InfiniBand (NVIDIA Quantum)
Low-latency scale-out fabric for GPU clusters
InfiniBand, dominated by NVIDIA's Quantum switches (from the Mellanox acquisition), is the incumbent scale-out fabric connecting GPU servers into large training clusters via RDMA with very low latency and in-network computing (SHARP). Quantum-X800 delivers 800Gb/s per port for Blackwell-era clusters. It competes with Ethernet-based AI fabrics (Spectrum-X, Ultra Ethernet), which are gaining share for cost and openness at hyperscale.
Owner
NVIDIA (via Mellanox)
Quantum-X800
800Gb/s per port
Edge
RDMA + SHARP in-network compute
How it fits the stack
InfiniBand (NVIDIA Quantum) with what it depends on (above) and what it feeds (below). The figure renders as a crawlable diagram and upgrades to an interactive 3D graph as it scrolls into view.
InfiniBand (NVIDIA Quantum) in the AI stack. InfiniBand (NVIDIA Quantum) with its immediate upstream dependencies (top) and downstream dependents (bottom) in the AI value chain. Hover a node in 3D, or read the full relationships below.
Graph data (text) — 6 entities, 6 relationships
- CoreWeave Data Centers —uses→ InfiniBand (NVIDIA Quantum)
- NCCL / RCCL collectives —depends on→ InfiniBand (NVIDIA Quantum)
- Nvidia —used by→ InfiniBand (NVIDIA Quantum)
- Nvidia —uses→ InfiniBand (NVIDIA Quantum)
- InfiniBand (NVIDIA Quantum) —competes with→ Arista Networks
- InfiniBand (NVIDIA Quantum) —competes with→ Ultra Ethernet / UALink consortia