Sid Sheth, d-Matrix | AI Infrastructure Silicon Valley - Executive Series
summary
Company builds inference-focused silicon and accelerator cards that target generative AI by combining four compute engines — a licensed RISC-V data mover, an in-memory compute engine for the ~50–70% matrix-math workload, and two transformer-specific transpose engines — aiming to cut power and cost vs GPUs. Product launches in H2, sells as server cards to cloud, sovereign and specialized data centers, supports low-latency always-on inference now and planned fine-tuning later.
AI-written summary · about
