Meta and Oracle are adopting NVIDIA’s Spectrum-X Ethernet switches to power giga-scale AI data centers, marking another expansion of NVIDIA’s footprint in hyperscale networking. The two companies plan to standardize on Spectrum-X as an open, accelerated Ethernet architecture to increase efficiency, predictability, and scalability across massive GPU deployments.
Meta will integrate Spectrum Ethernet into its Facebook Open Switching System (FBOSS) and the Minipack3N switch platform to manage AI networks at unprecedented scale. Oracle, meanwhile, will use Spectrum-X to interconnect millions of GPUs within its giga-scale AI factories built on NVIDIA’s Vera Rubin architecture. Both aim to shorten training times for trillion-parameter models and enhance performance for generative and reasoning AI workloads.
NVIDIA said its Spectrum-X Ethernet delivers up to 95% data throughput at AI cluster scale — far exceeding standard Ethernet’s typical 60% efficiency — by leveraging advanced congestion control, adaptive routing, and AI-driven telemetry. The technology extends across data centers and geographies through Spectrum-XGS, linking multiple AI clusters into unified “AI super-factories.”
• Meta integrates Spectrum-X into FBOSS and Minipack3N for open, large-scale AI networking
• Oracle deploys Spectrum-X to connect Vera Rubin–based AI factories across regions
• Spectrum-X delivers 95% throughput efficiency using adaptive routing and congestion control
• NVIDIA positions Spectrum-X as part of a full-stack AI fabric spanning compute, interconnect, and telemetry
“Trillion-parameter models are transforming data centers into giga-scale AI factories, and industry leaders like Meta and Oracle are standardizing on Spectrum-X Ethernet to drive this industrial revolution,” said Jensen Huang, founder and CEO of NVIDIA.
🌐 Analysis: The move solidifies NVIDIA’s growing presence in Ethernet-based AI networking, complementing its InfiniBand leadership. Meta’s integration of Spectrum-X into FBOSS reflects a growing alignment between open networking and accelerated fabrics. Oracle’s choice reinforces its partnership with NVIDIA for Vera Rubin architecture AI factories, echoing similar hyperscaler adoption trends seen across AWS, Microsoft, and Google Cloud.
