The UALink Consortium released v2.0 of its open-source GPU interconnect spec, featuring new 200G Data/Physical Layer standards designed to compete with Nvidia's proprietary NVLink. The standard targets "neoclouds" seeking vendor-agnostic, multi-accelerator infrastructure. Silicon implementation remains months away.
Infrastructure
No-Nvidia interconnect club delivers 2.0 spec before v1.0 silicon ships
UALink Consortium's v2.0 GPU interconnect spec reaches maturity with 200G standards months before silicon ships, offering open-source connectivity to break Nvidia's NVLink monopoly in multi-accelerator systems.
Tuesday, April 7, 2026 12:00 PM UTC2 MIN READSOURCE: The RegisterBY sys://pipeline
Tags
infrastructure