Nvidia, MediaTek Bring Custom Chips to AI Racks

Nvidia is making its rack-scale AI infrastructure available to companies that want to build custom AI accelerators, rather than relying solely on Nvidia's own processors. This partnership with MediaTek allows customers to develop custom accelerators using Nvidia's NVLink technology for connectivity while retaining control over their own chip designs. The move matters because major cloud providers are already building their own processors for AI workloads, and this approach lets them do so without having to rebuild their entire AI infrastructure from scratch. The technology competes with an open-standard alternative called UALink, which is being developed by a group of chip and infrastructure companies as another way to connect accelerators at scale.