The Ultra Accelerator Link Consortium (UALink), an industry organization taking a collaborative approach to advance high-speed interconnect standards for next-generation AI workloads, announced an expansion to its Board of Directors, welcoming Alibaba Cloud, Apple, and Synopsys – joining existing member companies like AMD, AWS, Cisco, Google, HPE, Intel, Meta, and Microsoft.
The UALink specification will establish an open industry standard, enabling AI accelerators and switches to communicate more effectively and expand memory access to meet the requirements of large AI models.
The addition of these new board members illustrates the growing momentum behind UALink’s efforts to address the connectivity demands of modern AI infrastructure. It’s also a direct response to NVIDIA’s increasing dominance in the space.
Announcement Details
Since its incorporation in October 2024, UALink has grown to include over 65 member companies across sectors as diverse as cloud services, silicon design, and software development.
Adding Alibaba Cloud, Apple, and Synopsys to the effort brings together a diverse group of companies with expertise in cloud computing, silicon design, and interface IP solutions. By joining forces, these new members will help drive the development and adoption of UALink technology.
The addition of Alibaba Cloud brings a wealth of expertise in cloud computing and applications to the consortium. The announcement quotes Qiang Liu, VP of Alibaba Cloud and GM of Alibaba Cloud Server Infrastructure: “UALink shows great promise in addressing connectivity challenges and creating new opportunities for expanding AI capabilities and demands. By defining core needs and solutions from the perspective of cloud computing and applications, we can build a more competitive intelligent computing supernode ecosystem.”
Apple’s Director of Platform Architecture, Becky Loop, echoed this sentiment in the release, stating that UALink “shows great promise in addressing connectivity challenges and creating new opportunities for expanding AI capabilities and demands.” Apple has a long history of pioneering innovations that drive the industry forward, and its involvement will undoubtedly bring significant expertise to the consortium.
Asa leading provider of best-in-class interface IP solutions, Synopsys is committed to contributing its expertise to the UALink Consortium. According to Richard Solomon, UALink Board Member and senior staff Product Manager, “UALink will be critical in addressing the performance and bandwidth communication demands of hyperscale data centers, enabling the high-speed interconnects needed to scale up pods and clusters.”
UALink vs NVLink
NVIDIA’s NVLink is well-established for high-bandwidth interconnects in GPU-based AI systems. NVLink is tightly integrated with NVIDIA’s hardware ecosystem, delivering optimized platform performance. UALink, by contrast, seeks to establish an open standard that enables interoperability across a broader range of hardware and software solutions.
UALink’s open approach will benefit organizations seeking flexibility in their AI infrastructure, particularly those employing diverse hardware environments. Today, nearly every public cloud provider and on-prem server OEM supports a diversity of AI approaches.
The top CSPs, for example, are all developing their own AI accelerators that are not based on NVIDIA technology. The OEM market also sees OEM diversity with accelerators from suppliers like AMD, Qualcomm, and Cerebus.
Analysis
The UALink Consortium is on track to release the UALink 1.0 specification for general industry review within the first quarter of 2025. Its specification aims to enable up to 200 Gbps per lane scale-up connections, supporting configurations of up to 1,024 accelerators within an AI pod.
The Consortium’s contributing members have had access to the specification since late 2024, allowing them to provide feedback and prepare for implementation. The upcoming public release will enable broader industry participation and adoption, fostering collaboration and innovation in AI infrastructure.
The UALink Consortium’s announcement shows the momentum and importance of having an open, high-performance interconnect standard for AI infrastructure. With the UALink 1.0 Specification coming soon, its success will depend on industry adoption and its ability to deliver on performance and scalability.
The growing list of consortium participants shows that adoption isn’t likely to be a challenge. It also illustrates the pent-up appetite for a solution that addresses networking challenges within AI infrastructure. We’re looking forward to seeing where this effort leads.
