Meta to Build Vera Rubin AI Factory with NVIDIA Spectrum-6 Ethernet Switches
At the Open Compute Project (OCP) Global Summit on Oct 12, NVIDIA announced that Meta will use its Spectrum-6 102.4Tb/s Ethernet switches to build a Vera Rubin AI factory based on an open network. The Spectrum-6 delivers twice the switching capacity and per-lane bandwidth of the Spectrum-4, and with Spectrum-X Multiplane technology can support clusters of more than 100,000 GPUs.
The announcement was part of NVIDIA's broader open infrastructure push at the summit. NVIDIA said it is contributing the Vera Rubin rack and compute tray designs to OCP, and updated its MGX 1.2 specification with new rack-level blind-mate liquid cooling, interoperable rack interconnects, liquid-cooled 5000A busbars and native 19-inch equipment support, enabling racks up to 250kW. It also released four open-source software components for DSX AI Factory covering switch configuration, GPU cluster stress testing, rack firmware management and bare-metal deployment, which NVIDIA said can help cloud providers turn hardware into production clusters within hours.
NVIDIA also announced the Open Data Center Specification v0.7 with AMD, Google and Meta to unify power, cooling and physical layout designs for AI factories, and a grid resilience specification with Google, Meta and Microsoft requiring large AI facilities to stay connected during grid disturbances and recover to near full load in under two seconds. With Cisco, NVIDIA introduced the USFP standard for 400Gb/s network architectures, a pluggable optical module offering 16 full-duplex lanes in a form factor close to OSFP, with four times the bandwidth density of traditional 200Gb/s single-lane pluggable modules and about 1.3 times better energy efficiency.
