Networking for AI and data centers

NVIDIA Networking equipment

We supply NVIDIA switches, network adapters, DPUs, cables and transceivers for AI clusters, high-performance computing, cloud platforms and data centers. We can configure the complete compatible network solution.

NVIDIA equipment

Choose the right equipment

Compare the available models or send your requirements for a tailored recommendation.

Quantum InfiniBand switches
InfiniBand

Quantum InfiniBand switches

Use casesAI / HPC clusters
PortsQSFP / OSFP
Parameterslatency, bandwidth
Documentationofficial

For large GPU clusters, switch generation, port configuration and cabling must be verified together.

Spectrum Ethernet switches
Ethernet

Spectrum Ethernet switches

Use casesAI Ethernet fabric
Speed100/200/400/800G
Parametersswitching capacity
OSCumulus / SONiC

Spectrum switches for AI, cloud, storage and leaf-spine Ethernet fabrics.

ConnectX Ethernet adapters
Network adapters

ConnectX Ethernet adapters

FamilyConnectX
InterfacePCIe
Speedup to 400/800G by SKU
Key checkfirmware

For servers and GPU nodes: verify PCIe, port, optics or cable, RoCE and drivers.

ConnectX InfiniBand adapters
Network adapters

ConnectX InfiniBand adapters

Use casesHPC / AI fabric
Portssingle/dual
Key checkHCA generation
Documentationofficial

Sourced against Quantum fabric compatibility and the exact server platform.

BlueField DPU
DPU

BlueField DPU

Functionsnetwork/storage/security
Use casesdata center offload
Key checkDOCA
VerificationSKU + firmware

DPU option for CPU offload, network security and infrastructure services.

LinkX cables & transceivers
Cables

LinkX cables & transceivers

Speeds100/200/400/800G
TypesDAC/AOC/optics
Key checklength + port
Verificationcompatibility

The cabling layer often determines whether a networking shipment is actually usable.

Accurate configuration

  • We configure switches, adapters, cables and transceivers together for the required ports, speed and connection lengths.
  • For AI clusters, we design around the server count, target bandwidth, topology and redundancy requirements.

What we need for pricing

Send the models, quantities, delivery country and preferred timeline. If the exact configuration is not yet defined, describe the application and we will recommend the right solution.

Tailored commercial offer

Get NVIDIA pricing and delivery terms

Send a model list or project specification. If the exact model is not yet selected, describe your application and we will recommend a suitable configuration.