Nvidia Reportedly Planning Investment in AI Inference Startup d-Matrix
Nvidia is reportedly planning a strategic investment in d-Matrix, a startup specializing in AI inference ASIC hardware. This move deepens an existing partnership following d-Matrix's earlier integration into Nvidia's hardware technology ecosystem. The potential investment highlights Nvidia's strategy of coopting specialized AI chip designers into its proprietary ecosystem rather than facing direct competition. By bringing dedicated inference XPUs into its interconnect standards, Nvidia reinforces its position as the central platform provider for enterprise AI data centers. Under their technical collaboration, d-Matrix's next-generation inference XPUs are designed to support Nvidia's NVLink Fusion ecosystem and adopt the MGX modular rack reference architecture. This compatibility enables third-party inference hardware to interface directly with Nvidia's high-speed interconnects and rack infrastructure.
## BACKGROUND
AI workloads are divided into model training and model inference, with inference demanding high memory bandwidth and low latency for real-time applications. NVLink Fusion is Nvidia's ecosystem initiative that allows semi-custom third-party chips to connect to NVLink interconnects, while Nvidia MGX provides a standardized modular architecture for data center rack designs.