Nvidia to Invest in Inference Rival d-Matrix via NVLink Fusion

Nvidia plans to invest in d-Matrix, the Santa Clara inference-chip startup that a month ago agreed to plug its Raptor processors into Nvidia's rack-scale gear through NVLink Fusion. The Information reported the plan; neither company has disclosed a dollar amount, timing, or the valuation Nvidia is accepting.

d-Matrix was last marked at a $2 billion valuation when it closed a $275 million Series C in November 2025. It has raised about $500 million since 2019, and in July The Information separately reported it was out raising again at a $5 billion target. The company's pitch is that its chips can run inference workloads "10 times faster and using five times less energy than a standalone graphics processing unit from Nvidia." That is the claim Nvidia is now writing a check into.

The pattern is familiar. Nvidia put $2 billion into Marvell and $3.5 billion into MediaTek earlier this year, both tied to NVLink Fusion. d-Matrix joins Qualcomm, Arm, Marvell, Amazon, Fujitsu, and MediaTek as what The Register called "NVLink true believers." The common thread: every NVLink Fusion platform must include at least one Nvidia product, whether a CPU, GPU, or switch, "ensuring Nvidia maintains revenue even as competitors design custom chips."

First d-Matrix Raptor systems inside Nvidia MGX racks are expected in Q4 2027, with "up to 144 Raptor accelerators within a single all-to-all NVLink domain." The money arrives well before the product does.

It is the second Nvidia-into-a-rival story on our radar today, after reported talks to deepen its stake in Reflection AI.