D-MATRIX ADOPTS $NVDA NVLINK FUSION FOR RACK-SCALE AI INFERENCE AI chipmaker d-Matrix will connect its next-generation Raptor XPUs to NVIDIA’s AI infrastructure platform using NVLink Fusion, giving its custom inference silicon a direct path into NVIDIA MGX-based AI factories. The integration includes NVLink scale-up networking, Spectrum-X Ethernet, Vera CPUs, ConnectX-9 SuperNICs and BlueField-4 DPUs, while allowing d-Matrix systems to operate alongside NVIDIA GPU platforms such as Vera Rubin NVL72. NVIDIA says NVLink Fusion can deliver 3x lower XPU-to-XPU latency than standard Ethernet, 10x higher packet rates and up to 3 TB/s of all-to-all bandwidth per XPU. The bigger goal is to let custom chipmakers focus on their own processors while using NVIDIA’s rack, power, cooling, networking, software and supply-chain infrastructure to deploy at scale faster.