NVIDIA Opens Vera Rubin Rack Designs and Partners on AI Standards
At the OCP Global Summit, NVIDIA announced the open-sourcing of its Vera Rubin rack and compute tray designs alongside updates to the MGX 1.2 specification, partnering with tech giants like Google, Meta, and Microsoft to establish open infrastructure standards. This move establishes unified open standards for next-generation AI factories, streamlining power, cooling, and hardware deployment across major cloud providers. By collaborating with industry leaders, NVIDIA is helping shape scalable, modular data center ecosystems. The MGX 1.2 specification introduces rack-level blind-mate liquid cooling, 5000A liquid-cooled busbars, and support for up to 250kW racks, while new open software components and network standards like Spectrum-6 Ethernet aim to support clusters exceeding 100,000 GPUs.
## BACKGROUND
The Open Compute Project (OCP) is a collaborative initiative founded in 2011 to share efficient data center hardware designs and industry best practices. NVIDIA's MGX is a modular server architecture designed to help system manufacturers quickly build customized accelerated computing systems for diverse enterprise workloads.