UALink

Ultra Accelerator Link (UALink) is an open specification for a die-to-die interconnect and serial bus between AI accelerators. It is co-developed by Alibaba, AMD, Apple, Astera Labs, AWS, Cisco, Google, Hewlett Packard Enterprise, Intel, Meta, Microsoft and Synopsys.^[1] The UALink Consortium officially incorporated as an organization and electronics industry consortium in 2024, for promoting and advancing UALink.

Its first specification will provide interconnectivity specifically for a scalable network. The initial 1.0 version 200Gbps UALink specification, is based on the IEEE P802.3dj 200 Gb/s (Ultra Ethernet) PHY Layer. Each system node is made up of a host and as many accelerators as needed, and is managed by one OS image. The accelerators can be connected to the host using a variety of interconnect protocols (CXL, PCIe, XGMI, CHI c2c, Infinity Fabric). UALink Switches (ULS) connect up to 1024 accelerators within an AI 'pod', where each Accelerator is assigned a unique 10-bit routing identifier. Each UALink Switch port connects to a distinct Accelerator.^[2] AMD Infinity Fabric is expected to be the main shared-memory protocol.

The specification was due to be available to Contributor Members in 2024.^[3] The 1.0 version (UALink_200) was released to the public in 2025. Members of the public can download an "evaluation copy" after providing contact information.^[4]