Apple has joined the Ultra Accelerator Link Consortium, which is creating a new connection standard for AI accelerators in data centers. This standard will be an analogue of Nvidia's proprietary NVLink technology.
The Ultra Accelerator Link Consortium is developing a standard called UALink that will unify the AI accelerator chips that are increasingly found in server computers. Apple is now a board member of the consortium, along with Alibaba and semiconductor company Synopsys.
In a statement, Apple's director of platform architecture, Becky Loup, said that UALink “shows great promise” in solving’the challenges of connectivity and creating new opportunities for AI.
«Apple has a long history of pioneering and collaborating to drive innovations that drive our industry forward. We are excited to join the board of directors of UALink», — Loop said in a separate statement.
UALink — What is it and how does it impact the AI market?
UALink aims to connect chips ranging from GPUs to custom solutions to accelerate the training, fine-tuning, and operation of AI models. The first UALink products are expected to be released over the next few years based on open standards, including AMD Infinity Fabric.
Intel, AMD, Google, AWS, Microsoft, and Meta are members of the UALink consortium. Nvidia, the largest maker of AI accelerators, is not. Instead, Nvidia offers its own NVLink interconnect technology for connecting chips in a data center cluster.