NVIDIA Opens Vera Rubin Architecture at OCP Summit to Scale AI Factories
NVIDIA shares rack designs and networking standards at OCP Global Summit to accelerate global deployment of next-generation AI infrastructure systems.
Key points
- NVIDIA contributed Vera Rubin rack designs to the OCP MGX architecture for open AI factory infrastructure.
- New partnerships announced include Cisco USFP standards and Meta Spectrum-X networking innovations.
- The MGX 1.2 specification supports rack deployments up to 250kW with blind-mate liquid cooling.
NVIDIA announced new partnerships and open architecture contributions at the Open Compute Project Global Summit in San Jose. The company shared key Vera Rubin rack and compute tray designs through the MGX architecture. This move aims to provide a common foundation for system builders developing next-generation AI factory infrastructure worldwide.
Strategic Context
The drive to scale AI factories from megawatts to gigawatts requires more than advanced computing hardware, according to NVIDIA blog posts. Open standards are helping the industry build infrastructure faster and more reliably. Ian Buck, vice president of hyperscale computing at NVIDIA, stated that open infrastructure is a critical ingredient for scaling these facilities globally.
Partnership Announcements
NVIDIA and partners published open standards for AI factory design and grid resilience. Meta and NVIDIA advanced open AI networking with Spectrum-X Ethernet SuperNIC and Spectrum-6 innovations. Visionbay introduced an 800V DC AI factory design for the Vera Rubin era. These efforts span the utility grid, rack-scale architecture and the AI factory itself.
Technical Specifications
The updated MGX 1.2 specification extends the open architecture with rack-scale blind-mate liquid cooling. It includes a liquid-cooled 5,000A busbar for higher-power systems and native 19-inch equipment support. The architecture supports rack deployments up to 250kW. It also adds configurable I/O bays and fanless liquid-cooled compute trays to improve modularity.
New Standards and Hardware
NVIDIA and Cisco announced the Ultra-Dense Small Form Factor Pluggable standard for 400Gb/s-based networking fabrics. NVIDIA and Google announced Falcon Transport support on future ConnectX-10 SuperNICs and BlueField DPUs. The Vera CPU rack design extends the MGX architecture beyond accelerated computing. It provides a platform for tool calling and orchestration in the agentic AI era.
What to do and how to stay safe: NVIDIA
- Review current data center power capacity plans against new 250kW rack deployment standards.
- Evaluate compatibility of existing cooling infrastructure with blind-mate liquid cooling requirements.
- Assess networking hardware readiness for 400Gb/s USFP module standards and Spectrum-X innovations.
- Monitor OCP publications for updates to MGX 1.2 specifications and interoperable interconnect designs.
Step-by-step guide: Prevent Prompt Injection Attacks: Practical Defences for AI Systems
General security guidance from the TechSec Observer newsroom. It is not confirmed advice from the organisations named in this story.
Frequently asked questions
What is the MGX 1.2 specification?
It is an updated open architecture for AI factories that adds blind-mate liquid cooling, 5,000A busbars and support for 250kW rack deployments.
Which companies partnered with NVIDIA at the OCP Summit?
Partners included Meta, Cisco, Google, Visionbay and various hyperscalers who published open standards for AI factory design and grid resilience.
Does the Vera CPU rack support accelerated computing?
No, the Vera CPU rack extends the architecture beyond accelerated computing to support tool calling, orchestration and sandboxed execution for AI agents.
