NVIDIA also announced that CoreWeave, Microsoft, Nebius, SpaceXAI and Tesla will be among the first organizations to deploy Spectrum-6 in their AI factories. The company said the technology is intended to allow more computing resources to function as a unified system, helping customers train AI models and deliver inference services more efficiently.
“CoreWeave is built for the most demanding AI workloads, and networking is central to delivering that performance at scale,” said Min Jun, director of product for networking at CoreWeave. “Bringing NVIDIA Spectrum-6 and liquid-cooled Spectrum-X Ethernet infrastructure into our AI factories will help us deliver the bandwidth, resilience and efficiency customers need to train frontier models and deploy inference faster.”
“At gigascale, performance comes down to coordination: keeping every GPU in lockstep so one slow link doesn’t stall an entire job,” said Laurelle Roseman, vice president of global partnerships at Nebius. “That’s what NVIDIA Spectrum-6 goes after, and why we brought it in early — a fabric that stays fast and resilient as our customers’ most demanding workloads scale.”
According to NVIDIA, CoreWeave, Microsoft and Nebius will be among the first cloud providers to deploy Vera Rubin-based infrastructure featuring Spectrum-6, expanding access to the platform for developers, startups and enterprise customers.
The company said AI networking has become increasingly important because large-scale training and inference workloads rely on continuous communication between thousands of accelerators. These workloads generate significant east-west network traffic as GPUs synchronize data across large computing clusters, creating demands that traditional Ethernet networks were not originally designed to handle.
Spectrum-6 combines with the NVIDIA ConnectX-9 SuperNIC to form the next generation of Spectrum-X Ethernet. As part of the Vera Rubin platform, it integrates with the NVIDIA Vera CPU, Rubin GPU, NVLink 6 Switch, BlueField-4 DPU and ConnectX-9 SuperNIC into a unified AI infrastructure.
NVIDIA said Spectrum-6 supports both pluggable and co-packaged optics, along with liquid cooling designed to improve power efficiency across AI data centers. The company also said Spectrum-X includes traffic optimization capabilities, intelligent load balancing, rapid failure bypass and data recovery features, while maintaining support for standard Ethernet, open network operating systems, open protocols and multiple RDMA transport models.
The company said the platform delivers up to 1.6 times higher AI networking performance than standard Ethernet deployments and can sustain up to 95% network efficiency across installations exceeding 100,000 GPUs. NVIDIA also said hardware-accelerated Spectrum-X multiplane topologies reduce the number of switches required in AI data centers by 1.7 times, while Spectrum-X Ethernet Photonics provides up to five times higher power efficiency and improves mean time between incidents by up to tenfold.
This analysis is based on reporting from NVIDIA.
Image courtesy of NVIDIA.
This article was generated with AI assistance and reviewed for accuracy and quality.
