What d-Matrix does
d-Matrix is a private semiconductor and systems company building digital in-memory compute (DIMC) technology for generative AI inference in data centers. The company’s stated product direction is an end-to-end inference platform that combines Corsair inference accelerators, JetStream networking/I-O acceleration, and Aviator software, with rack-scale reference designs such as SquadRack to integrate into standard data-center environments. The core technical theme is compute–memory integration on silicon (DIMC) using chiplet-style scaling and high-bandwidth die-to-die / chiplet interconnects, with the goal of reducing inference latency, improving tokens-per-second and performance-per-watt, and improving total cost of ownership versus GPU-centric approaches for production inference workloads.
Go-to-market positioning emphasizes serving hyperscale and enterprise AI infrastructure buyers who need cost and energy efficiency at inference time, including low-latency “interactive” and agentic use cases. d-Matrix’s products are described as scaling from cards to racks using standard fabrics, and the company has increasingly broadened its “platform” scope via acquisitions aimed at deployment/orchestration software (Wallaroo.ai) and rack-scale systems engineering (GigaIO’s data center business). As of 2026, d-Matrix announced that its Corsair inference accelerator platform is in full production with shipping in volume to priority customers, and it has published multiple partnerships and integration announcements with inference service providers and cloud environments (e.g., Parasail, Gimlet Labs).