Sid Sheth, d-Matrix | Robotics & AI Infrastructure Leaders
AI Summary
This video is an in-depth interview from theCUBE Robotics and AI Leaders program featuring Sid Sheth, CEO of d-Matrix, discussing their Corsair generative AI inference platform. They cover the market demand shift from training massive AI models to cost-efficient inference, with d-Matrix’s focus on efficiency, integrating compute and memory via innovative in-memory computing chiplets. The Corsair card is designed for seamless data center integration without infrastructure changes, targeting applications like interactive large language models (LLMs), agentic reasoning models, and real-time interactive video generation. The company roadmap aims to scale with memory stacking on compute, focusing on achieving low latency and high efficiency tailored for inference workloads. Sid emphasizes the advantage of new numerical formats and co-location of compute and memory to maximize energy efficiency and performance in AI inference deployments.