d-Matrix moves its inference chips onto NVIDIA's NVLink Fusion

Inference chipmaker d-Matrix said it will connect its next-generation Raptor processors to NVIDIA's infrastructure through NVLink Fusion. The move lets d-Matrix use NVIDIA's MGX rack architecture, liquid-cooled designs, Spectrum-X networking and supply chain for its own silicon. The company plans to run its racks alongside NVIDIA Vera Rubin NVL72 systems for disaggregated inference, and to integrate Vera CPUs, ConnectX-9 SuperNICs and BlueField-4 DPUs. NVLink Fusion opens the same rack to non-NVIDIA processors, so a data center can be built once and host GPUs, CPUs and custom chips together. NVIDIA says NVLink Fusion lets chip companies focus on their own processors while using NVIDIA infrastructure to deploy them, which both firms describe as a faster, lower-risk path to large-scale deployment.