Apple has officially joined the Ultra Accelerator Link Consortium (UALink), a group of over 65 members working on next-generation AI accelerator architecture. This move allows Apple to influence the new standard and promote its adoption, effectively creating a framework for connecting multiple GPUs in data centers powering AI tasks.
The UALink consortium aims to compete with Nvidia’s dominance in AI hardware by establishing an open standard for scale-up connections of many AI accelerators. This initiative seeks to challenge Nvidia’s proprietary NVLink technology, which relies on Infiniband communication technology.
Backed by AMD and Intel, the UALink Consortium includes prominent members such as AWS, Google, Meta, Microsoft, and now Apple. The consortium has released its initial 1.0 version of the 200Gbps UALink specification, enabling the connection of up to 1K accelerators within an AI pod.
The specification is based on the IEEE P802.3dj PHY Layer and will be available to contributor members in 2024, with public release planned for the first quarter of 2025. With Apple’s involvement, the consortium aims to drive innovation and expand AI capabilities, creating new opportunities for industries to adopt this technology.
Source: https://9to5mac.com/2025/01/14/apple-joins-ualink-group-tasked-with-taking-on-nvidias-ai-hardware-dominance