NVIDIA networking hits $11B as Spectrum-X, NVLink lead

NVLink, Spectrum-X Ethernet, InfiniBand ramp drove $11B networking quarter

NVIDIA’s networking segment generated about $11 billion in quarterly revenue, rising roughly 267% year over year, according to CRN (https://www.crn.com/news/ai/2026/nvidia-hits-record-quarterly-growth-says-it-s-the-world-s-largest-networking-business).CRN also reports the quarter was propelled by the continued ramp of NVLink as the compute fabric for Grace‑Blackwell GB200/GB300 rack‑scale platforms, plus growth in Spectrum‑X Ethernet and the Quantum InfiniBand portfolio.

Why NVIDIA’s networking surge matters for AI data centers

The results signal a structural shift toward AI “factory” designs in which compute, fabric, and I/O are procured as an integrated system rather than piecemeal switches or NICs. The networking attach rate is now close to 90%, as analyzed by The Motley Fool (https://www.fool.com/investing/2026/01/15/90-of-nvidias-customers-now-buy-this-and-its-not-g//).Leadership messaging reflects this pivot from discrete switches to AI‑optimized fabrics. “Spectrum‑X is a ‘home run’,” said Jensen Huang, CEO of NVIDIA.

For incumbent switch vendors, spend is consolidating around AI‑optimized fabrics and rack‑scale systems, which influences procurement priorities, interoperability testing, and lifecycle planning. Integration between compute, fabric, and acceleration is becoming a primary selection criterion.Cisco has expanded its partnership with NVIDIA to integrate Silicon One with Spectrum‑X and SuperNIC technology to deliver unified, AI‑ready networking architectures, according to Cisco’s newsroom (https://newsroom.cisco.com/c/r/newsroom/en/us/a/y2025/m02/cisco-expands-partnership-with-nvidia-to-accelerate-ai-adoption-in-the-enterprise.html).For buyers, this suggests architecture choices will hinge on workload patterns, fabric integration, and operational model rather than standalone box metrics. Supply timing and qualification cycles remain practical constraints in near‑term deployments.

AI fabric architecture beyond GPUs: Ethernet, InfiniBand, NVLink

AI data centers now pair accelerators with dedicated fabrics. Spectrum‑X defines NVIDIA’s Ethernet approach; Quantum represents its InfiniBand platform; NVLink provides the compute fabric connecting accelerators in rack‑scale GB200/GB300 designs.BlueField DPUs and SuperNICs sit alongside these fabrics within server and rack systems. The outcome is a more integrated, factory‑like deployment model intended to reduce bottlenecks between training and inference stages.

When to choose Spectrum-X Ethernet versus InfiniBand in deployments

Organizations anchored in Ethernet operations, tooling, and skills may favor Spectrum‑X to align with existing data‑center practices and L2/L3 ecosystems.Environments prioritizing tightly integrated AI fabrics often maintain InfiniBand‑based clusters via NVIDIA’s Quantum platform, especially where established workflows and infrastructure are already in place.

NVLink’s role with GB200/GB300 and BlueField DPUs

As noted above, NVLink functions as the compute fabric for GB200/GB300 rack‑scale platforms, tying accelerators together within the system design.BlueField DPUs complement this within server nodes, while the Ethernet or InfiniBand cluster fabric carries inter‑node traffic across the AI cluster.

FAQ about Spectrum-X Ethernet

Did NVIDIA actually surpass Cisco in networking, and in which specific segments or metrics did this occur?

Based on NextPlatform/IDC data (https://www.nextplatform.com/2025/06/23/nvidia-passes-cisco-and-rivals-arista-in-datacenter-ethernet-sales/), NVIDIA topped Cisco in select datacenter Ethernet metrics during the quarter; this pertains to segment-level comparisons, not overall company revenue.

How do Spectrum-X Ethernet, NVLink, and InfiniBand work together in AI datacenter design, and when should each be used?

Benzinga’s analyst notes (https://www.benzinga.com/analyst-ratings/analyst-color/25/02/44005208/nvidia-ai-juggernaut-rolls-on-but-can-arista-cisco-keep-up) describe a transition to NVLink 72 and Spectrum‑X alongside Quantum InfiniBand; use NVLink for GPU fabric, Spectrum‑X for Ethernet clusters, InfiniBand for specialized AI.