- Apple has joined the board of the Ultra Accelerator Link consortium
- The link is a key technology that binds GPUs, not unlike synapses on neurons
- UALink emerges as the biggest rival to Nvidia’s proprietary NVLink
Back in June 2024, we reported how a number of big tech names had joined forces to form the Ultra Accelerator Link (UALink) Promoter Group, a strategic move aimed at reducing Nvidia’s dominance in the AI accelerator market.
UALink competes directly with Nvidia’s proprietary NVLink technology and seeks to develop a new industry standard for high-speed, low-latency communications for scale-up AI systems in data centers. It already has support from Intel, AMD, Google, Microsoft, Meta, HPE, Cisco and Broadcom, but now Apple has also joined the UALink board.
“UALink shows great promise in addressing connectivity challenges and creating new opportunities to expand AI capabilities and requirements,” said Becky Loop, Director of Platform Architecture at Apple. “Apple has a long history of pioneering and collaborating on innovations that drive our industry forward, and we are excited to join the UALink board.”
Project ACDC
By joining the consortium, it appears that Apple plans to use UALink technology in “Project ACDC” (Apple Chips in the Data Center), also known as “Baltra”.
This rumored initiative, in collaboration with TSMC and Broadcom, aims to develop proprietary AI chips for Apple’s data centers, boosting the capabilities of its new Apple Intelligence.
Apple isn’t the only new company joining the consortium; additional supporters include Alibaba Cloud Computing and Synopsys.
Expected to arrive in the first quarter of 2025, the UALink 1.0 specification will enable up to 200 Gbps per orbit scale-up connection for up to 1,024 accelerators in an AI pod.
“We are pleased to welcome Alibaba, Apple and Synopsys to the UALink Consortium Board of Directors,” said Kurtis Bowman, UALink Consortium Board Chair. “Since our incorporation, the consortium has grown to more than 65 total members, spanning cloud, silicon and IP providers, software companies, system OEMs and others. The continued support of the consortium will help accelerate the adoption of this important industry standard and define next-generation interconnect for AI workloads.”
