NVIDIA opened its Vera Rubin architecture to OCP to scale AI factories. It contributed MGX rack and compute tray designs for next-generation infrastructure. Partners like Cisco, Google, Meta, Microsoft and Visionbay joined the effort. NVIDIA and hyperscalers published open standards for AI factory design and grid resilience. NVIDIA launched open source software components for its DSX AI Factory platform. Meta and NVIDIA advanced open AI networking with Spectrum-6 and Multiplane technology. Visionbay introduced an 800V DC AI factory design for Vera Rubin. NVIDIA and Cisco announced the USFP standard for 400Gb/s networking fabrics. NVIDIA and Google announced Falcon Transport support on future SuperNICs. The MGX 1.2 specification extends liquid cooling and modularity for higher-power systems. DSX OS now includes tools for network configuration, cluster validation and rack management. Open Data Center Specification v0.7 defines a unified blueprint for AI factory facilities. OCP Grid Resilience Specification sets behavior standards for AI infrastructure during grid disturbances.
What changed
AI factory designs and standards were largely proprietary or vendor-specific.
What it unlocks
Building interoperable, high-density AI factories using open standards and modular components.
- 250kW rack deployments
- 800 VDC architecture
- 400Gb/s per lane networking
- 100,000+ GPUs in AI factories
What you need to act on it
- adoption of NVIDIA MGX architecture
- use of open OCP specifications
- compatibility with DSX reference design
Sources