d-Matrix Adopts NVIDIA NVLink Fusion for Rack-Scale XPU Deployment

September 10, 2026 | Source: NVIDIA Blog

Tags: NVLink Fusion, NVIDIA, d-Matrix, XPU, AI inference, MGX, Spectrum-X

d-Matrix will integrate its Raptor XPUs into NVIDIA's NVLink Fusion platform, gaining 6th-gen NVLink's 3 TB/s per-XPU all-to-all bandwidth and 3x lower latency than Ethernet — letting the inference-focused chipmaker skip building rack-scale infrastructure from scratch.

Details

AI inference chipmaker d-Matrix announced its next-generation Raptor XPUs will connect to NVIDIA's infrastructure via NVLink Fusion, a platform that lets third-party silicon companies plug into NVIDIA's proven rack-scale AI factory without building networking, cooling, and supply-chain infrastructure independently.\n\nNVLink Fusion delivers 6th-generation NVLink connectivity: 3 TB/s per-XPU all-to-all bandwidth, 3x lower XPU-to-XPU latency compared to off-the-shelf Ethernet, and 10x higher packet rates. d-Matrix's Raptor XPUs will connect via NVLink scale-up and Spectrum-X scale-out networking, slotting into the MGX rack architecture NVIDIA deploys at AI factory scale.\n\nThe practical appeal for custom silicon companies: building a competitive XPU is already hard — deploying it at data-center scale (sourcing chips, validating scale-up networking, certifying rack architecture, managing cooling and power) historically adds years and cost. NVLink Fusion absorbs those layers, letting chipmakers focus on differentiated compute rather than infrastructure.\n\nd-Matrix joins a substantial partner roster that includes AWS, Intel, Arm, Fujitsu, Marvell, MediaTek, Ayar Labs, and Lightmatter. The platform supports Arm, x86, and RISC-V CPU architectures. CEO Sid Sheth cited 'soaring demand for inference' as the driver for the partnership.