Apple Joins AI Hardware Standard Consortium To Improve Server Performance

Apple has joined the board of directors of Ultra Accelerator Link Consortium (UALink) (photo: x @Consomac)

JAKARTA- Apple has joined the board of directors of Ultra Accelerator Link Consortium (UALink), an industry standard group open to developing UALink specifications. This move allows Apple to exert influence on how the AI server infrastructure architecture will develop.

UALink is a high-speed accelerator interconnection technology that focuses on developing technical specifications for high bandwidth communication between processing units, such as GPUs or AI chips. In other words, this technology allows some GPUs or AI chips to communicate quickly and work together as if they are one large chip.

UALink now consists of more than 65 member companies, including Alibaba and Synopsys, who have also recently joined this consortium with Apple. "UALink shows great potential in overcoming connectivity challenges and creating new opportunities to expand AI's capabilities and demand," saidten Loop, Platform Architecture Director at Apple.

The chairman of the UALink Consortium Council, Kurtis Badminton, welcomed the addition of the new member: "Continuous support for the Consortium will help accelerate adoption of this industry standard, which defines next-generation interconnections for AI's workload."

Benefits for Apple

Apple sees the huge benefits of its involvement in UALink, especially in developing high-performance AI servers. UALink allows connections up to 200Gbps per lane and can connect up to 1,024 accelerators in one AI pod.

With this technology, Apple can optimize how chips in their systems-on-chips work, which will improve performance and benefit users directly.

By joining Apple into UALink, the company demonstrates a strong commitment to encouraging innovation in AI technology and improving its infrastructure performance. This participation also helps open up new opportunities in the development of industry standards that are more efficient and effective for the future of AI.