d-Matrix to integrate inference XPUs with Nvidia GPUs via NVLink Fusion

d-Matrix CEO Sid Sheth said the startup has formed a partnership with Nvidia that will integrate its inference-focused XPUs with Nvidia GPUs using NVLink Fusion, according to a Bloomberg Technology interview. The arrangement places d-Matrix's inference accelerators inside Nvidia's interconnect ecosystem as demand shifts toward specialized systems built for inference rather than training. Nvidia is embracing the rise of custom AI chips, and Sheth described inference as emerging as the next major battleground in AI compute. The segment did not disclose commercial terms, deployment volumes, or a timeline for the integration.