AI inference chipmaker d-Matrix is plugging its forthcoming Raptor accelerators into Nvidia’s NVLink Fusion ecosystem, extending Nvidia’s reach beyond its own GPUs.
According to d-Matrix the multi-year deal will bring d-Matrix XPUs into Nvidia’s rack-scale AI infrastructure, with Raptor connecting directly to the MGX architecture using high-bandwidth, low-latency NVLink Fusion.
Systems will combine Raptor with Nvidia Vera CPUs, NVLink switches, BlueField-4 DPUs, ConnectX-9 SuperNICs and Spectrum-X Ethernet networking, with Astera Labs providing additional connectivity.
For d-Matrix, building an accelerator is difficult enough without recreating Nvidia’s racks, networking, cooling and supply chain from scratch. For Nvidia, NVLink Fusion means companies can use rival accelerators while Nvidia still supplies much of the plumbing holding the system together.
d-Matrix founder and chief executive Sid Sheth said: “Demand for inference is soaring, but capital, time and energy remain finite. With NVLink Fusion and MGX, we can integrate our Raptor XPUs into a broadly deployed, liquid-cooled architecture, giving customers a faster, lower-risk path to deploy and scale ultralow-latency inference.”
Raptor targets latency-sensitive workloads including coding assistants, real-time chatbots and voice agents. Nvidia GPUs could handle compute-heavy prefill work while d-Matrix processors tackle the latency-sensitive decode stage.
The chip uses a memory-centric architecture with 3D DRAM stacking, combining a DRAM memory chip with an SRAM compute chip to reduce latency and the energy wasted shifting data around.
Nvidia founder and chief executive Jensen Huang said: “NVLink Fusion enables partners to integrate custom silicon with NVIDIA’s deep ecosystem of NVLink, advanced packaging, rack-scale systems and networking technologies.” Raptor is expected to tape out before the end of 2026.
The first Raptor XPUs integrated into Nvidia MGX racks are expected in the fourth quarter of 2027, giving accelerator rivals an easier route into large AI installations while making Nvidia’s infrastructure increasingly difficult to avoid.
