NVIDIA Unveils Spectrum-XGS Ethernet for Giga-Scale AI Super-Factories



Lawrence Jengar
Aug 22, 2025 12:55

NVIDIA has launched Spectrum-XGS Ethernet, enabling distributed data centers to unify into giga-scale AI super-factories, enhancing scalability and performance across vast distances.



NVIDIA Unveils Spectrum-XGS Ethernet for Giga-Scale AI Super-Factories

NVIDIA has introduced its latest innovation, the Spectrum-XGS Ethernet, a groundbreaking technology aimed at transforming distributed data centers into unified, giga-scale AI super-factories. This announcement, made at the Hot Chips conference, marks a significant leap in networking technology, promising to overcome the limitations of conventional Ethernet infrastructures that hinder scalability and performance.

Revolutionizing AI Infrastructure

As the demand for artificial intelligence continues to surge, individual data centers are reaching their limits in terms of power and capacity. NVIDIA’s Spectrum-XGS Ethernet addresses this challenge by enabling data centers to scale beyond the constraints of a single facility. This technology integrates with the existing NVIDIA Spectrum-X™ Ethernet platform, introducing a scale-across infrastructure that connects multiple, geographically distributed data centers into a cohesive AI super-factory.

Jensen Huang, founder and CEO of NVIDIA, emphasized the significance of this development, stating, “The AI industrial revolution is here, and giant-scale AI factories are the essential infrastructure.” Spectrum-XGS Ethernet adds a new dimension to data center capabilities by linking facilities across cities, nations, and continents, thereby facilitating the creation of vast AI super-factories.

Enhanced Performance and Connectivity

The Spectrum-XGS Ethernet is fully integrated into the Spectrum-X platform and features advanced algorithms that dynamically adapt to the distances between data center facilities. This integration enhances performance through auto-adjusted distance congestion control, precision latency management, and end-to-end telemetry. Consequently, the technology nearly doubles the performance of the NVIDIA Collective Communications Library, optimizing communication across multi-GPU and multi-node setups.

CoreWeave, a pioneer in hyperscale AI infrastructure, is among the first to adopt Spectrum-XGS Ethernet. Peter Salanki, cofounder and CTO of CoreWeave, highlighted the benefits, stating, “With NVIDIA Spectrum-XGS, we can connect our data centers into a single, unified supercomputer, providing our customers access to giga-scale AI that will accelerate breakthroughs across every industry.”

Innovative Networking Solutions

The Spectrum-X Ethernet networking platform offers 1.6 times greater bandwidth density compared to standard Ethernet, making it ideal for hyperscale AI factories. The platform includes NVIDIA Spectrum-X switches and NVIDIA ConnectX®-8 SuperNICs, ensuring seamless scalability, ultralow latency, and impressive performance for enterprises building the future of AI.

This latest announcement follows a series of networking innovations from NVIDIA, including Spectrum-X and Quantum-X silicon photonics networking switches, designed to connect millions of GPUs across sites while minimizing energy consumption and operational costs.

For those interested in exploring the capabilities of Spectrum-XGS Ethernet, further information is available at the NVIDIA Newsroom.

Image source: Shutterstock


Source: https://blockchain.news/news/nvidia-spectrum-xgs-ethernet-giga-scale-ai