Nvidia Teams Up with d-Matrix on Custom AI Chips

Nvidia is expanding its focus on AI inference by partnering with startup d-Matrix. The partnership will integrate d-Matrix's inference-focused XPUs with Nvidia GPUs using NVLink Fusion, enabling faster and more efficient processing of AI workloads.

d-Matrix CEO Sid Sheth discussed the collaboration on 'Bloomberg Tech', emphasizing the growing importance of specialized systems for AI inference. According to Sheth, inference is becoming a key battleground in AI compute as demand shifts toward custom chips optimized for specific tasks.