All three sources report that d-Matrix, an AI inference chipmaker, will adopt NVIDIA's NVLink Fusion for its upcoming Raptor XPUs. The move lets d-Matrix's accelerators connect to NVIDIA's rack-scale infrastructure, using NVLink for scale-up and Spectrum-X for scale-out. ServeTheHome and NVIDIA both describe this as joining a growing roster of ecosystem partners.
The Next Platform frames the same deal as pairing Raptor with "Nvidia Rackscale Iron," while NVIDIA's blog emphasizes a "rack-scale XPU deployment." The terminology differs slightly, but the substance is identical: d-Matrix's memory-based XPUs will interoperate with NVIDIA's networking fabric, allowing them to be deployed alongside NVIDIA hardware in large AI systems.
This is notable because d-Matrix's Raptor is designed for inference, not training. By hooking into NVLink Fusion, the startup can offer a complementary option in NVIDIA-dominated data centers without building its own interconnect stack. The sources agree on the core facts and offer no conflicting details.