NVIDIA Ethernet SuperNICs deliver powerful networking for AI factories and cloud data centers. Designed for network-intensive, massively parallel computing, these SuperNICs accelerate GPU-to-GPU communication, ensuring high-bandwidth, low-latency data transfers and enhancing AI performance and scalability. Integrated with NVIDIA Spectrum-X™, they form a critical component of modern AI infrastructure, driving unprecedented efficiency and speed in AI applications.
AI cloud service providers are embracing NVIDIA Ethernet SuperNICs within Spectrum-X—now broadly available from NVIDIA system partners—to bring extreme networking performance to their AI infrastructures.
The NVIDIA BlueField®-3 SuperNIC is an advanced network accelerator at speeds up to 400Gb/s, designed for accelerating hyperscale AI workloads while delivering deterministic, isolated performance, with secure cloud multi-tenancy.
The ConnectX®-8 SuperNIC offers up to 800Gb/s of total network bandwidth, transforming generative AI workloads and next-generation data center performance for massive-scale AI compute fabrics.