English 
NVIDIA announced that Meta and Oracle will deploy its Spectrum-X™ Ethernet networking switches to accelerate AI data center networks. Both companies aim to leverage the high-performance switches to build open, scalable network architectures that dramatically enhance AI training efficiency and speed up time to insights.
Oracle plans to construct a gigascale AI factory accelerated by NVIDIA’s Vera Rubin architecture and interconnected by Spectrum-X Ethernet. Meta will integrate Spectrum-X switches into its Facebook Open Switching System (FBOSS), a platform designed for managing and controlling large-scale network switches, enabling faster deployments and improved efficiency for AI workloads.
について エヌビディア Spectrum-X platform, which includes Spectrum-X Ethernet switches and Spectrum-X Ethernet SuperNICs, is purpose-built for the era of trillion-parameter models and generative AI. It allows hyperscalers to interconnect millions of GPUs with unprecedented efficiency and scale. Spectrum-X has demonstrated record data throughput, achieving 95% efficiency on the world’s largest AI supercomputer, compared with approximately 60% on conventional Ethernet due to flow collisions.
こちらもお読みください： ABC Corporation Launches Free Email Consultation for MCP Construction Ahead of AI Agent Era
Spectrum-X uses smart congestion control, adaptive routing, and AI telemetry. This setup offers fast, dependable connections between data centers. It supports massive AI superfactories. These factories span cities, countries, and continents. The platform works seamlessly with NVIDIA GPUs, CPUs, NVLink™, and software. This setup ensures great performance, from computing to networking. It changes how we view the costs and abilities of AI-scale networking.
日本語