At Xcelerated Compute 2026, the UALink Consortium joined industry leaders to advance an important conversation shaping the future of AI infrastructure: how open, scale-up networking will support the next generation of AI workloads.
Xcelerated Compute provided a valuable forum for our representatives to engage with the broader AI infrastructure community and share progress on the evolution of UALink technology and solutions. UALink Consortium representatives spoke with attendees on the importance of enabling open, optimized, high-performance scale-up connectivity to support transformative AI workloads.
During the UALink Consortium’s special presentation “Building the Open Scale-Up Fabric for AI”, Kurtis Bowman, UALink Board Chair, AMD, highlighted the growing importance of open standards in addressing the performance, scalability, and ecosystem flexibility demands of AI deployments. The discussion reinforced why scale-up interconnects have become a focal point of AI system architecture as model sizes grow and infrastructure requirements intensify.
The conversation explored how UALink technology is purpose-built to provide that foundation and deliver an open, high-performance scale-up fabric while enabling innovation at higher layers of the stack. Attendees also learned how UALink technology is advancing, building an ecosystem around interoperable accelerator-to-accelerator communication and supporting a path toward more flexible and scalable AI systems. You can view the session recording on the UALink Consortium YouTube channel.
Engagement at the UALink session and at our kiosk reinforced the market’s recognition that open scale-up networking is becoming an increasingly important part of future AI system design. Thank you to everyone who joined us in New York and contributed to the discussion. We look forward to continuing the conversation at upcoming industry events, including OCP EMEA Summit, FMS, Hot Interconnects and AI Infra Summit, as we advance UALink technology as the open protocol for AI scale-up networking.