The Open Compute Project Foundation (OCP) and the Ultra Accelerator Link™ (UALink™) Consortium announced a strategic collaboration to integrate high-performance scale-up interconnects into next-generation AI and HPC clusters. The initiative will combine UALink’s recently launched 1.0 specification with OCP’s Open Systems for AI infrastructure development, aiming to deliver low-latency, high-bandwidth, and energy-efficient interconnects tailored for large-scale AI training and inference. With AI workloads pushing the limits of compute density and interconnect performance, the alliance seeks to drive faster deployment of open, interoperable system designs.
As hyperscale operators invest heavily in AI infrastructure, the collaboration ensures that UALink’s scale-up interconnect technologies are embedded within OCP’s open reference architectures and future data center blueprints. The joint effort will also intersect with OCP’s Future Technologies Initiative, including its Short-Reach Optical Interconnect workstream, enabling new approaches to manage bandwidth bottlenecks in AI clusters. With backing from industry leaders such as AMD, Intel, Microsoft, Meta, AWS, and Google, the UALink Consortium positions itself as a foundational layer in the evolving AI infrastructure stack.
- OCP and UALink will jointly promote open, high-performance scale-up interconnects for AI and HPC workloads
- The partnership integrates UALink 1.0 specification into OCP’s Open Systems for AI Strategic Initiative
- UALink addresses demands for ultra-low latency and massive bandwidth in accelerated compute environments
- Collaboration aligns with OCP’s Future Technologies Initiative, including optical interconnect workstreams
- UALink Consortium includes founding members such as AMD, Intel, Microsoft, Meta, Google, AWS, and Cisco
- Goal is to advance interconnect bandwidth and scalability for generative AI and LLM infrastructure
- OCP and UALink aim to support the rapid evolution of open, sustainable AI-capable data center systems
“The rapid adoption of AI across industries… has created a pivotal moment for data center investments. By collaborating, the UALink Consortium and the OCP Community can shape system specifications to address critical challenges in interconnect bandwidth and scalability posed by advanced AI models,” said George Tchaparian, CEO, Open Compute Project Foundation.