Home » Resources » Blog » Advancing AI with AMD: Delivering Purpose-built Connectivity for Scale-up Architecture with UALink

Advancing AI with AMD: Delivering Purpose-built Connectivity for Scale-up Architecture with UALink

Mike Hendricks, AVP, Solutions & Ecosystem

Back in 2017, the founders of Astera Labs identified a looming bottleneck in data center infrastructure. While AI models were evolving at an unprecedented pace, traditional connectivity infrastructure wouldn’t keep up. From the beginning, Astera Labs set out with a clear mission: to deliver purpose-built connectivity solutions for the entire AI rack.

Yesterday, we were honored to join AMD on the keynote stage at AMD Advancing AI event – a showcase of ecosystem collaboration powering the next era of AI. Together, we highlighted how UALink™ connectivity solutions are critical for enabling open, scale-up architectures for AI.

As we’ve discussed in previous blogs, UALink addresses fundamental AI system challenges—delivering low latency, high bandwidth, and memory semantic communication while also driving down power consumption and total cost of ownership (TCO). This open, scale-up connectivity standard is a key enabler of the composable infrastructure AI demands—and exactly the kind of innovation the AMD Advancing AI event was all about.

Astera Labs is proud to be at the forefront of this transformative moment. As a UALink Consortium Board Member, Astera Labs is at the forefront in helping to define and drive the UALink specification to meet the performance, scalability, and openness that the AI era demands.  

During AMD’s Advancing AI keynote, Astera Labs CEO, Jitendra Mohan, shared why we are excited about UALink technology. I can break it down into three key points:

  1. Open. UALink is the only open standarddesigned specifically for scale-up applications. This means that it is supported by a broad ecosystem of vendors, providing our customers with more choice and flexibility.
  2. Simple. UALink’s architecture is inherently simple. It is not burdened with excess overhead of Ethernet-based approaches. UALink is designed to be efficient, fast and robust – it combines the best of many protocols.
  3. Optimized. UALink is purpose-built for AI scale up and completely aligns with Astera Labs’ mission, expertise and roadmap.

A full connectivity portfolio of smart UALink solutions

Astera Labs is already at the forefront of scale-up connectivity innovations with our Scorpio-X Series Fabric Switches, and our Aries Smart Retimers. As a trusted partner for our hyperscale customers, we have a close-up view of the features and timeframes needed to realize their vision of deploying UALink architectures. Our vision is to provide a complete infrastructure for the entire AI rack. This includes purpose-built silicon, hardware, and software to support AI platforms based on custom ASICs and merchant GPUs including AMD’s Instinct solutions. Working closely with AMD and XPU partners, we plan to offer a complete portfolio of UALink products to support deployment at scale, including smart fabric switches, signal conditioners, controllers, and more.

And of course, all of Astera Labs’ products are supported by our COSMOS software suite that provides an unparalleled view into the health of the entire rack. To ensure seamless deployment and accelerate time-to-market for our customers, our solutions undergo rigorous interoperability testing with a broad array of devices in our Cloud-Scale Interop Lab. This robust validation environment ensures interoperability at rack-scale.

Open collaboration for the AI era

In her closing remarks of the Advancing AI keynote, AMD CEO Lisa Su shared “The future of AI is not going to be built by any one company or in a closed ecosystem. It’s going to be shaped by open collaboration across the industry. It’s going to be shaped because everyone is bringing their best ideas. and it’s going to be shaped because we’re innovating together.”

We couldn’t agree more. By working with AMD and our industry partners on UALink technology, we are driving the next wave of AI innovation for scale-up architecture.

Resources:

Share:

Related Articles