What happened
At the OCP Global Summit, NVIDIA announced the contribution of its Vera Rubin rack and tray designs to the Open Compute Project (OCP) via the MGX architecture. The company also released open-source components for its DSX AI factory platform and co-published new open specifications for data center facility design and grid resilience with AMD, Google, and Meta.
NVIDIA announced at the Open Project (OCP) Global Summit that it is contributing key Vera Rubin rack and compute tray designs to the OCP through the open NVIDIA MGX architecture. This contribution provides system builders with a common foundation for developing next-generation AI factory infrastructure, extending the open architecture with rack-scale blind-mate liquid cooling, interoperable interconnects, and a liquid-cooled 5,000A busbar designed for higher-power systems.
The updated MGX 1.2 specification supports rack deployments up to 250kW and introduces a Vera CPU rack design intended for the agentic AI era. This specific rack is designed to support tool calling, orchestration, and sandboxed execution, allowing AI agents to interact with applications and data in real time, distinct from the Vera Rubin NVL72 which focuses on large-scale training and .
NVIDIA also expanded its DSX AI factory platform by releasing open-source components for configuring networks, validating clusters, managing racks, and provisioning bare metal. These components, developed from NVIDIA’s experience running DGX Cloud, are available on GitHub and enable cloud providers and software partners to integrate AI factory capabilities directly into their multi-tenant platforms.
In a collaborative effort, NVIDIA joined AMD, Google, and Meta to publish the Open Data Center Specification v0.7, a unified blueprint for next-generation AI factory facility design covering power architecture, cooling, and physical layout. Additionally, NVIDIA, Google, Meta, and Microsoft published the OCP Grid Resilience Specification, which defines how large-scale AI infrastructure should respond to utility grid disturbances, aiming to restore near-full load within less than two seconds.
Source details: blogs.nvidia.com ↗
Why it matters
This move shifts AI infrastructure from proprietary, vendor-locked systems to open, interoperable standards. By publishing the Vera Rubin design and DSX OS components, NVIDIA enables third-party manufacturers and cloud providers to build compatible AI factories, potentially accelerating global deployment and reducing integration costs for operators scaling to gigawatt-level facilities.
The publication of these open standards addresses a critical bottleneck in AI scaling: the lack of a common foundation for hardware and facility design. By making the Vera Rubin architecture and DSX OS components open, NVIDIA allows the broader ecosystem to build, manufacture, and scale AI infrastructure without waiting for proprietary specifications from individual customers.
This shift to open standards is particularly significant for the transition from megawatt to gigawatt-scale AI factories. The new specifications provide a validated, deployable implementation for facility layout and power architecture, enabling equipment makers and data center operators to design and deliver infrastructure that is interoperable and efficient.
The inclusion of the Vera CPU rack design signals a strategic focus on the infrastructure requirements for agentic AI, which demands real-time interaction with external tools and services. By standardizing this aspect, NVIDIA is helping to define the hardware baseline for the next generation of AI applications that rely on complex, multi-step agent workflows.
Interactive Mechanism: How It Actually Works
Explore the underlying technology behind this development interactively.
crm_get_transaction(id='4092').Which component of an AI application is the machine-learning model itself?
What to watch next
Monitor the adoption of the MGX 1.2 specification by non-NVIDIA hardware partners and the practical implementation of the 800 VDC power architecture in existing data centers. Watch for the release of the USFP optical transceiver standard and its impact on networking costs for large-scale AI clusters.
Track the adoption of the MGX 1.2 specification by third-party system builders and whether it leads to a broader ecosystem of compatible AI factory hardware beyond NVIDIA’s own products.
Observe the implementation of the 800 VDC power architecture, particularly through partners like Visionbay and Vertiv, to see how effectively it addresses the challenges of high-density power distribution in existing data center facilities.
Monitor the development and adoption of the USFP optical transceiver form factor, announced with Cisco, as it aims to improve bandwidth density and power efficiency for next-generation AI fabrics.
Watch for the practical application of the OCP Grid Resilience Specification in real-world deployments to determine if the sub-two-second recovery time for grid disturbances is achievable at scale.