Israel-built Nvidia switch heads to the world's largest AI data centers
Microsoft, xAI, Tesla and Nebius are among the first companies deploying the next-generation networking technology.
The world's largest AI infrastructure operators, including Microsoft, xAI, Tesla, CoreWeave and Nebius, will deploy Nvidia's latest networking switch, Spectrum-X6, developed at the company's Israeli R&D center, in their data centers, the chip giant announced on Tuesday.
The switch, unveiled as part of Nvidia's Vera Rubin platform, delivers data transfer speeds of 102.4 terabits per second, twice the bandwidth of the previous generation.
"Spectrum-6 anchors the next generation of the NVIDIA Spectrum-X Ethernet platform, delivering the bandwidth, scale and intelligence needed to operate an AI factory as one end-to-end computing system," the company said.
Training and running large AI models requires thousands of processors to work in lockstep, constantly exchanging vast amounts of data. High-speed networking has therefore become as critical as the AI chips themselves, ensuring that processors remain synchronized and that data moves efficiently across increasingly large clusters.
Traditional Ethernet switches were designed to manage traffic between users, servers and storage systems, not the complex, synchronized communication required between thousands of AI processors. Nvidia's Spectrum networking family, developed in Israel, was built specifically to address this challenge, enabling ultra-fast communication between processors inside AI data centers.
According to Nvidia, Spectrum-X6 enables larger groups of AI processors to operate as a single computing system, improving utilization of computing resources while reducing bottlenecks. The result is faster model training and inference, as well as greater efficiency at scale.
“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.”
