation of both specs, from facility layout and power architecture to the software that monitors and responds to grid signals in real time.

Visionbay's 800 VDC K-1 Platform

Visionbay unveiled plans for K-1, a next-generation AI factory platform built around an 800 VDC architecture and designed with future NVIDIA Vera Rubin deployments in mind. The project combines NVIDIA's 800 VDC reference architecture with power and thermal technologies from Vertiv. As AI training, inference and reasoning workloads expand, rack power densities keep rising, and traditional low-voltage distribution becomes harder to scale, requiring more copper, more cabling and more power-conversion equipment.

K-1 is designed to phase in 800 VDC distribution where it delivers the greatest benefit, preserving flexibility for existing deployments rather than requiring a complete transition on day one. Its architecture uses cluster-level power centers that centralize AC-to-DC conversion and distribute 800 VDC closer to compute resources, which simplifies downstream power delivery, supports modular expansion and creates repeatable building blocks for future growth. The platform also supports liquid-cooled infrastructure and scalable power distribution.

Meta and NVIDIA Open Networking

Meta is adopting NVIDIA Spectrum-6 102.4Tb/s Ethernet switches to build larger, higher-performance Vera Rubin AI factories on open networking. NVIDIA says Spectrum-6 delivers twice the switching capacity and per-lane bandwidth of Spectrum-4, and provides a lossless Ethernet AI fabric at 200G per lane, with high effective bandwidth and performance isolation that keeps one job's traffic from slowing another's.

Meta's Spectrum-6-based switch, N-IceCube, slots into IcePack, Meta's 100T liquid-cooled rack. N-IceCube runs the Facebook Open Switching System (FBOSS), Meta's open source network operating system, contributed to OCP to support efficient networking at hyperscale. FBOSS is built on the Switch Abstraction Interface, an OCP standard NVIDIA helped develop that provides a vendor-independent way of controlling switch hardware. Together, these technologies support agentic AI factories of more than 100,000 GPUs that train and serve large mixture-of-experts models.

Key Features at a Glance

MGX 1.2 rack density: Supports rack deployments up to 250kW – higher power per rack without redesigning the whole facility.
Liquid-cooled 5,000A busbar: Designed for higher-power AI systems – carries the current that dense Vera Rubin racks demand.
Vera CPU rack: Built on third-generation MGX, supports tool calling, orchestration and sandboxed execution – gives agentic AI workloads their own purpose-built platform.
Grid resilience spec: Restores to near-full load within less than two seconds – helps AI campuses ride through utility disturbances.
Spectrum-6 switches: 102.4Tb/s Ethernet with 200G per lane – twice the switching capacity and per-lane bandwidth of Spectrum-4.
Open source DSX OS components: Config Manager, Cluster Readiness Engine, Rack Management Service and Cluster API Provider – automation for networks, cluster testing, rack firmware and bare-metal provisioning.
USFP module: Industry's first module for 400Gb/s-based networking fabrics, co-announced with Cisco – a shared form factor for next-generation fabrics.
K-1 phased 800 VDC: Converts AC to DC at cluster-level power centers – lets operators adopt higher-voltage distribution incrementally.

Availability

The article does not give ship dates or pricing for these items. The Vera Rubin-related designs, the Visionbay K-1 platform and the ConnectX-10 and BlueField support are described as plans or future products. The MGX 1.2 specification, the Vera CPU rack specification, the Open Data Center Specification v0.7 and the OCP Grid Resilience Specification are presented as published or open standards, and the DSX OS open source components are available through the GitHub repositories linked in the announcement.

Source: blogs.nvidia.com