Wall St Engine(@wallstengine):$NVDA OPENS VERA RUBIN RACK DESIGNS, EXPANDS AI NETWORKING WITH META, GOOGLE AND CISCO NVIDIA announced new partnerships and open infrastructure contributions at the OCP Global Summit, covering rack designs, networking, software and power systems for next-generation AI data centers. The company contributed key Vera Rubin rack and compute tray designs through its MGX architecture. The updated MGX 1.2 specification supports racks up to 250kW, with standardized power connections, liquid cooling and fanless compute trays. $META is adopting NVIDIA’s Spectrum-6 Ethernet switches, which deliver 102.4Tb/s of switching capacity. Combined with Spectrum-X Multiplane technology, the architecture is designed to connect more than 100,000 GPUs with fewer switch layers. Traffic is distributed across up to eight independent network paths, allowing workloads to keep running if a component fails. Cisco $CSCO and NVIDIA announced plans for USFP, a new pluggable optical module designed for 400Gb/s-per-lane networking. The companies say it will deliver 4x the bandwidth density and approximately 1.3x the power efficiency of 200Gb/s-per-lane modules, with up to 12.8Tb/s of overall capacity in roughly the same physical footprint. TE Connectivity and Amphenol are helping develop USFP connectors, cages and modules. The companies plan to establish an open industry specification, giving multiple suppliers a common design to build around. Google’s Falcon Transport will be supported on future NVIDIA ConnectX-10 SuperNICs and BlueField DPUs. The collaboration enables NVIDIA and Google networking hardware to communicate over a common transport, with features for managing congestion, recovering dropped packets and protecting connections. NVIDIA also released new open-source DSX software components for configuring networks, stress-testing GPU clusters, managing racks and preparing hardware for deployment. Its Vera CPU rack design specifically targets AI agent workloads, including tool calling, orchestration and sandboxed execution. On power infrastructure, NVIDIA joined AMD, Google and Meta on an open data-center design specification, and Google, Meta and Microsoft on grid-resilience standards. Separately, Visionbay introduced plans for K-1, an 800V DC AI factory platform combining NVIDIA’s reference architecture with Vertiv’s power and cooling technologies for future Vera Rubin deployments.

2026.10.12 15:26

$NVDA OPENS VERA RUBIN RACK DESIGNS, EXPANDS AI NETWORKING WITH META, GOOGLE AND CISCO NVIDIA announced new partnerships and open infrastructure contributions at the OCP Global Summit, covering rack designs, networking, software and power systems for next-generation AI data centers. The company contributed key Vera Rubin rack and compute tray designs through its MGX architecture. The updated MGX 1.2 specification supports racks up to 250kW, with standardized power connections, liquid cooling and fanless compute trays. $META is adopting NVIDIA’s Spectrum-6 Ethernet switches, which deliver 102.4Tb/s of switching capacity. Combined with Spectrum-X Multiplane technology, the architecture is designed to connect more than 100,000 GPUs with fewer switch layers. Traffic is distributed across up to eight independent network paths, allowing workloads to keep running if a component fails. Cisco $CSCO and NVIDIA announced plans for USFP, a new pluggable optical module designed for 400Gb/s-per-lane networking. The companies say it will deliver 4x the bandwidth density and approximately 1.3x the power efficiency of 200Gb/s-per-lane modules, with up to 12.8Tb/s of overall capacity in roughly the same physical footprint. TE Connectivity and Amphenol are helping develop USFP connectors, cages and modules. The companies plan to establish an open industry specification, giving multiple suppliers a common design to build around. Google’s Falcon Transport will be supported on future NVIDIA ConnectX-10 SuperNICs and BlueField DPUs. The collaboration enables NVIDIA and Google networking hardware to communicate over a common transport, with features for managing congestion, recovering dropped packets and protecting connections. NVIDIA also released new open-source DSX software components for configuring networks, stress-testing GPU clusters, managing racks and preparing hardware for deployment. Its Vera CPU rack design specifically targets AI agent workloads, including tool calling, orchestration and sandboxed execution. On power infrastructure, NVIDIA joined AMD, Google and Meta on an open data-center design specification, and Google, Meta and Microsoft on grid-resilience standards. Separately, Visionbay introduced plans for K-1, an 800V DC AI factory platform combining NVIDIA’s reference architecture with Vertiv’s power and cooling technologies for future Vera Rubin deployments.